CN118466397A - Monitoring system of numerical control machine tool - Google Patents
Monitoring system of numerical control machine tool Download PDFInfo
- Publication number
- CN118466397A CN118466397A CN202410935037.7A CN202410935037A CN118466397A CN 118466397 A CN118466397 A CN 118466397A CN 202410935037 A CN202410935037 A CN 202410935037A CN 118466397 A CN118466397 A CN 118466397A
- Authority
- CN
- China
- Prior art keywords
- data
- processing parameters
- parameters
- product
- machine tool
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/18—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form
- G05B19/406—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form characterised by monitoring or safety
- G05B19/4065—Monitoring tool breakage, life or condition
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/37—Measurements
- G05B2219/37616—Use same monitoring tools to monitor tool and workpiece
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Manufacturing & Machinery (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- General Factory Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a monitoring system of a numerical control machine tool, which records processing parameters of the numerical control machine tool in the processing process of a die and the quality condition of a final product; according to the recorded and stored numerical control machine tool data, a random forest model is adopted to identify key processing parameters which have obvious influence on the quality condition of the product; quantifying the relation between the key processing parameters and the product quality condition by adopting a Bayesian network model; and according to the relation between the processing parameters and the quality condition of the product, monitoring the current processing parameters and adjusting the processing parameters to ensure that the qualification rate of the product reaches the standard. The invention solves the problems that the prior art is difficult to identify the processing parameters which have obvious influence on the quality of the product, and the relation between the processing parameters and the product percent of pass is difficult to quantify, so that the product quality monitoring efficiency of the numerical control machine tool is low and the processing parameters are controlled inaccurately.
Description
Technical Field
The invention belongs to the technical field of numerical control machine tools, and relates to a monitoring system of a numerical control machine tool.
Background
A numerically controlled machine tool (Numerical Control Machine Tool) is an automated machine tool that controls the operation of the machine tool by a pre-programmed program. Compared with the traditional manual operation machine tool, the numerical control machine tool has higher precision, repeatability and production efficiency. The operation process of the numerical control machine tool is controlled by a computer system, and a user can use computer software to generate a machine tool machining path and convert the path into a numerical control program. The numerical control program contains various instructions to guide the machine tool to execute machining operations such as cutting, drilling, milling, turning and the like.
For injection mold processing, a numerical control machine tool needs to adjust a plurality of parameters to ensure the quality of products. However, due to the numerous processing parameters, it is difficult to identify the processing parameters that have a significant impact on the quality of the product, and it is difficult to quantify the relationship between the processing parameters and the product yield, resulting in the problems of low product quality monitoring efficiency and inaccurate control of the processing parameters for the numerical control machine.
Disclosure of Invention
The invention provides a monitoring system of a numerical control machine tool, which aims to realize real-time monitoring and post-tracing of product quality by recording data of the numerical control machine tool in a die machining process and combining a statistical process control theory (SPC), and solves the problems of low product quality monitoring efficiency and inaccurate machining parameter control of the numerical control machine tool.
The aim of the invention can be achieved by the following technical scheme:
the application provides a monitoring system of a numerical control machine tool, which comprises a data recording module, a data analysis module and a parameter adjustment module, wherein the data recording module, the data analysis module and the parameter adjustment module are in communication connection, and the monitoring system comprises the following components:
The data recording module is used for recording processing parameters of the numerical control machine tool in the processing process of the die and the quality condition of a final product, wherein the quality condition of the product comprises qualification and disqualification;
the data analysis module is used for analyzing the relation between the processing parameters and the product quality conditions according to the recorded and stored numerical control machine tool data, and comprises the following steps:
s1, identifying key processing parameters which have obvious influence on the quality condition of a product by adopting a random forest model;
s2, quantifying the relation between the key processing parameters and the product quality condition by adopting a Bayesian network model;
and the parameter adjustment module is used for monitoring the current processing parameters and adjusting the processing parameters according to the relation between the processing parameters and the quality condition of the product so as to ensure that the qualification rate of the product reaches the standard.
Further, the machining parameters include cutting speed, feed speed, spindle speed, cutting depth and width, tool path, cooling and lubrication, tool type and material.
Specifically, in step S1, the random forest model includes the following construction steps:
s11, randomly extracting L samples from a data set to serve as a training set;
s12, randomly and repeatedly extracting N processing parameter variables as partition points, and determining the optimal partition points of the product quality condition variables by utilizing the coefficient of the radix to divide so as to generate a decision tree;
S13, repeating the operations of the steps S11-S12 for M times to obtain M decision trees and generate a random forest;
S14, calculating a prediction error of each decision tree in the random forest on the out-of-bag data;
S15, adding random disturbance of processing parameter variables in the out-of-bag data, and calculating a prediction error of each decision tree of the random forest on the random disturbance;
s16, sorting the importance of the processing parameter variables according to the prediction error of the random forest on the out-of-bag data;
S17, selecting n processing parameter variables before importance ranking as key processing parameters.
