WO2025238536A1

WO2025238536A1 - Method for creating a database for a continuous cycle production line for decorated ceramic products and computer server containing the database

Info

Publication number: WO2025238536A1
Application number: PCT/IB2025/054991
Authority: WO
Inventors: Paola GATTI
Original assignee: Sacmi Imola SC
Current assignee: Sacmi Imola SC
Priority date: 2024-05-14
Filing date: 2025-05-13
Publication date: 2025-11-20
Anticipated expiration: 2026-11-14

Abstract

Described is a method for making a database for a line (100) for the production of decorated ceramic products in a continuous cycle comprising: a plurality of processing units, configured for performing respective processes in succession, wherein each processing unit includes a system of sensors and a plurality of operating configurations, the method comprising the following steps: preparing a non-volatile memory and a processor; using the processor, storing in the memory values of adjustment parameters, the operating configurations of the processing unit being determined by values of the adjustment parameters, input parameter values, at least some of which being detected by the system of sensors, performance parameter values, representing performance of the processing unit, and time stamps for temporally synchronising the values of the adjustment parameters, input parameters and performance parameters.

Description

DESCRIPTION

METHOD FOR CREATING A DATABASE FOR A CONTINUOUS CYCLE PRODUCTION LINE FOR DECORATED CERAMIC PRODUCTS AND COMPUTER SERVER CONTAINING THE DATABASE

Technical field

This invention relates to a method for making a database for a line for the production of decorated ceramic products in a continuous cycle and a computer server containing the database.

Background art

The lines are used for making decorated ceramic products, in particular tiles or slabs.

Said production lines comprise a plurality of interconnected machines. Each machine performs a process according to a respective configuration, determined by suitable setting parameters of the machine. In particular, each machine produces a semi-finished product which is transferred to the next machine; further works are performed on the semi-finished product.

As is known, the ceramic products are made starting from a mass of atomized ceramic powders, which are fed and pressed inside a suitable mould, in such a way as to obtain a raw pressed support generally having a flat and thin shape.

The pressed support is then subjected to a drying step which may precede one or more decorating steps, for example glazing or moulding, designed to recreate on the visible face of the pressed support a colouring, a drawing or a graphical effect.

The raw pressed support is finally subjected to a firing step with which the finished ceramic product is obtained.

Examples of the lines or parts of them are described in the following patent documents in the name of the same Applicant as this invention: EP1356909A2 and WO2013/045989.

The lines include a plurality of steps, including: preparing paste, pressing, drying, decorating and glazing, automatic movement, firing, selection and packaging.

In particular, each machine has a relative local control panel on fitted on the machine, by which users may configure the machine.

Normally, these production lines have large dimensions, in particular in a direction along which the machines interconnected together are positioned. A drawback of these lines consists in the fact that it is relatively probable that there are errors in the configuration of the machines, since there is a large number of settings necessary for each machine and that, often, the configuration of a machine affects the configuration of another machine far from the first.

For this reason, in this sector it is necessary to perform a constant and reliable control and supervision on the settings of the machines and other operating parameters along the production line to ensure that each machine operates with the correct values.

In this context, patent documents IT RE2003A000081 , ITRE2008A000087, ITB02014A000518, 102015000041012, IT 2016 06 131 1697,

1020160001051 17, IT 2020 1 1 213021 , 102021000024458, EP3198345B1 , EP3198345B1 , by the same Applicant, the documents GB2390154A1 , W02023174007A1 , EP4241951 A1 , US2017349012A1 , W003031370A2, W003031370A3, W003032096A2, W003032096A3, WO2019025852A1 , W02019135017A1 describe methods for the supervision on the production line of ceramic products for controlling one or more parameters of the process.

Furthermore, patent document CN1 15453997 provides an additional example of methods for supervision on the production line of ceramic products.

CN1 15453997 concerns the intelligent management of a factory, specifically an intelligent management system for ceramic production. CN1 15453997 involves the intelligent production of ceramics, with monitoring of various areas within the smart ceramic factory through a ceramic production monitoring unit, which sends monitoring data to the intelligent ceramic production management platform.

The intelligent management platform analyses and evaluates the monitoring data and, when abnormal data is detected, it identifies the type and level of the anomaly and notifies the management personnel, thereby facilitating their work. At the same time, the anomaly information is sent to the production monitoring unit, which identifies the location of the anomaly and transmits it back to the intelligent management platform, allowing the managers to quickly locate the fault for maintenance purposes.

In this sector, a need is that of providing an apparatus and a method for the production of decorated ceramic products in a continuous cycle with a particularly high efficiency and reliability.

In particular, it is necessary to provide an apparatus and a method which are able to control in a constant, automatic and reliable fashion various operating parameters during the production cycles of ceramic products and ensure that these parameters are maintained in the right intervals.

For this reason, it is of fundamental importance to derive performance values for each phase and each machine of the line which allow the quality requirements of the end products or of the semi-finished products to be met. Such a derivation should be particularly reliable, fast and efficient.

In particular, one of the problems addressed during the controls and analysis of the operating data in these lines is that some data which are used in these controls are not/can be detected on the line, for example because the devices designed for this detection are not available, or due to the particularly high costs.

That is to say, it can impact on the efficiency and the reliability of these controls on the line. It is also essential to provide a database which can be used to perform a control on the operating parameters of the line in a reliable and efficient manner.

Disclosure of the invention The aim of this invention is to provide a line and a method for the production of decorated ceramic products in a continuous cycle, a method for making a database for a line for the production of decorated ceramic products in a continuous cycle and a computer server which overcome the above- mentioned drawbacks of the prior art.

According to an aspect of the invention, the invention provides a line for the production of decorated ceramic products in a continuous cycle (hereinafter referred to as “line”). The line comprises a plurality of processing units. The processing units of the plurality of processing units are configured to perform respective processing operations in succession. Each processing unit includes a system of sensors and each processing unit includes a plurality of operating configurations.

The line comprises a supervision unit. The supervision unit includes a processor. The supervision unit has access, for each processing unit, to adjustment parameters. In particular, the operating configurations of the processing unit are determined by values of the adjustment parameters. Moreover, the supervision unit has access, for each processing unit, to input parameters. At least some of the input parameters are detected by the system of sensors.

The supervision unit has access, for each processing unit, to performance parameters. The performance parameters represent the performance of the processing unit.

The supervision unit has access to a non-volatile memory. The non-volatile memory includes instructions that can be read by a machine. The readable instructions ensure that the processor acquires control data. The control data includes values of the adjustment parameters for the plurality of processing units. The control data includes input parameter values for the plurality of processing units.

The instructions which can be read by a machine ensure that the processor feeds the control data to a predetermined algorithm. In one example said predetermined algorithm includes a machine-trained model and/or a physics-based model. The physics-based model may be a simplified physics-based model representative of the behaviour of the system or one of its components, used to derive said estimated values. The predetermined algorithm (in one example the machine-trained model) is configured to derive estimated values for one or more of the performance parameters. In particular, the pre-determined algorithm model is configured for deriving estimated values for one or more of the performance parameters depending on the control data.

Moreover, the instructions which can be read by a machine ensure that the processor generates an alarm signal, on the basis of said estimated values. This solution makes it possible to control various data along the production line in an automatic, fast and efficient manner and to obtain a production line which is particularly efficient and precise.

It should be noted that although, in the present description, reference is made by way of example to a machine-trained model, it is clarified that all functional and operational aspects of the invention also apply in cases where a deterministic or predefined algorithm is used, not based on machine learning techniques.