Further, in step S13, the determination of the M values of the M decision trees is specifically as follows:
generating random forests with different decision tree numbers by taking O decision trees as step sizes;
calculating a prediction error of the random forest on the test sample;
And selecting a random forest with the minimum prediction error, and determining the number M of decision trees.
Further, in step S16, the importance of the processing parameter variables is ordered according to the prediction error of the random forest on the out-of-bag data, and the calculation formula is as follows:
，
wherein: PIM represents the degree of importance of the process parameter variables; m is the number of decision trees in the random forest; Representing the prediction error of the kth decision tree on the data outside the bag added with the random disturbance of the ith processing parameter variable; representing the prediction error of the kth decision on the out-of-bag data without random disturbance.
Further, in step S2, the bayesian network model includes the following construction steps:
s21, determining a model connection relation: the key processing parameters are used as father nodes, the product quality condition is used as child nodes, and the father nodes and the child nodes are connected to form a Bayesian network model structure;
S22, data parameterization: estimating key processing parameters and model parameters of product quality conditions according to preset standards;
S23, constructing a conditional probability table: and constructing a condition probability table of correlation between each key processing parameter and the model parameter of the product quality condition according to the model parameters obtained by the data parameterization.
Further, in step S22, the key processing parameters and the model parameters of the transmission quality data are estimated according to the preset standard, specifically, the key processing parameters are divided into a plurality of class parameters by adopting a K-means clustering method; the "pass" and "fail" in the product quality case are divided into two logic parameters, yes and no, respectively.
Further, the method for classifying the key processing parameters into a plurality of grade parameters by adopting the K-means clustering method comprises the following steps:
t1, normalization treatment: normalizing the key processing parameters;
t2, initializing a clustering center: randomly selecting a plurality of data objects in a data space as an initial clustering center;
t3, initializing a cluster: the Euclidean distance between all data objects and an initial clustering center is calculated, and each data object is divided into the initial clustering center with the minimum Euclidean distance to form an initial clustering cluster;
and T4, updating a clustering center: calculating the average value of the data objects of each initial cluster, and taking the average value as a new cluster center;
t5, determining a final cluster: repeatedly calculating Euclidean distances between all the data objects and the new clustering center again, dividing each data object into the new clustering center with the minimum Euclidean distance to form a new clustering cluster; and calculating the average value of each new cluster again to serve as a new cluster center until the new cluster center is not changed any more, determining a final cluster, and determining the grade parameter corresponding to each final cluster according to the key processing parameter data characteristics of the final cluster.
Further, in step T1, the normalization process has a calculation formula:
，
Wherein Y i is the normalized ith key processing parameter data sample; k i is an ith key processing parameter data sample; k max represents the maximum value of all critical process parameter data samples; k min represents the minimum of all critical process parameter data samples.
Further, in step T3, the euclidean distance has a calculation formula:
，
Wherein d (x, C i) represents a Euclidean distance function; x is the normalized data object; c i represents an ith cluster center; n is the number of samples of the dataset; x j represents the data object of the j-th variable in the dataset; c ij represents the ith cluster center of the jth variable.
Further, in step S23, a conditional probability table is constructed for correlation between each key processing parameter and the model parameter of the product quality condition according to the model parameter obtained by the data parameterization, specifically: for the parameters of each variable, under the condition of giving the parameters to the father node, calculating the joint conditional probability of the parameters at the father node, wherein the formula is as follows:
，
，
Where P (U) is expressed as a set of joint conditional probabilities of node u=b 1,B2,……,Bn; pa (B i) is the parent node set of node B i in the bayesian network; p (B i) represents the prior probability of the target node B i.
Further, the Bayesian network model also adopts residual analysis or model fitting degree to evaluate the prediction performance of the model.
Further, in the parameter adjustment module, according to the relationship between the processing parameter and the quality condition of the product, the current processing parameter is monitored and the processing parameter is adjusted, including the following steps:
Acquiring key processing parameters of a time sequence at a specific time step;
inputting the key processing parameters into a Bayesian network model after training, and predicting the quality condition of the product;
and when the qualification rate of the product quality condition is lower than a preset qualification rate threshold value, adjusting the key processing parameters.
The invention has the beneficial effects that:
Recording processing parameters in the processing process of the die by a numerical control machine tool and the quality condition of a final product; according to the recorded and stored numerical control machine tool data, a random forest model is adopted to identify key processing parameters which have obvious influence on the quality condition of the product; quantifying the relation between the key processing parameters and the product quality condition by adopting a Bayesian network model; and according to the relation between the processing parameters and the quality condition of the product, monitoring the current processing parameters and adjusting the processing parameters to ensure that the qualification rate of the product reaches the standard. The invention solves the problems that the prior art is difficult to identify the processing parameters which have obvious influence on the quality of the product, and the relation between the processing parameters and the product percent of pass is difficult to quantify, so that the product quality monitoring efficiency of the numerical control machine tool is low and the processing parameters are controlled inaccurately.