The use of a machine learning model therefore represents a preferred embodiment of the invention, but should not be considered as limiting.

The processing of control data and the generation of estimated values for performance parameters may also be implemented using other models, such as physics-based models, statistical algorithms, decision rules, threshold logic, or other equivalent computational techniques.

In one example, the system may use simplified physics-based models, either in addition to (at least partially) or as an alternative to machine-trained models.

These physics-based models are representations of the system or parts thereof, built on physical principles and analytical simplifications, offering lower computational times compared to complex numerical models (such as CFD models), and are therefore suitable for real-time applications in the daily operation of the plant.

The proposed method allows flexible operation, relying on:

• Machine-trained models (data-driven), built from operational datasets,

• Or a combination of trained models and physics-based models, in order to integrate the predictive capabilities of machine learning with the descriptive robustness of cause-effect constraints inherent in physical models.

This hybrid architecture enables greater robustness, transparency, and adaptability of the system to various operating conditions and levels of data availability.

Accordingly, in one example, the supervision unit is programmed to jointly use the machine-trained model and the simplified physics-based model. In this example, the machine-trained model may be configured to estimate one or more performance parameters based on control data, while the physicsbased model may be used to refine or validate said estimated values based on deterministic system constraints.

In one example, when the availability of input data and/or actual control parameters is limited or difficult to obtain, the supervisory unit may be configured to generate simulated or predicted data related to the operating conditions of the process (in particular, control data).

Such data may be obtained, for instance, through a physical model, a statistical model, or a machine-trained model, and used either as a replacement for or in addition to the data actually acquired from sensors.

In this way, the system is capable of deriving — even in the absence of direct measurements — the necessary parameters to feed the predictive model, and still provide said estimated values, preferably in real time, and optionally generate alert signals and/or suggestions or adjustments to the control parameters according to an estimated recipe. This allows for greater flexibility and reliability in process management, even under conditions of partial or unavailable data. Accordingly, in one example, the supervision unit is configured, in the presence of limited or unavailable input data and/or actual control parameters, to generate simulated or predicted data related to the operating conditions of the process.

Said data may be obtained through a physical model, a statistical model, or a machine-trained model, and may be used as a replacement for or in addition to the data acquired from sensors, in order to feed the predefined model and generate, preferably in real time, the estimated values, the alert signal, and/or suggestions or adjustments to the control parameters according to a derived recipe.

According to an example, the supervision unit is programmed to compare the performance parameters with one or more estimated values for one or more of the performance parameters and to generate the alarm signal in response to a difference between the performance parameters and one or more estimated values for one or more of the performance parameters which exceeds a predetermined threshold.

In one example, the machine-trained model is configured to derive estimated values of one or more performance parameters before the occurrence of an anomalous event. This allows the system to proactively generate an alert signal, improving the production line's ability to respond promptly to potentially critical operating conditions.

In one example, the alert signal is generated when a difference is detected between the measured performance parameters and the corresponding estimated values, and this difference exceeds a predefined threshold. This setup enables the system to operate in predictive mode, triggering corrective actions or maintenance interventions before a significant performance degradation or failure occurs.

Therefore, according to one aspect of the present description, the system exhibits predictive behaviour. In particular, unlike known solutions in the field, where the detection of anomalous conditions is based on a direct comparison between real-time data and a set of predefined conditions stored in a database, the present invention introduces a predictive approach.

Specifically, the production line described in the present invention uses a predefined algorithm, such as a model trained using machine learning techniques, configured to receive as input control data including regulation parameters and input parameters (in one example, collected in real time from individual processing units). Based on this data, the model is capable of deriving estimated values for one or more performance parameters of each unit, prior to the occurrence of a potential anomaly. Therefore, in one example, the estimation can be a forecast.

This predictive capability allows the system to anticipate abnormal behaviour and proactively generate an alert signal whenever a significant difference is detected between the actual and estimated performance parameter, exceeding a predefined threshold. This results in a substantial improvement in terms of reliability, safety, and operational continuity compared to conventional systems that merely detect anomalies once they have already occurred.

According to an example, the supervision unit is programmed to estimate a trend over time of one or more of the performance parameters as a function of the control data. According to an example, the supervision unit is programmed to suggest, preferably in real time, changing one or more parameters of the adjustment parameters, according to a recipe derived on the basis of the estimated values.

For this reason, according to an example, the supervision unit is programmed to derive estimated values for one or more of the performance parameters, depending on the control data, and suggests to the operator optimum values for the adjustment parameters which make it possible to maintain or obtain said estimated values for one or more of the performance parameters.

The supervision unit is programmed for adjusting, preferably in real time, one or more parameters of the adjustment parameters according to a recipe derived on the basis of the estimated values. For this reason, according to an example, the supervision unit is programmed to derive estimated values for one or more of the performance parameters, depending on the control data, and automatically adjusts the adjustment parameters so as to maintain or obtain said estimated values for one or more of the performance parameters.

In this document, the expression “real time” refers to the ability of a system to process data and generate an output (such as a decision, command, or recommendation) within a time interval short enough to enable an effective response to the current operating conditions of the monitored process or system. This interval may vary depending on the application but generally implies that the system responds with a latency that is compatible with the dynamics of the ongoing process, without delays that would compromise its effectiveness or stability.

According to an example, the supervision unit is programmed to perform a feedback control on the performance parameters. The feedback control is performed with a predetermined frequency. The supervision unit is configured for adjusting the adjustment parameters in the case of identifying a deviation between the performance parameters and the estimated values. According to an example, one or more parameters of the input parameters or of the adjustment parameters are input by an operator or by an offline device.

In this example, said data input by an operator or by an offline device are temporally correlated to corresponding values acquired for one or more control data.

For this reason, it is possible to obtain a particularly reliable data control along the line.

According to an example, the supervision unit is programmed to receive target values for one or more of the performance parameters. Said target values for one or more of the performance parameters are temporally correlated with corresponding values acquired for the adjustment parameters and for the input parameters. Moreover, the supervision unit is configured to generate training data for training the machine-trained model, based on the target values. Moreover, the supervision unit is configured to generate training data for training the machine-trained model on the basis of the corresponding values of the adjustment parameters and of the input parameters.

According to an example, the plurality of processing units includes one or more of the following units:

- forming/drying,

- decorating,

- firing.

As regards these processing units, it should be noted that lines for making decorated ceramic tiles, or parts of these lines, are described in the following patent documents in the name of the same Applicant and incorporated herein by reference: WO2016/046724, WO2013/050845, EP1356909A2 and WO2013/045989.

The firing unit comprises an oven and a conveyor which passes through the oven. The conveyor includes conveyor rollers which support and move the tiles inside the oven. The oven is equipped with a plurality of heating elements. The heating elements are adjustable to define a temperature profile of the oven.

The input parameters include one or more of the following parameters: humidity of the incoming product; the type of paste to be used; the type of decoration; and the thickness and/or format and/or planarity of the product;

The adjustment parameters include one or more of the following parameters: the compacting pressure; number of drying cycles; distance interval of the pressing rollers; the temperature profile of the oven; firing combustion air.

The performance parameters include one or more of the following parameters: the thickness of the product; the humidity of the outgoing product; the temperature of the outgoing product; energy consumption (of one or more processing units or parts thereof); surface defects for the tiles or semi-finished tiles.

The management of the production process in the ceramic lines comprises a control (in real time), using a system of sensors, of one or more (or all) the processing steps, in order to be able to assess the suitability of the semifinished products in order to maximise the quality of the finished product.