Drawings
The present invention is further described below with reference to the accompanying drawings for the convenience of understanding by those skilled in the art.
Fig. 1 is a block diagram of a monitoring system of a numerical control machine tool according to the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the present invention for achieving the intended purpose, the following detailed description will refer to the specific implementation, structure, characteristics and effects according to the present invention with reference to the accompanying drawings and preferred embodiments.
Referring to fig. 1, the application provides a monitoring system of a numerical control machine tool, which comprises a data recording module, a data analysis module and a parameter adjustment module, wherein the data recording module, the data analysis module and the parameter adjustment module are in communication connection, and the monitoring system comprises:
The data recording module is used for recording processing parameters of the numerical control machine tool in the processing process of the die and the quality condition of a final product, wherein the quality condition of the product comprises qualification and disqualification;
In the present embodiment, when a numerical control machine tool is used to machine a die, a plurality of parameters need to be adjusted to ensure machining accuracy, surface quality, and machining efficiency. The following key parameters and their adjustment points are as follows:
1) Cutting speed: the cutting speed depends on the type of material and the tool material. The cutting speed is generally high during processing of the injection mold to avoid melting of the material, and the speed range is generally 150-300 m/min.
2) Feed rate: the feed rate affects the cutting load and surface finish. For plastic materials, a feed speed of 200-800 mm/min is typically used, the specific speed being adjusted according to the mould complexity and the type of tool.
3) Spindle rotational speed: the number of revolutions per minute of the spindle is expressed in Revolutions Per Minute (RPM). The spindle rotational speed is set in relation to the cutting speed and the workpiece diameter.
4) Depth of cut and width: the depth of cut and width affect cutting force and machining efficiency. For injection mold tooling, shallow cutting depths and smaller cutting widths are typically selected to reduce material deformation and tool wear.
5) Cutter path: the tool path is rationally planned to reduce cutting vibrations and stress concentrations, especially when machining complex geometries. Common path strategies include parallel cutting, helical cutting, and contour cutting.
6) Cooling and lubricating: an appropriate amount of coolant or lubricant is used to prevent overheating of the cutting area. Air cooling or water-soluble cutting fluids are commonly used to avoid plastic material deformation or sticking to the tool.
7) Cutter type and material: the proper cutter material and geometry are selected. The hard alloy or the high-speed steel tool is suitable for processing an injection mold. The sharpness of the cutter has an important influence on the processing quality.
Thus, further, the machining parameters include cutting speed, feed speed, spindle speed, cutting depth and width, tool path, cooling and lubrication, tool type and material.
The data analysis module is used for analyzing the relation between the processing parameters and the product quality conditions according to the recorded and stored numerical control machine tool data, and comprises the following steps:
s1, identifying key processing parameters which have obvious influence on the quality condition of a product by adopting a random forest model;
s2, quantifying the relation between the key processing parameters and the product quality condition by adopting a Bayesian network model;
and the parameter adjustment module is used for monitoring the current processing parameters and adjusting the processing parameters according to the relation between the processing parameters and the quality condition of the product so as to ensure that the qualification rate of the product reaches the standard.
In this embodiment, since there are too many processing parameters involved, the degree of influence on the quality of the product is different, and it is difficult for people to identify key processing parameters with a larger degree of influence therefrom. For this reason, in this embodiment, the processing parameters are used as explanatory variables, the product quality condition is used as response variables, a random forest model is constructed to rank the importance of the processing parameters, and the processing parameters with the importance ranked forward are selected as key processing parameters. Random forests are a classical machine learning algorithm that can be used for feature extraction and variable importance ranking. Random forests are trained and predicted by building multiple decision trees, each modeled using only a portion of the training data and a portion of the features. A strong classification or regression model is finally obtained by constructing a plurality of decision trees, wherein each decision tree has weight contribution. The random forest may give a ranking of the importance of each input feature in the model, the importance scores of these features being calculated from their degree of influence in the random forest on the model's prediction results. Typically, feature importance scores in random forests are calculated based on out-of-bag errors.
Specifically, in step S1, the random forest model includes the following construction steps:
s11, randomly extracting L samples from a data set to serve as a training set;
s12, randomly and repeatedly extracting N processing parameter variables as partition points, and determining the optimal partition points of the product quality condition variables by utilizing the coefficient of the radix to divide so as to generate a decision tree;
S13, repeating the operations of the steps S11-S12 for M times to obtain M decision trees and generate a random forest;
S14, calculating a prediction error of each decision tree in the random forest on the out-of-bag data;
the data outside the bag refers to data which was not used in training other than model training. In machine learning and statistical modeling, out-of-bag data is one of the important indicators used to evaluate the performance and generalization ability of a model. The effect of the out-of-bag data is to verify the behavior of the model in the face of the unseen sample. In the training process, only partial data sets are used for parameter estimation and model fitting, and the outside-bag data provides an independent data set which can be used for evaluating the prediction accuracy and reliability of the model. In some machine learning algorithms, such as random forests, gradient lifting trees, cross-validation, etc., model evaluation and tuning can be performed by using out-of-bag data to better estimate the performance of the model and help identify problems such as over-fitting or under-fitting.
S15, adding random disturbance of processing parameter variables in the out-of-bag data, and calculating a prediction error of each decision tree of the random forest on the random disturbance;
It should be noted that adding random perturbation of the data outside the bag means that some degree of random modification or perturbation is performed on the data outside the bag when training the model. The purpose of this perturbation is to increase the generalization ability of the model to unknown data and to reduce the overfitting to training data. By adding random disturbance of the data outside the bag, the model can learn and adapt to different types of data changes better, and the dependence of the model on specific samples and specific features in the training set is reduced, so that the expressive capacity of the model in a real scene is improved. Meanwhile, reasonable data disturbance outside the bag can effectively prevent the phenomenon of overfitting and improve the generalization capability of the model.
S16, sorting the importance of the processing parameter variables according to the prediction error of the random forest on the out-of-bag data;
S17, selecting n processing parameter variables before importance ranking as key processing parameters.
It should be noted that in this embodiment, the n processing parameter variables before the importance ranking, n, may depend on the value by which the prediction error starts to become larger after the model joins the nth marketing campaign information variable.
Further, in step S13, the determination of the M values of the M decision trees is specifically as follows:
generating random forests with different decision tree numbers by taking O decision trees as step sizes;
calculating a prediction error of the random forest on the test sample;
And selecting a random forest with the minimum prediction error, and determining the number M of decision trees.
Further, in step S16, the importance of the processing parameter variables is ordered according to the prediction error of the random forest on the out-of-bag data, and the calculation formula is as follows:
，
wherein: PIM represents the degree of importance of the process parameter variables; m is the number of decision trees in the random forest; Representing the prediction error of the kth decision tree on the data outside the bag added with the random disturbance of the ith processing parameter variable; representing the prediction error of the kth decision on the out-of-bag data without random disturbance.
In this embodiment, by quantifying the relationship between the key processing parameters and the product quality condition in the numerical control machine tool die processing using the bayesian network model, the important influence of the key processing parameters on the product yield can be more deeply understood. A bayesian network is a probabilistic graph model that can be used to represent the dependencies between variables and analyze them through probabilistic reasoning. In the relation between the key processing parameters and the product quality conditions, the key processing parameters can be used as input variables, the product quality conditions can be used as output variables, and the product qualification rate can be predicted by analyzing the dependency relation between the key processing parameters and the product quality conditions.
In particular, a bayesian network model is trained from historic data that can learn probability distributions between key process parameters and product quality scenarios. By learning these probability distributions, it is possible to understand how much the product is acceptable under different key process parameters, thereby quantifying the relationship between them.
After a trained Bayesian network model is obtained, it can be used to predict product yield. By observing the change of the key processing parameters within a period of time, the observation data can be input into a Bayesian network model, and probability reasoning is performed, so that probability distribution of the product quality condition is obtained. These probability distributions can help us evaluate the risk of product failure and adjust key process parameters accordingly to prevent or solve potential product quality problems.
Further, in step S2, the bayesian network model includes the following construction steps:
s21, determining a model connection relation: the key processing parameters are used as father nodes, the product quality condition is used as child nodes, and the father nodes and the child nodes are connected to form a Bayesian network model structure;
In the process of constructing the Bayesian network model, an explanatory variable is generally taken as a father node, a response variable is taken as a child node, and the influence relationship among the variables is represented. In the embodiment, the key processing parameters are used as father nodes, the product quality condition is used as child nodes, and the father nodes and the child nodes are connected to form a Bayesian network model structure, and in the structure, the key processing parameters have direct influence on the product quality condition.
S22, data parameterization: estimating key processing parameters and model parameters of product quality conditions according to preset standards;
In this embodiment, the Bayesian network is a graph model for modeling and inferring probabilistic relationships. Parameterization is an important task in bayesian networks, whose purpose is to estimate parameters of probability distributions in the network from observed data. These parameters are used to describe the conditional probabilities and edge probabilities between variables. The obtained test data sample values are high or low, and the embodiment adopts preset standard to estimate key processing parameters and model parameters of transmission quality data in the parameterization process:
Further, in step S22, the parameters of the key processing parameters and the transmission quality data are estimated according to the preset standard, specifically, the key processing parameters are divided into a plurality of grade parameters by adopting a K-means clustering method; the "pass" and "fail" in the product quality case are divided into two logic parameters, yes and no, respectively.
In this embodiment, the product quality case is divided into two logic parameters, namely "yes" and "no" which can be used to determine whether an event occurs, for example, for the product quality case, "yes" indicates a probability that the product quality is acceptable, and "no" indicates a probability that the product quality is unacceptable.