For example, the surface defects are detected by video cameras and processors programmed to process the images taken by the video cameras. Generally speaking, the control unit is programmed to identify (processing the images taken by the video cameras) the following types of defects on the tiles (especially after firing), corresponding to the same number of parameters:

- geometrical conformity parameters (for example: planarity, orthogonality, rectilinearity);

- structural conformity parameters (for example, representing the presence of cracks, laminations or chips);

- surface defect parameters (for example, representing the presence of dips, craters, bubbles, crazing);

- surface finishing parameters (representing the presence of possible surface aesthetic defects, such as, for example, marks, contamination, decoration errors).

According to an example, for one or more of the processing units of the plurality of processing units, the processing unit includes a control card. The control card has access to the respective adjustment parameters. The control card has access to the respective adjustment parameters. The control card has access to the respective input parameters. The control card has access to the respective performance parameters. The control card is programmed for deriving values of the adjustment parameters depending on the input parameters and on the performance parameters.

According to an aspect of the invention, the instructions which can be read by a machine ensure that the processor receives target values for one or more of the performance parameters, temporally correlated with corresponding values acquired for the adjustment parameters and for the input parameters. Moreover, instructions which can be read by a machine ensure that the processor generates training data to train the machine- trained model, based on the target values and the corresponding values of the adjustment parameters and of the input parameters.

This aspect may be combined with one or more aspects of the invention. It should be noted that this aspect makes the line available according to one or more aspects of the invention having the above-mentioned features for generating training data type for training the machine-trained model and may be combined or not with the generation of the alarm signal.

According to an example, the performance parameters include parameters acquired online and/or offline. In particular, the parameters detected offline are temporally correlated to the corresponding adjustment parameters and input parameters.

According to an example, at least one of the following conditions occurs:

- the line includes a registration device for recording the date and/or the time and/or the station of the acquisition of the parameters acquired offline;

- the line includes a device for reading an identification code uniquely associated to a particular product, wherein the code contains the date and/or the time and/or the station of the acquisition of the parameters acquired offline for said specific product.

According to an example, the supervision unit is programmed to associate a unique code to each product in the line and track the product using said unique code.

In one example, among the input parameters used by the system, values associated with the unique code assigned to each product or batch may also be included. This unique code is correlated with the other input parameters detected along the production line and is used to feed a structured database.

Specifically, the unique code makes it possible to correlate data acquired at different times and/or at different points along the production line, thereby facilitating the complete reconstruction of the processing cycle for each product. This data structure enables the system to track the evolution of operating parameters and to establish cause-effect relationships, improving the accuracy of both predictive models and real-time decisions.

In one example, the unique code is used to correlate the values of input parameters detected at different points of the production line and/or at different times, specifically to feed a structured database and allow for the reconstruction of the processing cycle and the analysis of cause-effect relationships.

According to an example, the line is equipped with a marking device. The marking device is configured to control the generation of said unique code for a product. Moreover, the marking device is configured to start the training. According to an example, the unique code contains data representing the control data and/or performance parameters and/or target values and/or the temporal correlation between said values and parameters. According to an aspect of the invention, a method is provided for the production of decorated ceramic product. The method comprises a step of acquiring control data, in a supervision unit. The control data includes values of adjustment parameters and input parameters for the plurality of processing units. According to an example, at least some of the input parameters are detected in the production line in which the ceramic products are produced. The values of the adjustment parameters determine operating configurations of a plurality of processing units which perform respective processing operations in succession along the production line.

The method comprises a step of feeding the control data to a machine- trained model. The machine-trained model is trained to derive estimated values for one or more performance parameters. In particular, the machine- trained model is trained to derive estimated values for one or more performance parameters depending on the control data. The performance parameters represent the performance of the processing unit.

The method comprises a step of generating an alarm signal, based on said estimated values.

According to an example, the method comprises a step of comparing the performance parameters with one or more estimated values for one or more of the performance parameters. The method may comprise a step of generating the alarm signal when a difference between the performance parameters and one or more estimated values for one or more of the performance parameters exceeds a predetermined threshold.

According to an example, a feedback control is performed on the performance parameters. The feedback control is performed according to a predetermined frequency. In particular, when a deviation between the performance parameters and the estimated values is identified, the performance parameters are adjusted.

The method may comprise a step of receiving target values in the supervision unit for one or more of the performance parameters. The target values for one or more of the performance parameters are temporally correlated with corresponding values acquired for the adjustment parameters and for the input parameters. The method may comprise a step of generating training data for training the machine-trained model, on the basis of the target values and the corresponding values of the adjustment parameters and of the input parameters.

According to an aspect of the invention, the method includes a step of receiving, from the supervision unit, target values for one or more of the performance parameters. The target values for one or more of the performance parameters are temporally correlated with corresponding values acquired for the adjustment parameters and for the input parameters. Moreover, according to this aspect of the invention, the method includes a step of generating training data for training the machine-trained model, on the basis of the target values and the corresponding values of the adjustment parameters and of the input parameters.

It should be noted that this aspect can be combined with one or more steps of the method according to one or more aspects of the invention. In particular, this aspect relates to the step of training the machine-trained model and may be combined less with the step of generating the alarm signal.

According to an example, the performance parameters include parameters acquired online and/or offline. The parameters detected outside the line are temporally correlated to the corresponding adjustment parameters and input parameters.

According to an example, at least one of the following conditions occurs:

- the date and/or the time and/or the station of the acquisition of the parameters acquired offline is recorded by means of a registration device positioned along the line;

- an identification code uniquely associated to a particular product, containing the date and/or the time and/or the station of the acquisition of offline purchasing parameters for said specific product, is read and recorded by means of a device for reading the identification code.

According to an example, a unique code is associated with each product in the line and the product is tracked along the line using said unique code.

According to an example, the unique code contains data representing the control data and/or performance parameters and/or target values and/or the temporal correlation between said values and parameters.

According to an aspect of the invention, this invention provides a method for making a database for a line for the production of decorated ceramic products in a continuous cycle. The line is according to one or more aspects of the invention. The method comprises a step of preparing a non-volatile memory. The method comprises a step of preparing a processor.

The method comprises a step of, by means of the processor, storing in the memory values of adjustment parameters. The operating configurations of the processing unit are determined by values of the adjustment parameters. The method comprises a step of, by means of the processor, storing in the memory values of input parameters. At least some of the input parameters are detected by the system of sensors.

The method comprises a step of, by means of the processor, storing in the memory values of performance parameters. The performance parameters represent the performance of the processing unit.

The method comprises a step of, by means of the processor, storing in the memory time stamps for temporally synchronising the values of the adjustment parameters, input parameters and performance parameters.

According to an example, the parameters acquired offline are input manually by a user and/or by a device located offline.

According to an example, at least one of the following conditions occurs:

According to an example, a unique code is associated with each product in the line. The unique code contains data representing the control data and/or performance parameters and/or target values and/or the temporal correlation between said values and parameters.

According to an example, the code is a bar code, or a QR code, or an RFID label.

According to an example, the non-volatile memory and the processor are provided with a knowledge base formed by data representative or descriptive of the operation of the line.

According to an example, the data are collected from machine manuals of the line and/or results of laboratory tests, and/or electrical diagrams, and/or mechanical drawings of parts of machines of the line.

According to an example, at least one of the values of the adjustment parameters, of the input parameters and of the performance parameters comes from data relating to the production of the line. In particular, data relating to the production of the line is collected by line information systems and/or production management software in the line.

According to an example, a machine-trained model is contained in the nonvolatile memory. The machine-trained model is trained to derive estimated values for one or more of the performance parameters, depending on the control data.