Further, the method for classifying the key processing parameters into a plurality of grade parameters by adopting the K-means clustering method comprises the following steps:
t1, normalization treatment: normalizing the key processing parameters;
in this embodiment, the purpose of normalizing the key processing parameters is to unify the data ranges between different features so that they have the same scale. Therefore, the problem of unbalanced weight caused by different data ranges among different features can be avoided, and each feature can be properly reflected in the clustering process. Through normalization processing, the dimensional influence among the data can be eliminated, so that the data can be more easily compared and analyzed, and the accuracy and the efficiency of clustering are improved. In addition, the normalization processing can further enable the clustering algorithm to be more stable and faster in convergence speed, and is beneficial to obtaining more accurate and reliable clustering results. Therefore, before the cluster analysis of the key processing parameters is performed, the data is generally normalized to achieve a better clustering effect.
T2, initializing a clustering center: randomly selecting a plurality of data objects in a data space as an initial clustering center;
In this embodiment, when initializing a cluster center, it is a common practice to randomly select a number of data objects in a data space as the initial cluster center. The reason for this is that by randomly selecting the initial cluster center, it can be ensured that the algorithm is not affected by the position of the specific initial cluster center, thereby avoiding sinking into a locally optimal solution. By randomly selecting the data objects in the data space as the initial clustering center, the algorithm has certain randomness, is helpful for exploring different aspects of the data, can increase the robustness and reliability of the algorithm, and improves the quality of the final clustering result. Randomly selecting data objects as initial cluster centers is a common and efficient initialization method.
T3, initializing a cluster: the Euclidean distance between all data objects and an initial clustering center is calculated, and each data object is divided into the initial clustering center with the minimum Euclidean distance to form an initial clustering cluster;
and T4, updating a clustering center: calculating the average value of the data objects of each initial cluster, and taking the average value as a new cluster center;
t5, determining a final cluster: repeatedly calculating Euclidean distances between all the data objects and the new clustering center again, dividing each data object into the new clustering center with the minimum Euclidean distance to form a new clustering cluster; and calculating the average value of each new cluster again to serve as a new cluster center until the new cluster center is not changed any more, determining a final cluster, and determining the grade parameter corresponding to each final cluster according to the key processing parameter data characteristics of the final cluster.
In this embodiment, K-means clustering is a common unsupervised learning algorithm used to divide data points into different clusters. By calculating the similarity between data points, the K-means algorithm can classify the data points into clusters that are closest together. The method is helpful for discovering hidden modes and structures in the data, so that cluster analysis and classification of the data are realized. The key process parameters are divided into different clusters using the K-means algorithm, each cluster representing key process parameters with similar characteristics.
Further, in step T1, the normalization process has a calculation formula:
，
Wherein Y i is the normalized ith key processing parameter data sample; k i is an ith key processing parameter data sample; k max represents the maximum value of all critical process parameter data samples; k min represents the minimum of all critical process parameter data samples.
Further, in step T3, the euclidean distance has a calculation formula:
，
Wherein d (x, C i) represents a Euclidean distance function; x is the normalized data object; c i represents an ith cluster center; n is the number of samples of the dataset; x j represents the data object of the j-th variable in the dataset; c ij represents the ith cluster center of the jth variable.
S23, constructing a conditional probability table: and constructing a condition probability table of correlation between each key processing parameter and the model parameter of the product quality condition according to the model parameters obtained by the data parameterization.
In the present embodiment, the conditional probability table is a table for representing the conditional probability relation between variables, which shows the probability that the variables occur given certain conditions. Assuming that two discrete variables A and B are present, the conditional probability table will be presented in the form of a table, wherein the rows represent the values of variable A and the columns represent the values of variable B. Each cell in the table represents the probability that a certain value of a will occur given a certain value of B. The purpose of the conditional probability table is to describe the conditional probability relationship between the and the metric variables, which can be used to infer, predict and analyze the dependency relationship between the variables. By observing and analyzing the conditional probability tables, correlations between variables can be known and efficient reasoning and decision-making can be performed based on these information.
When conditional probabilities with multiple parent nodes are involved, an expanded form of conditional probability tables, referred to as joint conditional probability tables or joint probability tables, is typically used. A joint conditional probability table is a table for describing the relationship between a plurality of variables, which considers the combination of values of all parent nodes and shows the probability of child nodes given the values of these parent nodes. In the joint conditional probability table, each column represents the value of one child node, and each value combination of the parent nodes corresponds to one row. Each cell represents the probability of a child node given a combination of parent node values. The joint conditional probability table is very useful for modeling complex probability relationships and making inferences. By observing and analyzing the joint condition probability table, the dependency relationship among a plurality of variables can be known, and effective probability reasoning and decision can be performed.