According to an example, there is a natural language processing engine, NLP. The processing engine has access to the texts contained in the memory. There is also a step of processing the texts contained in the memory, using NLP.

According to an example, a request text, input by a user, is processed using the NLP engine. The database can be used by artificial intelligence algorithms to provide natural language responses to said request.

In one example, a conversational interface (chatbot) is provided, integrated with a natural language processing (NLP) engine. The chatbot is configured to interact with the user through a natural language conversational interface, allowing the user to input requests, commands, or queries related to the status of the production line, and/or regulation parameters, and/or estimated values, and/or generated alerts, or other process data. The chatbot is operationally connected to both the database and the NLP engine.

The chatbot is connected to the database, which, as previously described, can store input parameters, regulation parameters, performance parameters, estimated values, regulation recipes, and other operational data. The NLP engine is configured to extract relevant information from memory and from the database, for example based on a unique product or batch identifier, and to generate consistent and context-aware responses in natural language.

In this way, the operator can quickly obtain explanations, suggestions, or diagnoses through natural language dialogue, even in the presence of complex and articulated data, thus improving the accessibility and efficiency of the supervision system.

The chatbot is capable of generating contextualized operational suggestions based on a plurality of information sources. These sources include, by way of example and without limitation, one or more of the following:

• Contextual technical documentation, such as machine and process manuals;

• Real-time process data;

• Historical databases related to production cycles;

• Optimization and/or predictive models, based on machine learning and/or physics-based approaches;

• Structured lists of performance parameters relevant to various systems.

The information content generated by the chatbot may include one or more of the following:

• Suggestions for modifying operational parameters (e.g., temperature setpoints);

• Optimized quantitative indications (e.g., variation of process values);

• Detailed operating procedures to implement the suggestion; • Estimates of the remaining useful life of components in predictive maintenance scenarios;

• Proposals for reducing energy consumption during high-intensity phases.

In some implementations, the chatbot may propose the automatic execution of the recommended changes, subject to explicit approval by the human operator.

These functionalities can be extended to multiple use cases, such as, by way of example: in-line quality control, process performance optimization, predictive maintenance, and energy efficiency management.

According to an aspect of the invention, a computer server is provided. The computer server comprises a processor. The computer server includes a non-volatile memory. The memory contains a database relevant to a line for the production of decorated ceramic products in a continuous cycle. The line is according to one or more aspects of the invention. The database contains values of adjustment parameters. The operating configurations of the processing unit are determined by values of the adjustment parameters. The database contains input parameter values, at least some of which being detected by the system of sensors.

The database contains values of performance parameters, representing performance of the processing unit.

The database contains time stamps for that data.

The database contains time stamps for temporally synchronising the values of the adjustment parameters, input parameters and performance parameters.

In particular, the computer server according to this aspect of the invention may be combined with one or more aspects of the invention.

Brief description of drawings

This and other features will become more apparent from the following description of a preferred embodiment of the invention, illustrated by way of non-limiting example in the accompanying tables of drawings, in which:

- Figure 1 illustrates a line for the production of decorated ceramic products in a continuous cycle according to one or more aspects of the invention;

- Figures 2 and 3 illustrate diagrams on the function of the supervision unit;

- Figures 4 and 5 show different processing units of the line with the traditional pressing and continuous pressing, respectively;

- Figure 6 illustrates a diagram on training and the generation of training data.

Detailed description of preferred embodiments of the invention

With reference to the accompanying drawings, the numeral 100 denotes a line for the production of decorated ceramic products in a continuous cycle. The line 100 includes a plurality of processing units 1 , 2, 3, 4, 5, 6. The processing units are configured to perform respective processing operations in succession. In particular, each processing unit produces a product or a semi-finished product which may be processed in the subsequent unit. The plurality of processing units includes a unit for preparing paste 1 , a pressing unit 2, a drying unit 3, a decorating and glazing unit 4, an automatic movement unit 5, a firing unit 6. The unit for preparing the paste 1 is aimed at obtaining the powder to be pressed and comprises three sections:

1 ) the raw material dosing line, which may be more or less automated;

2) the grinding, which also has different levels of automation. It can be designed in a traditional way with discontinuous mills or in a more automated way and controlled with continuous mills or continuous modular mills with high efficiency and reduced electrical absorption.

3) the atomizer, which dries the finely atomized slip to obtain the powders to be pressed. SACMI produces atomizers where the introduction of the slip into the drying tower is performed with a circular crown or alternatively by means of lances.

A complete range of machines is available for the preparation of paste for every production requirement.

In particular, there may be one or more T hoppers for storing powder (of various colours) which comes out from atomizers. The powders are then loaded in the pressing unit by means of powder loaders 9. The pressing is the processing step during which the ceramic product is formed. The pressing may be continuous or non-continuous.

Drying occurs after the pressing step. In the cases in which the pressing is continuous, there is a cutting unit 1 1 between the pressing unit 2 and the drying unit 3. Vertical drying units can be used in the drying unit 3, with roller cages or horizontal roller drying units in a single layer version, with two, three, five and seven planes. The drying unit may be designed for recovering hot air coming from the cooling flue of the firing unit (oven), thus allowing significant energy savings.

The products which leave the drying unit 3 are conveyed to the decorating and glazing unit 4 which includes one or more printers ST for decorating and printing images on the ceramic product.

The automatic movement unit 5 includes a system for moving and storing the raw and baked product.

The products may be stored in a storage block 12 after the decorating step or directly after the drying (still raw). The products may also be transported after drying and/or the decoration directly to the firing unit without any storage step.

The automatic movement unit allows the production of the glazing and selection lines to be separated from that of the oven, guaranteeing continuity of feeding and unloading of the oven. The storage makes it possible to organize production by batches, facilitating product changeovers and increasing ceramic production efficiency. The solutions which can be proposed comprise the movement of the raw product and the baked product using automatic vehicles.

The raw product is stored in roller boxes, whilst a high-density solution may also be used for the baked product with the direct stacking on shelves of the baked product.

The ceramic products are then brought to the firing unit.

The firing is the fundamental operation of the technological process; in it the irreversible reactions develop which transform the raw materials of the paste and the enamels or the dyes into the finished product.

A finishing unit 13 may be provided after the firing. At the finishing unit in a step of selection, packaging and palletizing, with various levels of automation, the production is controlled and classified according to dimension, planarity, defects and tones. It is then sorted and grouped by automatic stackers into homogeneous classes of product, and then packaged and placed on the pallets.

The packaging can be in dies for small/medium sizes or in trays for larger sizes. For the palletizing there are models with a track on the ground, or models with portal structure suitable for moving heavier loads.

Each processing unit includes a system of sensors S. Each processing unit includes a plurality of operating configurations.

The line includes a supervision unit 7, which includes a processor 8. The supervision unit has access, for each processing unit, to adjustment parameters, input parameters and performance parameters. The supervision unit can be located online or offline. In the same way, the processor can be located online or offline. One or both of the processor supervision units may be located on the line or offline, or one of the two is located on the line and the other is located offline.

At least some of the input parameters are detected by the system of sensors.

Some of the input parameters or adjustment parameters may be detected offline. For example, it may be that some parameters are detected by an offline machine, by means of virtual sensors, or they are detected and inserted manually by an operator or by an offline device.

The performance parameters represent the performance of the processing unit. Values of the adjustment parameters determine the operating configurations of the processing unit.

Furthermore, the supervision unit has access to a non-volatile memory. The non-volatile memory includes instructions that can be read by a machine.

In particular, the processor acquires control data. The control data includes values of the adjustment parameters and input parameters for the plurality of processing units. The data input by an operator or by an offline device are temporally correlated to corresponding values acquired for one or more control data.

The processor feeds the control data to a machine-trained model, which is trained to derive estimated values for one or more of the performance parameters, depending on the control data.

Moreover, the processor generates an alarm signal, based on said estimated values.

The supervision unit receives target values for one or more of the performance parameters, temporally correlated with corresponding values acquired for the adjustment parameters and for the input parameters, and generates training data for training the machine-trained model, based on the target values and the corresponding values of the adjustment parameters and of the input parameters.

The performance parameters can include parameters acquired online and/or offline. The parameters detected outside the line are temporally correlated to the corresponding adjustment parameters and input parameters.

According to an example, the line includes a registration device for recording information of the parameters which have been acquired offline; for example, the date and/or the time and/or the station of the acquisition of the parameters acquired offline.

Moreover, the line may include a device for reading an identification code uniquely associated to a particular product, wherein the code contains the date and/or the time and/or the station of the acquisition of the parameters acquired offline for said specific product. The supervision unit can associate a unique code to each product in the line and track the product using said unique code throughout the entire line.

The line may be equipped with a marking device which controls the generation of said unique code for a product and starts the training (or requests the starting).

In particular, the unique code contains data representing the control data and/or performance parameters and/or target values and/or the temporal correlation between said values and parameters. Data of different phases along the line can be associated with the unique code.

The supervision unit may compare the performance parameters with one or more estimated values for one or more of the performance parameters and generate the alarm signal in response to a difference between the performance parameters and one or more estimated values for one or more of the performance parameters which exceeds a predetermined threshold. The threshold may be defined by the operator.

The supervision unit can estimate a trend, or a trend over time, of one or more performance parameters as a function of the control data. For example, the supervision unit can estimate that a performance parameter remains in a certain range if the consumption value of one or more parts of the line is considered in a certain range.

The supervision unit may suggest changing one or more parameters of the adjustment parameters, according to a recipe derived on the basis of the estimated values. In particular, once the supervision unit obtains the estimated values, it can suggest which adjustment parameters it is necessary to work on to obtain/maintain said estimated values.

The supervision unit can adjust one or more parameters of the adjustment parameters according to a recipe derived on the basis of the estimated values.

It may be envisaged that after estimating the estimated values the supervision unit suggests a recipe to the operator (in particular for the adjustment values) and if the operator requests an adjustment, the supervision unit automatically adjusts the values of the adjustment parameters according to the recipe derived for a value of a performance parameter.

The supervision unit can perform a feedback control on the performance parameters, with a predetermined frequency, and adjust the adjustment parameters in the case of identifying a deviation between the performance parameters and the estimated values.

In particular, with the solution according to this invention it is possible to estimate optimum values for the performance parameters along the line on the basis of the data (representing operating conditions) which can be detected automatically by devices or machines of the line or set manually by the operator and to suggest or set the optimum settings of one or more machines to maintain optimum product, productivity and consumption characteristics.

The recipe for the settings can be derived as a function of a set of historical data available on products made in the past or of a current catalogue of the ceramic product and a set of information I quality targets I productivity/ consumption on new products to be made. The recipes can be stored. Various parameters, and in particular values estimated for the performance parameters and the recipes for the settings of the machines (processing units), and the alarm signal (deviation of the performance parameters from the estimated parameters) can be displayed on a panel of each machine, on a web application and a notification in push mode can be sent to the operator by email or other messaging applications.

Data used by the supervision unit may come from technological documentation and/or from corporate information systems, such as ERP or production management software (e.g. MES, MOM or similar), such as, for example, the production batches planned on the various production lines or the batches produced in a certain period of time, characteristics of the items being produced, etc, or they may be generated by virtual sensors.

In particular, a software application may generate the suggestions (recipes) as requested by the user and show them on a screen or export them to a file; it may send notifications to pre-set groups of users (for example: email, WhatsApp, Telegram, or other channels) and may directly set the machine parameters (with confirmation, or not, by the user).

The application can be used both by PCs and mobile devices such as tablets or smartphones. It may also exchange data with the machines and devices of the line, which in turn may, advantageously but not necessarily, show the suggestions produced by the application on its interface panel and generate alarms if there are deviations between the suggestions provided by the application and the current settings of the machine.

The data exchange between the software application and the devices of the line may take place, by way of non-limiting example, by means of automation protocols, webAPI, files of any format or databases of any type and by means of a wired or wireless network connection.

The application also, or alternatively, allows the optimum settings of the machines for new products to be suggested as a function of the technological features, such as, only by way of example, size and thickness, paste, glaze card or others (not necessarily all together: the significant features may depend on the processing cycle in which the machine is inserted). In this way, it generates a first version of a machine recipe, which may subsequently be sent to the machine. The application can use many algorithms with predetermined rules, algorithms belonging to the category of Artificial Intelligence, in particular to the category of Machine Learning.

According to an aspect, a database for a line 100 is made.

In particular, the processor stores in the memory the adjustment parameter values, the performance parameter values, the input parameter values and time stamps for temporally synchronising the values of the adjustment parameters, input parameters and performance parameters.

The database can reside in both the cloud and on premises and can be used by artificial intelligence models to answer questions from the operator about the conduction of the line. The non-volatile memory and the processor are provided with a knowledge base formed by data representative or descriptive of the operation of the line.

The data are collected from machine manuals of the line and/or results of laboratory tests, and/or electrical diagrams, and/or mechanical drawings of parts of machines of the line.

At least one of the values of the adjustment parameters, of the input parameters and of the performance parameters comes from data relating to production of the line. The data relating to the production of the line is collected by line information systems and/or production management software in the line.

The machine-trained model is contained in the non-volatile memory.

The database necessary to validate the algorithms implemented and, in the case of Machine Learning algorithms, for the training step of the algorithms, includes both data coming from historical series and data referred to a single product uniquely identified.

There is a natural language processing engine, NLP, having access to the texts contained in the memory. There is a step of processing the texts contained in the memory, using NLP.

A request text, input by a user, is processed using the NLP engine. The database can be used by artificial intelligence algorithms to provide natural language responses to said request.

The database used for training machine learning algorithms may include both data coming from real production lines and optionally data generated synthetically by physical simulation models of the process.

The data relating to the individual product require that each product be identified uniquely and this may occur in a virtual manner (identification transferred between the machines which process it) or in a physical manner (application of an identification code directly on the workpiece or on a device which moves in an integral fashion with the workpiece, such as, for example, labels, RFID). The database may also include data coming from other software functionalities included optionally in the application, such as, for example, work shifts, the production batch in progress, the article and its features. Data entry can be both manual and automatic, that is to say, there is the possibility of reconciling inside the same database both the results of measurements or controls performed manually offline and the measurements and controls performed by automatic devices, such as, for example, production machines, vision systems, or other types of sensors.

In the case of measurements or controls performed manually, it is possible to use a functionality of a software application which makes it possible to correctly identify the time of collection of the material to be analysed from a certain pick-up point; the operator responsible for performing offline tests or measurements may, at the same time or subsequently, report the results of the controls performed, even if performed at subsequent moments, and the application keeps track of both the date and time of collection and the date and time in which the results have been input. This result may be obtained by using a software application wherein the operator declares the instant of picking up material from a preconfigured pick-up point, or by using devices located close to the pick-up points, such as, by way of non-limiting example, the use of pushbuttons, functions included in a panel close to the line or functions included in the control interface of a machine of the line. According to an example, the line 100 comprises at least one portable control panel 10. The portable control panel 10 is, for example, a tablet, a smartphone, or another electronic device. The portable control panel 10 has a display unit and a user interface; the interface could be defined by the same display unit, with a touch screen technology of known type.

The portable control panel 10 has a processor and a connection module, for putting the portable control panel in wireless communication with each control unit of each processing unit of the line and with the supervision unit. The portable control panel makes it possible to input data obtained offline and to reconcile data obtained online with data obtained offline and to start generating training data and to supervise the line. These devices, which are interconnected with the software application, communicate the date and time of collection to the software application, so that the operator can input the measurements taken as soon as they become available. In this way, the data detected manually have a reference to the instant of picking up from the line and are correctly contextualised in the database with reference, for example, to the date and time of production or to the item which was in production at the time of picking up or other dimensions relevant for the classification of the data.

If the parts picked up for the manual measurements are uniquely identified with a code or a device applied on them, the operator in charge of the tests will have the possibility of inputting the identification code in the software application, manually or using devices for recognising the code, in such a way that in the database governed by the software application the measurements made can be correctly assigned to the part being measured and reconciled with all the other measurements, even automatic, previously or subsequently assigned to the same part.

The invention also provides a computer server which includes the processor and the non-volatile memory according to one or more aspects of the invention. The memory contains the database according to one or more aspects of the invention.

The input parameters include at least one of the following parameters:

- Current flow of raw materials (instantaneous flow rate of the incoming solids);

- Current flow rate of loose clays (instantaneous flow rate of the incoming clays);

- Water flow rate (instantaneous flow rate of the incoming water);

- Temperature of the air at the inlet of the atomizing tower;

- Temperature of the atomized powder at the outlet;

- Pressure in the presser cylinder, proportional to the compaction pressure;

- Water content (in mass) of the atomized powder entering;

- Water content (in mass) of the ceramic at the outlet; - Shape (height and diameter) of the nozzles and their number;

- Water content (in mass) of the ceramic at the inlet;

- Slip inlet temperature;

- Thickness of the product to be baked;

- Humidity of the product at the inlet;

- The type of paste to be used;

- The type of decoration;

- The thickness and/or the format and/or the planarity of the product;

- Dry composition of the PDM recipe;

- % dry content from dispersed clays;

- Moisture content of PDM raw materials;

- Density of dispersed clays;

- % deflocculant on dry content;

- Nominal residue range (%) at 63 or 45 microns;

- Density of recovered water;

- % of recovered water;

- Current raw material flow rate (kg/hour);

- Current dispersed clay flow rate;

- Current water flow rate (liters/hour);

- Current deflocculant flow rate (liters/hour);

- Current power (kW) of MMC modules;

- Nominal slurry density;

- Desired slurry viscosity (in cP);

- Slurry residue or D50;

- Nominal moisture content of atomized powder;

- Nominal particle size of atomized powder;

- Powder production (t/h);

- Inlet slurry density;

- Outlet powder temperature;

- % filtration waste;

- Current PPB pressure; - Number of nozzles;

- Type of nozzles;

- Current slurry flow rate;

- Current discharge valve opening setting

- Current air flow rate in operation;

- Current tower vacuum level;

- Nominal thickness of pressed tile;

- Productivity [cycles per minute;

- Mold dimensions;

- Number of mold outlets;

- Specific pressure I Cylinder pressure (1 st and 2nd pressing);

- Current soft loading thickness;

- Current cylinder pressure;

- Current final TM pressure;

- Current % of YV318-YV32;

- Current piston speed;

- Current SPE parameters;

- Current trolley speed;

- Current trolley position settings;

- Moisture of atomized powder;

- Particle size of atomized powder;

- Nominal moisture of the finished product;

- Nominal dry breaking load;

- Nominal output temperature;

- Body type;

- Forming type (traditional I continuous);

- Dimensions and weight of treated product;

- Current drying cycle;

- Ambient humidity;

- Inlet air;

- Product inlet moisture; - Inlet density;

- Green breaking load;

- Nominal density of pressed slab;

- Nominal thickness of pressed slab;

- Productivity [meters per minute];

- Nominal dimensions of slab and sub-formats;

- Roller spacing;

- Current soft loading thickness;

- Current APB settings;

- Current guide width;

- Current doctor blade values;

- Current roller pressure;

- Current compactor speed;

- Current calibration plate values;

- Current TPV cutter values;

- Body type;

- Decoration type;

- Product families (to be polished, matt material);

- Size (Format);

- Treated product weight;

- Current firing cycle;

- Product inlet moisture;

- Apparent density at inlet;

- Product inlet temperature;

- Product pyroplasticity;

- Dilatometry of support I engobe I glaze;

- Quantity of glaze I engobe;

- Ambient humidity;

- Inlet air;

- Target nominal format;

- Inlet format; - Inlet thickness;

- Water volume used;

- Tool type;

- Inlet flatness;

- Tensioning state;

- Inlet temperature;

- Productivity;

- Current number of incisions;

- Current conveyor belt feed speed for incision;

- Current disc incision;

- Current cycles of incision machine;

- Current cycles of splitting machine;

- Current number of splits on parent format;

- Inlet format;

- Inlet thickness;

- Water volume used;

- Tool type;

- Inlet flatness;

- Material type;

- Open porosity at inlet;

- Inlet gloss;

- Water volume used;

- Tool type;

- Inlet flatness;

- Inlet temperature.

The adjustment parameters may include at least one of the following parameters:

- Pressure of the slip entering the nozzles;

- The compaction pressure;

- Temperature-time curve of the dryer;

- Number of drying cycles; - Distance interval of the pressing rollers;

- The temperature profile of the oven

- Firing combustion air;

- Distance between the presser rollers (minimum clearance)

- Feed speed of the tiles relative to the tools (fixed);

- Setpoint flow rate of raw materials (kg/h);

- Setpoint flow rate of loose clays;

- Setpoint water flow rate (liters/hour);

- Setpoint deflocculant flow rate (liters/hour);

- Setpoint power of MMC modules (kW);

- Setpoint current PPB pressure;

- Setpoint number of nozzles;

- Setpoint nozzle type;

- Setpoint slurry flow rate;

- Setpoint discharge opening adjustment;

- Setpoint working air flow rate;

- Setpoint tower vacuum level;

- Setpoint soft loading thickness;

- Setpoint current cylinder pressure;

- Setpoint current final TM pressure;

- Setpoint % YV318-YV32;

- Setpoint current piston speed;

- Setpoint SPE;

- Setpoint current trolley speed;

- Setpoint current trolley positions;

- Drying cycle;

- Fan rotation rate by zone (Hz - Flow rates);

- Opening of “false” air in the zones;

- Opening/Closing of chimneys;

- Setpoint for soft loading thickness;

- Setpoint for APB measurements; - Setpoint for VNT I APD+;

- Setpoint for nominal guide width;

- Setpoint for doctor blades;

- Setpoint for roller pressure;

- Setpoint for roller distance range;

- Setpoint for PCR speed;

- Setpoint for calibration plate;

- Setpoint for TPV cutters;

- Drying cycle;

- Fan rotation rate by zone (Hz - Flow rates);

- Opening of “false” air in the zones;

- Opening/Closing of chimneys;

- Preheating temperatures;

- RR temperatures;

- RL temperatures;

- Temperatures under roller axis in RL;

- Maximum firing temperature;

- Combustion air in preheating;

- Combustion air in firing;

- Firing/cooling;

- Roller movement/advancement;