Further, in step S23, a conditional probability table is constructed for correlation between each key processing parameter and the model parameter of the product quality condition according to the model parameter obtained by the data parameterization, specifically: for the parameters of each variable, under the condition of giving the parameters to the father node, calculating the joint conditional probability of the parameters at the father node, wherein the formula is as follows:
，
，
Where P (U) is expressed as a set of joint conditional probabilities of node u=b 1,B2,……,Bn; pa (B i) is the parent node set of node B i in the bayesian network; p (B i) represents the prior probability of the target node B i.
Further, the Bayesian network model also adopts residual analysis or model fitting degree to evaluate the prediction performance of the model.
In bayesian network models, residual analysis and model fitness assessment are often used to assess the prediction accuracy and fitness of the model, which can help to understand the performance and predictive ability of the model on training data.
Residual analysis is to evaluate the accuracy of a model by calculating the difference between the actual observations and the model predictions. For bayesian network models, the residual (actual observations minus model predictions) for each observed variable may be calculated and then the distribution and pattern of the residual analyzed. If the residual meets certain assumptions (e.g., obeys normal distribution) and there is no apparent structure or trend, this may indicate that the prediction accuracy of the model is high. Conversely, if there is a systematic pattern, trend, or outlier in the residual, further improvement of the model or correction data may be required.
Model fitness evaluation is to evaluate the fitness of a model based on its behavior on training data. Common metrics include Log-Likelihood function values (Log-Likelihood) and bayesian information criteria (Bayesian Information Criterion, BIC). The log likelihood function value measures the fitting degree of the model on the observed data, and the higher the numerical value is, the better the fitting degree is. BIC is an index that comprehensively considers the complexity and the fitting degree of the model, and lower BIC values indicate that the model balances the complexity and the fitting degree better.
In addition to residual analysis and model fitness evaluation, there are many other methods that can be used to evaluate the predictive accuracy of bayesian network models, such as cross-validation, leave-one-out, etc.
Further, in the parameter adjustment module, according to the relationship between the processing parameter and the quality condition of the product, the current processing parameter is monitored and the processing parameter is adjusted, including the following steps:
Acquiring key processing parameters of a time sequence at a specific time step;
inputting the key processing parameters into a Bayesian network model after training, and predicting the quality condition of the product;
and when the qualification rate of the product quality condition is lower than a preset qualification rate threshold value, adjusting the key processing parameters.
The invention has the beneficial effects that:
Recording processing parameters in the processing process of the die by a numerical control machine tool and the quality condition of a final product; according to the recorded and stored numerical control machine tool data, a random forest model is adopted to identify key processing parameters which have obvious influence on the quality condition of the product; quantifying the relation between the key processing parameters and the product quality condition by adopting a Bayesian network model; and according to the relation between the processing parameters and the quality condition of the product, monitoring the current processing parameters and adjusting the processing parameters to ensure that the qualification rate of the product reaches the standard. The invention solves the problems that the prior art is difficult to identify the processing parameters which have obvious influence on the quality of the product, and the relation between the processing parameters and the product percent of pass is difficult to quantify, so that the product quality monitoring efficiency of the numerical control machine tool is low and the processing parameters are controlled inaccurately.
The present invention is not limited in any way by the above-described preferred embodiments, but is not limited to the above-described preferred embodiments, and any person skilled in the art will appreciate that the present invention can be embodied in the form of a program for carrying out the method of the present invention, while the above disclosure is directed to equivalent embodiments capable of being modified or altered in some ways, it is apparent that any modifications, equivalent variations and alterations made to the above embodiments according to the technical principles of the present invention fall within the scope of the present invention.
Claims (10)
1. A monitoring system of a numerical control machine tool is characterized in that: the system comprises a data recording module, a data analysis module and a parameter adjustment module, wherein the data recording module, the data analysis module and the parameter adjustment module are in communication connection, and the system comprises the following components:
The data recording module is used for recording processing parameters of the numerical control machine tool in the processing process of the die and the quality condition of a final product, wherein the quality condition of the product comprises qualification and disqualification;
the data analysis module is used for analyzing the relation between the processing parameters and the product quality conditions according to the recorded and stored numerical control machine tool data, and comprises the following steps:
s1, identifying key processing parameters which have obvious influence on the quality condition of a product by adopting a random forest model;
s2, quantifying the relation between the key processing parameters and the product quality condition by adopting a Bayesian network model;
and the parameter adjustment module is used for monitoring the current processing parameters and adjusting the processing parameters according to the relation between the processing parameters and the quality condition of the product so as to ensure that the qualification rate of the product reaches the standard.
2. The system for monitoring a numerically-controlled machine tool according to claim 1, wherein: the machining parameters include cutting speed, feed speed, spindle speed, cutting depth and width, tool path, cooling and lubrication, and tool type and material.