- Draft of AF fan;

- Draft of AAC fan with RCC hood;

- Opening of ANFU hood valve;

- Opening/closing of RR valves;

- Opening/closing of RGA valves;

- Opening/closing of L blowers;

- Blowing in RF;

- RR regulation;

- Pressure regime;

- RCC hood regulation; - Number of engravings;

- Engraving belt feed speed;

- Engraving with disk;

- Engraving machine cycles;

- Splitting machine cycles;

- Number of splits on parent format;

- Space between consecutive tiles;

- Feed speed of 1 st module;

- Number of acting calibrators;

- Number of acting bevelers;

- Bevel inclination;

- Grinding wheel rotation speed;

- Material removed per calibrator 1 st module;

- Pusher speed;

- Feed speed of 2nd module;

- Number of acting calibrators;

- Number of acting bevelers;

- Bevel inclination;

- Grinding wheel rotation speed;

- Material removed per calibrator 2nd module;

- Beam stroke;

- % beam oscillation

- Beam oscillation curve;

- Granulometric scale;

- Belt feed speed;

- Number of heads;

- Head rotation speed;

- Head pressure;

- Head backpressure. The performance parameters include one or more of the following parameters:

- The thickness and/or the format and/or the planarity of the product; - The humidity of the outgoing product;

- The temperature of the outgoing product;

- Sheen of the processed surface;

- Nominal size of the finished part

- Quantity of water absorbed by the material after firing;

- Mechanical strength (under bending) of the tile after firing;

- Water content (in mass) of the ceramic at the outlet

- Granulometric distribution of the atomized powder;

- Slip density (density of the ceramic suspension exiting the grinding);

- Slip viscosity (viscosity of the ceramic suspension exiting the grinding);

- Grinding residue of the mix (% of solids with particle size > 45 pm. Indicator of the grinding degree of the slip exiting the mills);

- Atomized powder moisture (moisture exiting the atomizer, must fall within an optimal range);

- Atomized powder granulometry (particle size distribution of the atomized particles, must fall within an optimal range);

- Specific consumption (thermal consumption of the atomizer per unit of evaporated water. Main indicator of process efficiency);

- Pressed tile/slab density (fundamental parameter for the efficiency of the pressing system, must fall within an optimal range);

- Pressed tile/slab thickness (fundamental dimensional parameter for the specific product in production);

- Pressed tile/slab dimensions (fundamental dimensional parameter for the specific product in production);

- Presence of cracks on the green body (Any cracks must be detected and the tile/slab discarded in the green state, before subsequent decoration or firing);

- Dry breaking load (Mechanical flexural strength of the tile after drying);

- Output temperature (Useful parameter for subsequent glazing and decorating stations);

- Dimensional variation after drying (Shrinkage or expansion affects crack formation during drying. Optimal value 0%. Expansion to be avoided);

- Output moisture (Must be low, but not 0%, to provide greater tile resilience and impact resistance);

- Presence of cracks after drying (Any cracks must be detected and the tile/slab discarded in the green state, before subsequent decoration or firing);

- Printing defects (Streaks, bands, lines caused by missing nozzles, misaligned digital graphics, halos, or surface stains);

- Amount of glaze applied (Consistency in the applied amount ensures final quality and surface performance);

- Color tone (The overall colour tone must be carefully checked to avoid unwanted defects when placing tiles of slightly different tones side by side); Black core;

-Tone (The overall colour tone must be carefully checked to avoid unwanted defects when placing tiles of slightly different tones side by side);

- Pre-heating cracks;

- Expansion (over-firing);

- Flatness (Variation in curvature - warping of the surface relative to nominal values);

- Cooling draw-off;

- Cooling stresses;

- Output temperature;

- Firing shrinkage (Linear dimensional variation. A fundamental parameter indicating correct sintering of the ceramic material);

- Water absorption (Amount of water absorbed by the material after firing. An important indicator to verify proper sintering of the ceramic material);

- Output temperature (Mechanical flexural strength of the tile after firing);

- Surface appearance;

- Dilatometry of body/engobe/glaze;

- Pyroplasticity;

- Density; - Flatness;

- Open porosity;

- Surface gloss after polishing (The ability of the polished surface to reflect incident light. An important parameter to verify polishing effectiveness);

- Productive yield (First-choice rate during polishing/lapping);

- Output temperature;

- Tool consumption (Indicator of the machine’s overall performance. Premature tool wear signals poor setup or material quality);

- Dimensional variations (Length, width, thickness variations from nominal values);

- Geometry variations (Straightness and squareness deviations of the edges from nominal values);

- Surface variations (Curvature/warping deviations from nominal surface values);

- Visible surface defects (Presence of cracks, scratches, small holes (pitting), etc.);

- Water absorption;

- Breaking strength;

- Flexural strength;

- Impact resistance;

- Abrasion resistance;

- Coefficient of thermal expansion;

- Thermal shock resistance;

- Load-bearing resistance;

- Moisture expansion;

- Frost resistance;

- Chemical attack resistance;

- Stain resistance;

- Lead and cadmium release;

- Coefficient of friction;

- Slurry density; - Slurry density (Density of ceramic suspension at the mill output);

- Measured slurry viscosity;

- Measured grinding residue of the mixture;

- Specific consumption Kcal/I;

- Moisture of atomized powder;

- Particle size distribution of atomized powder;

- Tile density distribution;

- Tile thickness;

- Tile dimensions;

- Current energy consumption;

- Dry breaking load;

- Presence of cracks;

- Dimensional variation after drying;

- Output moisture;

- Output temperature;

- Slab density distribution;

- Crack characteristics;

- Tile thickness;

- Tile dimensions;

- Current energy consumption;

- Dry breaking strength;

- Presence of cracks;

- Dimensional variation after drying;

- Output moisture;

- Output temperature;

- Black core;

- Caliber;

- Shade;

- Flatness;

- Shrinkage;

- Absence of preheating cracks; - Absence of cooling "breaks";

- Absence of cooling stresses;

- Product output temperature;

- Water absorption of final product;

- Breaking load after firing [range];

- Surface appearance;

- Apparent density at output;

- Energy consumption;

- Current consumption - electric I pneumatic I water;

- Tool consumption;

- Output format;

- Input flatness;

- Accuracy of rectified size;

- Optimization of electric I pneumatic I water consumption;

- Production;

- Optimization of tool consumption;

- Flatness;

- Open porosity;

- Optimization of electric I pneumatic I water consumption;

- Output gloss;

- Production;

- Output temperature;

- Optimization of tool consumption.

It should be noted that the processing units and the parameters (input, performance and adjustment) are not limited to the examples shown above. The following paragraphs, listed with alphanumeric references, represent illustrative and non-limiting ways of describing the present invention.

A. A line (100) for production of decorated ceramic products in a continuous cycle, comprising:

- a plurality of processing units (1 , 2, 3, 4, 5, 6), configured to perform respective processing operations in succession, wherein each processing unit includes a system of sensors (S) and a plurality of operating configurations;

- a supervision unit (7) including a processor (8), wherein the supervision unit has access, for each processing unit, to: adjustment parameters, the operation configurations of the processing unit being determined by the values of the adjustment parameters, input parameters, at least some of which are detected by the sensor system, performance parameters, representative of the performance of the processing unit, and has access to a non-volatile memory including machine-readable instructions that cause the processor to: acquire control data including values of the adjustment parameters and input parameters for the plurality of processing units, feed the control data to a machine-trained model, the machine- trained model being trained to derive estimated values for one or more of the performance parameters based on the control data, receive target values for one or more of the performance parameters, temporally correlated with corresponding acquired values of the adjustment parameters and input parameters, generate training data to train the machine-trained model based on the target values and the corresponding values of the adjustment and input parameters.