3. The system for monitoring a numerically-controlled machine tool according to claim 1, wherein: in step S1, the random forest model includes the following construction steps:
s11, randomly extracting L samples from a data set to serve as a training set;
s12, randomly and repeatedly extracting N processing parameter variables as partition points, and determining the optimal partition points of the product quality condition variables by utilizing the coefficient of the radix to divide so as to generate a decision tree;
S13, repeating the operations of the steps S11-S12 for M times to obtain M decision trees and generate a random forest;
S14, calculating a prediction error of each decision tree in the random forest on the out-of-bag data;
S15, adding random disturbance of processing parameter variables in the out-of-bag data, and calculating a prediction error of each decision tree of the random forest on the random disturbance;
s16, sorting the importance of the processing parameter variables according to the prediction error of the random forest on the out-of-bag data;
S17, selecting n processing parameter variables before importance ranking as key processing parameters.
4. A monitoring system for a numerically controlled machine tool as in claim 3, wherein: in step S13, the determination of the M values of the M decision trees is specifically as follows:
generating random forests with different decision tree numbers by taking O decision trees as step sizes;
calculating a prediction error of the random forest on the test sample;
And selecting a random forest with the minimum prediction error, and determining the number M of decision trees.
5. A monitoring system for a numerically controlled machine tool as in claim 3, wherein: in step S16, the importance of the processing parameter variables is ordered according to the prediction error of the random forest on the out-of-bag data, and the calculation formula is as follows:
，
wherein: PIM represents the degree of importance of the process parameter variables; m is the number of decision trees in the random forest; Representing the prediction error of the kth decision tree on the data outside the bag added with the random disturbance of the ith processing parameter variable; representing the prediction error of the kth decision on the out-of-bag data without random disturbance.
6. The system for monitoring a numerically-controlled machine tool according to claim 1, wherein: in step S2, the bayesian network model includes the following construction steps:
s21, determining a model connection relation: the key processing parameters are used as father nodes, the product quality condition is used as child nodes, and the father nodes and the child nodes are connected to form a Bayesian network model structure;
S22, data parameterization: estimating key processing parameters and model parameters of product quality conditions according to preset standards;
S23, constructing a conditional probability table: and constructing a condition probability table of correlation between each key processing parameter and the model parameter of the product quality condition according to the model parameters obtained by the data parameterization.
7. The system for monitoring a numerically-controlled machine tool according to claim 6, wherein: in step S22, the key processing parameters and the model parameters of the transmission quality data are estimated according to a preset standard, specifically, the key processing parameters are divided into a plurality of grade parameters by adopting a K-means clustering method; the "pass" and "fail" in the product quality case are divided into two logic parameters, yes and no, respectively.
8. The system for monitoring a numerically-controlled machine tool according to claim 7, wherein: the key processing parameters are divided into a plurality of grade parameters by adopting a K-means clustering method, and the method comprises the following steps:
Normalization: normalizing the key processing parameters;
Initializing a clustering center: randomly selecting a plurality of data objects in a data space as an initial clustering center;
initializing a cluster: the Euclidean distance between all data objects and an initial clustering center is calculated, and each data object is divided into the initial clustering center with the minimum Euclidean distance to form an initial clustering cluster;
updating a clustering center: calculating the average value of the data objects of each initial cluster, and taking the average value as a new cluster center;
Determining a final cluster: repeatedly calculating Euclidean distances between all the data objects and the new clustering center again, dividing each data object into the new clustering center with the minimum Euclidean distance to form a new clustering cluster; and calculating the average value of each new cluster again to serve as a new cluster center until the new cluster center is not changed any more, determining a final cluster, and determining the grade parameter corresponding to each final cluster according to the key processing parameter data characteristics of the final cluster.
9. The system for monitoring a numerically-controlled machine tool according to claim 6, wherein: in step S23, a conditional probability table is constructed for correlation between each key processing parameter and the model parameter of the product quality condition according to the model parameter obtained by the data parameterization, specifically: for the parameters of each variable, under the condition of giving the parameters to the father node, calculating the joint conditional probability of the parameters at the father node, wherein the formula is as follows:
，
，
Where P (U) is expressed as a set of joint conditional probabilities of node u=b 1,B2,……,Bn; pa (B i) is the parent node set of node B i in the bayesian network; p (B i) represents the prior probability of the target node B i.