A1. Line (100) according to paragraph A, wherein the performance parameters include in-line and/or off-line acquired parameters, wherein the off-line parameters are temporally correlated with the corresponding adjustment and input parameters.

A2. Line (100) according to paragraph A1 , wherein at least one of the following conditions applies: the line includes a registration device to record the date and/or time and/or location of the acquisition of the off-line acquired parameters; the line includes a device for reading an identification code uniquely assigned to a specific product, wherein the code contains the date and/or time and/or location of acquisition of the off-line parameters for said specific product.

A3. Line (100) according to any one of paragraphs A-A2, wherein the supervision unit is programmed to assign a unique code to each product on the line and to track the product using said unique code.

A4. Line (100) according to paragraph A3, wherein the line is equipped with a marking device for commanding the generation of said unique code for a product and for initiating training.

A5. Line according to paragraph A4, wherein the unique code contains data representative of the control data and/or performance parameters and/or target values and/or the temporal correlation between said values and parameters.

A6. Line (100) according to any one of the previous paragraphs, wherein the plurality of processing units includes:

- Forming/drying

- Decoration

- firing, wherein one or more of the following conditions applies:

- the input parameters include: moisture content of the outgoing product; and/or moisture content of the incoming product; and/or the type of mixture to be used; and/or the type of decoration; and/or the product thickness;

- the adjustment parameters include: compaction pressure; and/or number of drying cycles; and/or spacing between pressing rollers; and/or temperature profile of the oven above the pressing rollers; combustion air during firing;

- the performance parameters include:

- product thickness; and/or

- moisture of the outgoing product; and/or

- temperature of the outgoing product.

A7. Method for continuous-cycle production of decorated ceramic products, comprising the following steps:

- acquiring, at a supervision unit, control data including values of adjustment and input parameters for the plurality of processing units, at least some of which are detected along the production line on which the ceramic products are made, and wherein the adjustment parameter values determine the operating configurations of the plurality of processing units that carry out respective processing operations in sequence along the production line;

• feeding the control data to a machine-trained model, which is trained to derive estimated values for one or more performance parameters, based on the control data, wherein the performance parameters represent the performance of the processing unit;

• receiving, from the supervisory unit, target values for one or more of the performance parameters, temporally correlated with corresponding acquired values of the control and input parameters;

• generating training data to train the machine-trained model, based on the target values and the corresponding adjustment and input parameter values.

A8. Method according to paragraph A7, wherein the performance parameters include in-line and/or off-line acquired parameters, wherein the off-line parameters are temporally correlated with the corresponding control and input parameters.

A9. Method according to paragraph A8, wherein at least one of the following conditions applies: - the date and/or time and/or location of acquisition of the off-line acquired parameters is recorded via a recording device located along the line;

- an identification code uniquely assigned to a specific product, containing the date and/or time and/or location of acquisition of the off-line parameters for said specific product, is read and recorded via a code-reading device.

A10. Method according to any one of the preceding paragraphs, wherein:

- a unique code is assigned to each product on the line, and the product is tracked along the line using said unique code.

A11. Method according to paragraph A10, wherein the unique code contains data representative of the control data and/or performance parameters and/or target values and/or the temporal correlation between said values and parameters.

Claims

1. A method for making a database for a line (100) for the production of decorated ceramic products in a continuous cycle, the line comprising a plurality of processing units, configured to perform respective processing operations in succession, wherein each processing unit includes a system of sensors and a plurality of operating configurations, the method comprising the following steps:

- preparing a non-volatile memory and a processor;

- by means of the processor, storing in the memory values of adjustment parameters, the operating configurations of the processing unit being determined by values of the adjustment parameters, values of input parameter, at least some of which being detected by the system of sensors, values of performance parameters, representing the performance of the processing unit, time stamps for temporally synchronising the values of the adjustment parameters, input parameters and performance parameters.

2. The method according to claim 1 , wherein the performance parameters include parameters acquired online and/or offline, wherein parameters detected offline are temporally correlated to the corresponding adjustment parameters and input parameters.

3. The method according to claim 2, wherein the parameters acquired offline are entered manually by a user and/or by a device located offline.

4. The method according to claim 2 or 3, wherein at least one of the following conditions occurs:

5. The method according to any one of the preceding claims, wherein a unique code is associated with each product in the line and contains data representing the control data and/or performance parameters and/or target values and/or the temporal correlation between said values and parameters.

6. The method according to claim 5, wherein the code is a bar code, or a QR code, or an RFID label.

7. The method according to any one of the preceding claims, wherein the non-volatile memory and the processor are provided with a knowledge base formed by data representative or descriptive of the operation of the line.

8. The method according to claim 7, wherein the data are collected from machine manuals of the line and/or results of laboratory tests, and/or electrical diagrams, and/or mechanical drawings of parts of machines of the line.

9. The method according to any one of the preceding claims, wherein at least one of the values of the adjustment parameters, of the input parameters and of the performance parameters comes from data relating to the production of the line, the data relating to the production of the line being collected by information systems of the line and/or production management software in the line.

10. The method according to any one of the preceding claims, wherein a machine-trained model is contained in the non-volatile memory, the machine-trained model being trained to derive estimated values for one or more of the performance parameters, depending on the control data.

11 . The method according to any one of the preceding claims, wherein there is a natural language processing engine, NLP, having access to the texts contained in the memory, in which there is a step of processing texts contained in the memory, using NLP.

12. The method according to claim 1 1 , wherein a request text, entered by a user, is processed using the NLP engine, wherein databases can be used by artificial intelligence algorithms to provide natural language responses to said request.

13. The method according to any one of the preceding claims, comprising the following steps:

• integrating a conversational interface operatively connected with the database and with a natural language processing (NLP) engine, the chatbot being configured to receive, in natural language, requests, commands, or queries from a user.

14. The method according to claim 13, wherein the NLP engine is configured to process the requests received by the chatbot and extract relevant information from memory and the database.

15. The method according to any one of claims 13 or 14, wherein the chatbot is configured to return responses in natural language, comprising at least one of the following:

• explanations regarding the status of the production line,

• operational suggestions,

• diagnoses of anomalies or malfunctions,

• alarm notifications.

16. The method according to any one of the preceding claims, wherein the chatbot is configured to generate contextualized operational suggestions based on a plurality of stored information sources, comprising at least one of the following:

• technical documentation,

• real-time acquired process data,

• historical databases,

• predictive or optimization models,

• structured lists of performance parameters.

17. The method according to any one of the preceding claims from 13 to 16, wherein the responses provided by the chatbot comprise at least one of the following informational contents:

• suggestions for modifying operational parameters, • optimized quantitative indications,

• detailed operating procedures,

• estimates of the remaining useful life of components,

• proposals for reducing energy consumption.

18. A computer server, comprising a processor and a non-volatile memory, wherein the memory contains a database relevant to a line for the production of decorated ceramic products in a continuous cycle, the line comprising a plurality of processing units, configured to perform respective processing operations in succession, wherein each processing unit includes a system of sensors and a plurality of operating configurations, the database containing values of adjustment parameters, the operating configurations of the processing unit being determined by values of the adjustment parameters, values of input parameter, at least some of which being detected by the system of sensors, values of performance parameters, representing the performance of the processing unit, time stamps for that data time stamps for temporally synchronising the values of the adjustment parameters, input parameters and performance parameters.