10. The system for monitoring a numerically-controlled machine tool according to claim 1, wherein: in the parameter adjustment module, according to the relation between the processing parameter and the quality condition of the product, the current processing parameter is monitored and adjusted, and the method comprises the following steps:
Acquiring key processing parameters of a time sequence at a specific time step;
inputting the key processing parameters into a Bayesian network model after training, and predicting the quality condition of the product;
and when the qualification rate of the product quality condition is lower than a preset qualification rate threshold value, adjusting the key processing parameters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410935037.7A CN118466397B (en) | 2024-07-12 | 2024-07-12 | A monitoring system for CNC machine tools |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410935037.7A CN118466397B (en) | 2024-07-12 | 2024-07-12 | A monitoring system for CNC machine tools |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118466397A true CN118466397A (en) | 2024-08-09 |
| CN118466397B CN118466397B (en) | 2024-10-29 |
Family
ID=92164294
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410935037.7A Active CN118466397B (en) | 2024-07-12 | 2024-07-12 | A monitoring system for CNC machine tools |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118466397B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119537324A (en) * | 2025-01-23 | 2025-02-28 | 珠海飞企耀点科技有限公司 | A multi-level quality document management method based on big data |
| CN119648256A (en) * | 2025-02-19 | 2025-03-18 | 珠海飞企耀点科技有限公司 | A pharmaceutical quality data management platform based on microservice architecture |
| CN119918997A (en) * | 2024-12-31 | 2025-05-02 | 江油市扬帆模具科技有限公司 | Intelligent precision parts full process processing control system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109613891A (en) * | 2018-11-06 | 2019-04-12 | 北京航空航天大学 | Method, device and equipment for identifying key parameters in NC machining process |
| CN113646714A (en) * | 2019-04-29 | 2021-11-12 | 西门子股份公司 | Processing parameter setting method and device for production equipment and computer readable medium |
| US20230206054A1 (en) * | 2021-12-23 | 2023-06-29 | Luminide, Inc. | Expedited Assessment and Ranking of Model Quality in Machine Learning |
-
2024
- 2024-07-12 CN CN202410935037.7A patent/CN118466397B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109613891A (en) * | 2018-11-06 | 2019-04-12 | 北京航空航天大学 | Method, device and equipment for identifying key parameters in NC machining process |
| CN113646714A (en) * | 2019-04-29 | 2021-11-12 | 西门子股份公司 | Processing parameter setting method and device for production equipment and computer readable medium |
| US20230206054A1 (en) * | 2021-12-23 | 2023-06-29 | Luminide, Inc. | Expedited Assessment and Ranking of Model Quality in Machine Learning |
Non-Patent Citations (1)
| Title |
|---|
| 蒋金瑜: "面向离散型制造的生产数据采集与产品质量预测方法研究", 万方学位论文数据库, 13 September 2021 (2021-09-13) * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119918997A (en) * | 2024-12-31 | 2025-05-02 | 江油市扬帆模具科技有限公司 | Intelligent precision parts full process processing control system |
| CN119537324A (en) * | 2025-01-23 | 2025-02-28 | 珠海飞企耀点科技有限公司 | A multi-level quality document management method based on big data |
| CN119648256A (en) * | 2025-02-19 | 2025-03-18 | 珠海飞企耀点科技有限公司 | A pharmaceutical quality data management platform based on microservice architecture |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118466397B (en) | 2024-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN118466397B (en) | A monitoring system for CNC machine tools | |
| US9336484B1 (en) | System and method for outlier detection via estimating clusters | |
| US7363103B2 (en) | Method and apparatus for generating machining information, and computer product | |
| CN114492198B (en) | A cutting force prediction method based on improved PSO algorithm assisted SVM algorithm | |
| Parwal et al. | Machine learning based approach for process supervision to predict tool wear during machining | |
| JP2008546046A (en) | Mahalanobis distance genetic algorithm method and system | |
| CN117688458A (en) | A method and system for automatic optimization of injection mold process parameters | |
| CN118917215B (en) | Gear precision correction method, device, equipment and storage medium | |
| CN118884883B (en) | Intelligent numerical control processing monitoring method and system based on industrial Internet of things | |
| CN118690925B (en) | Quick mold changing method and system | |
| CN119065338B (en) | Precision Machining Technology Optimization Method for Machine Tool Spindle System | |
| CN119807937B (en) | Data collection and analysis method and system for 3D printing fixtures based on big data | |
| CN119692154A (en) | Method, device, electronic device and storage medium for determining process parameter data | |
| CN112966449A (en) | Manufacturing process quality defect root cause diagnosis method based on variable selection and probabilistic neural network | |
| CN119477036A (en) | Multi-feature quality prediction method based on improved Hippo algorithm to optimize FFN | |
| CN119806037A (en) | Control method for dynamic adjustment of drill and reamer seat on Y-axis of high-precision CNC lathe | |
| CN115981236B (en) | Method for predicting energy consumption in turning process of numerical control lathe | |
| Shurrab et al. | Tool wear prediction in computer numerical control milling operations via machine learning | |
| CN118195357A (en) | A key performance prediction method and intelligent management platform for intelligent manufacturing workshops under uncertain environments | |
| CN111730412A (en) | Ant colony optimization algorithm-based micro milling cutter wear state monitoring method of support vector machine | |
| Hung et al. | Hybrid Approach Combining Simulated Annealing and Deep Neural Network Models for Diagnosing and Predicting Potential Failures in Smart Manufacturing. | |
| Ferreiro et al. | A Bayesian network for burr detection in the drilling process | |
| Cuadrado-Gallego et al. | Analogies and differences between machine learning and expert based software project effort estimation | |
| CN118586782A (en) | A mathematical and statistical analysis method based on quality management decision-making system | |
| JP7697025B2 (en) | Method and system for industrial change-point detection - Patents.com |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |