CN116796142A - A tool wear status identification method, system, electronic device and storage medium - Google Patents

Abstract

The invention provides a tool wear status identification method, system, electronic equipment and storage medium. The method includes: collecting tool wear vibration signals; inputting the tool wear vibration signals into a preset convolutional neural network to perform feature extraction and one-time reduction. dimension to obtain vibration signal characteristics; based on the preset local linear embedding rules, perform a secondary dimensionality reduction on the vibration signal characteristics to obtain an observation sequence; input the observation sequence into a pre-trained hidden Markov model to perform tool Wear status identification and tool wear status prediction, obtaining identification results and prediction results; feeding back the identification results and prediction results to the associated terminal equipment. The tool wear status identification method provided by the present invention can automatically realize feature extraction of tool wear vibration signals, has high accuracy and strong real-time performance, and the tool wear status identification accuracy of this method is high.

Description

A tool wear status identification method, system, electronic device and storage medium

Technical field

The present invention relates to the field of signal recognition technology, and in particular to a tool wear status recognition method, system, electronic equipment and storage medium.

Background technique

Tool wear and tear is a common fault in CNC machining. The processing problems caused by it will directly affect the processing quality, production efficiency and processing cost of the product. In serious cases, it may even cause damage to the machine tool itself, resulting in relatively significant economic losses. Traditional tool wear status monitoring and detection methods mainly rely on offline measurement or judgment by skilled workers through personal experience. On the one hand, the monitoring efficiency is low, and on the other hand, it is difficult to draw accurate and scientific conclusions based on manual experience. Since the 20th century, science and technology have developed and progressed rapidly, and the monitoring level of CNC machining has become more stable and efficient as machine tools are equipped with more and more sensors and communication equipment as well as more advanced and intelligent monitoring methods.

The current tool wear status identification method mainly evaluates the tool wear status by manually extracting features from the collected signals and processing the extracted features based on simulation models or classifiers. However, this manual method of signal feature extraction has problems such as poor real-time performance and low accuracy, and is not well suited for online monitoring of tool wear. Moreover, the feature processing and identification method based on the policy model and classifier has a low accuracy and cannot accurately predict the tool wear status.

Contents of the invention

The present invention provides a tool wear status identification method, system, electronic equipment and storage medium to solve the problems in the prior art that signal feature extraction is performed manually during the tool wear status identification process, resulting in poor real-time performance and low accuracy. , as well as the problem that the existing tool wear status identification methods cannot predict the tool wear status more accurately.

The invention provides a tool wear status identification method, which includes:

Collect tool wear vibration signals;

Input the tool wear vibration signal into the preset convolutional neural network, perform feature extraction and one-time dimensionality reduction, and obtain the vibration signal characteristics;

Based on the preset local linear embedding rules, perform secondary dimensionality reduction on the vibration signal features to obtain the observation sequence;

Input the observation sequence into a pre-trained hidden Markov model, perform tool wear status identification and tool wear status prediction, and obtain identification results and prediction results;

The recognition results and prediction results are fed back to the associated terminal device.

Optionally, input the tool wear vibration signal into a preset convolutional neural network to perform feature extraction and primary dimensionality reduction. The steps of obtaining vibration signal features include:

Input the tool wear vibration signal into the convolution layer of the convolutional neural network, perform feature extraction, and obtain target features;

The target features are input into the pooling layer of the convolutional neural network, and dimensionality reduction is performed once to obtain the vibration signal features.

Optionally, based on the preset local linear embedding rules, perform secondary dimensionality reduction on the vibration signal features, and the steps of obtaining the observation sequence include:

Based on the preset neighbor point discriminant sub-rules, determine the initial neighbor point of each sample point in the vibration signal feature, and each sample point corresponds to one or more initial neighbor points;

Sort and filter the initial neighbor points according to the Euclidean distance between the sample point and the corresponding initial neighbor point, and obtain at least one target neighbor point;

By performing linear relationship fitting on the target neighbor points, obtain the local reconstruction weight matrix of the sample points corresponding to the target neighbor points;

Based on the local reconstruction weight matrix of each sample point and its target neighbor point, the observation sequence after secondary dimensionality reduction is obtained.

Optionally, the steps of obtaining the hidden Markov model include:

Obtain an observation sequence sample set, where the observation sequence sample set includes one or more observation sequence samples;

Initialize the parameters of the preset original model and set the number of hidden states, which include: initial wear, mid-term wear and severe wear;

Input the observation sequence sample into the initialized original model, and use the preset forward-backward algorithm to obtain the first intermediate quantity and the second intermediate quantity. The first intermediate quantity is based on the hidden state of each observation sequence sample. The forward probability and backward probability are obtained, and the second intermediate quantity is obtained based on the forward probability, backward probability and state transition probability of each hidden state of the observation sequence sample;

Based on the first intermediate quantity and the second intermediate quantity, the model parameters of the original model are updated and iterated to obtain the trained hidden Markov model.

Optionally, input the observation sequence into a pre-trained hidden Markov model to identify tool wear status and predict tool wear status. The steps of obtaining identification results and prediction results include:

Input the observation sequence into the hidden Markov model, perform initial forward probability calculation and/or initial backward probability calculation, and obtain the initial forward probability and/or initial backward probability of the observation sequence;

Perform forward recursion on the initial forward probability according to time sequence to obtain the target forward probability, and/or perform backward recursion on the initial backward probability according to time sequence to obtain the target backward probability;

The recognition result is obtained based on the target forward probability and/or the target backward probability.

Optionally, input the observation sequence into a pre-trained hidden Markov model to identify tool wear status and predict tool wear status. The steps of obtaining identification results and prediction results include:

Output the observation sequence to the hidden Markov model and obtain initialized local state data;

Based on the initialized local state data, perform local state dynamic planning to obtain predicted local states at multiple times;

Based on the predicted local state, obtain the target hidden state sequence probability and the target hidden state sequence at the target time;

Based on the predicted local state and the target hidden state sequence, backtracking is performed to obtain the final hidden state sequence; the final hidden state sequence is used as the prediction result.

Optionally, the step of feeding back the recognition results and prediction results to the associated terminal device includes:

Based on the identification results and prediction results, issue an alarm or generate warning information;

Feed back the warning information to the associated terminal device.

The invention also provides a tool wear status identification system, including:

Collection module, used to collect tool wear vibration signals;

The convolution module is used to input the tool wear vibration signal into a preset convolutional neural network, perform feature extraction and one-time dimensionality reduction, and obtain vibration signal characteristics;

A dimensionality reduction module, used to perform secondary dimensionality reduction on the vibration signal features based on preset local linear embedding rules to obtain an observation sequence;

An identification and prediction module, used to input the observation sequence into a pre-trained hidden Markov model, perform tool wear status identification and tool wear status prediction, and obtain identification results and prediction results;

A communication module, configured to feed back the recognition results and prediction results to the associated terminal device.

The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the tool wear state is realized as any one of the above. recognition methods.

The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements any one of the above-mentioned tool wear state identification methods.

Beneficial effects of the present invention: The present invention provides a tool wear status identification method, system, electronic equipment and storage medium, by collecting tool wear vibration signals; inputting the tool wear vibration signals into a preset convolutional neural network, and performing Feature extraction and primary dimensionality reduction are used to obtain vibration signal features; based on the preset local linear embedding rules, the vibration signal features are subjected to secondary dimensionality reduction to obtain an observation sequence; the observation sequence is input into the pre-trained hidden Marko Hu model, carry out tool wear status identification and tool wear status prediction, obtain identification results and prediction results; feed the identification results and prediction results to the associated terminal equipment. It can automatically realize feature extraction of tool wear vibration signals with high accuracy and strong real-time performance. Moreover, this method has high accuracy in identifying tool wear status and achieves more accurate prediction of tool wear status.

Description of the drawings

In order to explain the present invention or the technical solutions in the prior art more clearly, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are of the present invention. For some embodiments of the invention, those of ordinary skill in the art can also obtain other drawings based on these drawings without exerting creative efforts.

Figure 1 is a schematic flow chart of the tool wear status identification method provided by the present invention;

Figure 2 is a schematic flowchart of obtaining vibration signal characteristics in the tool wear status identification method provided by the present invention;

Figure 3 is a schematic flowchart of obtaining an observation sequence in the tool wear status identification method provided by the present invention;

Figure 4 is a schematic diagram of the implementation of local linear embedding in the tool wear status identification method provided by the present invention;

Figure 5 is a schematic flow chart of obtaining the hidden Markov model in the tool wear state identification method provided by the present invention;

Figure 6 is a schematic flow chart of tool wear status identification in the tool wear status identification method provided by the present invention;

Figure 7 is a schematic flow chart of tool wear state prediction in the tool wear state identification method provided by the present invention;

Figure 8 is a schematic structural diagram of the tool wear status identification system provided by the present invention;

Figure 9 is a schematic structural diagram of the electronic device provided by the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention more clear, the technical solutions in the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention. , not all examples. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.

The tool wear status identification method, system, electronic device and storage medium provided by the present invention will be described below in the form of embodiments with reference to FIGS. 1 to 9 .

Please refer to Figure 1. In this embodiment, a tool wear status identification method includes:

S101: Collect tool wear vibration signals.

Specifically, the tool wear vibration signal is a vibration signal of tool wear of a CNC machine tool. The tool wear vibration signal can be collected by installing a vibration sensor on the CNC machine tool. The sampling rate can be set according to the actual situation, such as 6000Hz (Hertz), etc., which will not be described here. By collecting the tool wear vibration signals of CNC machine tools, it can facilitate subsequent identification and processing based on the collected tool wear vibration signals, determine the tool wear status, and realize real-time monitoring of the tool wear status.

S102: Input the tool wear vibration signal into the preset convolutional neural network, perform feature extraction and one-time dimensionality reduction, and obtain vibration signal features.

It should be noted that by inputting the tool wear vibration signal into the preset convolutional neural network for feature extraction, the automatic extraction of tool wear vibration signal features can be achieved better, with higher accuracy and effectively reducing the problems caused by data noise. The impact is that the feature extraction is more comprehensive and avoids the problems of low accuracy and poor real-time performance caused by manual extraction or mechanical extraction. Moreover, by using a convolutional neural network to perform a dimensionality reduction on the target features obtained through feature extraction, the dimensionality of the target features can be reduced to a certain extent, thereby reducing the complexity of data processing.

S103: Based on the preset local linear embedding rules, perform secondary dimensionality reduction on the vibration signal features to obtain an observation sequence.

Specifically, the local linear embedding rule is: perform local linear embedding based on the neighbor points of each sample point in the vibration signal characteristics, that is, any sample point can be linearly represented by its corresponding neighbor point, thereby realizing the vibration signal Secondary dimensionality reduction of features to obtain the observation sequence. By performing secondary dimensionality reduction on the vibration signal features based on preset local linear embedding rules, the local features of the tool wear state can be better preserved. It should be noted that tool wear status usually has obvious local features and manifold structures. The above-mentioned local linear embedding rules can retain these local features and manifold structures while reducing dimensionality, thereby better reflecting changes in tool wear status. Moreover, the vibration signal characteristics after secondary dimensionality reduction, that is, the observation sequence, are easier to classify and identify. By mapping high-dimensional data into a low-dimensional space and retaining the local features and manifold structure of the original data, the reduced-dimensional features are easier to classify and identify. The accuracy and reliability of identification of tool wear status are better improved. Reduce the impact of noise and outliers on tool wear status identification, and improve the robustness of tool wear status identification.

S104: Input the observation sequence into the pre-trained hidden Markov model, perform tool wear status identification and tool wear status prediction, and obtain identification results and prediction results.

Specifically, by inputting the observation sequence into a pre-trained hidden Markov model for tool wear status identification and tool wear status prediction, the tool wear status can be more accurately identified and the tool wear status can be predicted.

S105: Feed back the recognition results and prediction results to the associated terminal device. The associated terminal device may be a pre-associated terminal device, such as a computer, a mobile phone, etc. By feeding the identification results and prediction results back to the associated terminal equipment, it is possible for relevant personnel to monitor the tool wear status in real time.

Please refer to Figure 2. In some embodiments, the tool wear vibration signal is input into a preset convolutional neural network to perform feature extraction and primary dimensionality reduction. The steps of obtaining vibration signal features include:

S201: Input the tool wear vibration signal into the convolution layer of the convolutional neural network, perform feature extraction, and obtain target features. By inputting the tool wear vibration signal into the convolution layer of the convolutional neural network for feature extraction, the tool wear vibration signal feature can be automatically extracted with high accuracy, and the poor recognition effect caused by manual feature extraction can be avoided. Problems such as difference.

In some embodiments, the mathematical expression of feature extraction by the convolutional layer is:

Among them, f represents the activation function, Represents the convolution kernel,/> Represents the m-th feature map in the l-th layer, /> Indicates the preset deviation value,/> Represents the extracted target features.

S202: Input the target features into the pooling layer of the convolutional neural network, perform dimensionality reduction once, and obtain the vibration signal features. Specifically, the target features are input into the pooling layer, the maximum pooling method is used to perform pooling processing on the target features output by the convolution layer, and the maximum statistical value of each pooling unit is selected to represent the corresponding feature, thereby obtaining the vibration signal characteristics. In some embodiments, the mathematical expression of the pooling process is:

in, Represents the vibration signal characteristics after secondary dimensionality reduction, max represents obtaining the maximum statistical value,/> represents the t-th neuron in the h channel in the l-th layer, and T represents the pooling step size.

Please refer to Figure 3. In some embodiments, based on preset local linear embedding rules, the vibration signal features are subjected to secondary dimensionality reduction. The steps of obtaining the observation sequence include:

S301: Based on the preset neighbor point discriminant sub-rules, determine the initial neighbor point of each sample point in the vibration signal feature, and each sample point corresponds to one or more initial neighbor points. The neighboring point discriminant sub-rules can be set according to actual conditions, such as setting the number of neighboring points to 3, 5, etc.

S302: Sort and filter the initial neighbor points according to the Euclidean distance between the sample point and the corresponding initial neighbor point, and obtain at least one target neighbor point. That is, according to the Euclidean distance between the sample point and its corresponding initial neighbor point, the initial neighbor points of the sample point are sorted, and the preset number of neighbor points ranked first are selected as the target neighbor points.

S303: Obtain the local reconstruction weight matrix of the sample point corresponding to the target neighbor point by performing linear relationship fitting on the target neighbor point. That is, by expressing the relationship between the sample point and its target neighbor point as a linear relationship, the local reconstruction weight matrix of the sample point is obtained. The local reconstruction weight matrix can be obtained by using the locally linear embedding algorithm (LLE, Locally linear embedding) in the existing technology, which will not be described again here.

S304: Based on the local reconstruction weight matrix of each sample point and its target neighbor point, obtain the observation sequence after secondary dimensionality reduction. That is, the local reconstruction weight matrix and the target neighbor points are used to map the sample points to the preset embedding coordinates, perform secondary dimensionality reduction, and use the final data as the observation sequence.

Regarding the implementation of local linear embedding, please refer _to Figure 4. First, _select the nearest neighbor, that is, select the initial neighbor point of the sample point Sort and filter the neighboring points to obtain at least one target neighboring point. Secondly, linear reconstruction is to obtain the local reconstruction weight matrix of the sample point corresponding to the target neighbor point by fitting a linear relationship to the target neighbor point (X _k , X _j ). _Wik in Figure 4 is W _ij respectively represents the weight between the sample point _Xi and the target neighbor point (X _k , X _j ). Then, mapping to embedding coordinates, that is, based on the local reconstruction weight matrix of each sample point and its target neighbor point, the sample point _Xi is mapped to the preset embedding coordinate into _Yi , and its target neighbor point X _k , X _j is mapped to Y _k , Y _j , and the observation sequence after secondary dimensionality reduction is obtained.

Please refer to Figure 5. In some embodiments, the step of obtaining the hidden Markov model includes:

S501: Obtain an observation sequence sample set, where the observation sequence sample set includes one or more observation sequence samples. Among them, the observation sequence sample is such as O={O ₁ , O ₂ ,... _OT }, and _OT represents the characteristics after secondary dimensionality reduction. By obtaining the observation sequence sample set, subsequent model training is facilitated.

S502: Initialize the parameters of the preset original model, and set the number of hidden states. The hidden states include: initial wear, mid-term wear and severe wear.

It should be noted that the original model is the original hidden Markov model λ, and its initial parameters include: initial state probability matrix Л, state transition probability matrix A, and emission matrix B. The hidden state can be set according to actual needs, such as normal, completely damaged, etc., which will not be described again here.

S503: Input the observation sequence sample into the initialized original model, and use the preset forward-backward algorithm to obtain the first intermediate quantity and the second intermediate quantity. The first intermediate quantity is based on each hidden value of the observation sequence sample. The forward probability and backward probability of the state are obtained, and the second intermediate quantity is obtained based on the forward probability, backward probability and state transition probability of each hidden state of the observation sequence sample.

Specifically, the mathematical expression of the first intermediate quantity is:

Among them, γ _t (i) represents the first intermediate quantity, P (i _t =q _i ,O; λ) represents the probability of the observation sequence sample of the hidden state of q _i , i _t represents the hidden state, and P (O; λ) represents The output probability of the observation sequence sample, α _t (i) represents the forward probability obtained by recursion, and β _t (i) represents the backward probability obtained by recursion.

The mathematical expression of the second intermediate quantity is:

Among them, ξ _t (i,j) represents the second intermediate quantity, P(i _t =q _i ,i _t+1 =q _j ,O; λ) represents the hidden state at time t is q _i , and time t+1 The probability of the observation sequence sample whose hidden state is q _j , a _ij represents the probability of converting from hidden state q _i to hidden state q _j , b _j (O _t+1 ) represents the feature O _t+1 is the hidden state q _j The corresponding emission matrix, β _t+1 (j), represents the backward probability at time t+1 obtained by recursion.

S504: Based on the first intermediate quantity and the second intermediate quantity, update and iterate the model parameters of the original model to obtain the trained hidden Markov model.

Please refer to Figure 6. In some embodiments, the observation sequence is input into a pre-trained hidden Markov model to identify tool wear status. The steps of obtaining the identification results include:

S601: Input the observation sequence into the hidden Markov model, perform initial forward probability calculation and/or initial backward probability calculation, and obtain the initial forward probability and/or initial backward probability of the observation sequence.

Specifically, given the hidden Markov model λ = (A, B, Л), A is the state transition probability matrix, B is the emission matrix, and Л is the initial state probability matrix. Given the observation sequence O={O ₁ , O ₂ ,...O _T }.

The steps for calculating the initial forward probability include: inputting the observation sequence into the hidden Markov model, calculating the forward probability of each hidden state at the initial moment, and obtaining:

α ₁ (i)＝π _i b _i (O ₁ ),i＝1,2,…N

Among them, α ₁ (i) represents the initial forward probability of the i-th hidden state at time 1, π _i represents the initial state probability matrix corresponding to the i-th hidden state, and b _i (o ₁ ) represents the feature O ₁ as the i-th hidden state. The emission matrix of hidden states, N represents the number of hidden states.

Furthermore, the steps for calculating the initial backward probability include: inputting the observation sequence into the hidden Markov model, calculating the backward probability of each hidden state at the initial moment, and obtaining:

β _T (i)＝1

Among them, β _T (i) represents the initial backward probability of the i-th hidden state at time T.

It can be understood that the initial time for forward probability calculation is usually time 1, and the initial time for backward probability calculation is usually time T, which is the last time.

S602: Perform forward recursion on the initial forward probability according to time sequence to obtain the target forward probability, and/or perform backward recursion on the initial backward probability according to time sequence to obtain the target backward probability.

Specifically, forward recursion is performed on the initial forward probability according to time sequence, that is, the forward probabilities at recursion times 2, 3,...T are recursed to obtain the target forward probability. The mathematical expression of the step of forward recursion on the initial forward probability according to time sequence is:

Among them, α _t+1 (i) represents the probability obtained by forward recursion, α _t (j) represents the probability that the hidden state is q _j at time t, and a _ji represents the transition from hidden state q _j to hidden state q _i . Probability, that is, the probability that the transition from the j-th hidden state to the i-th hidden state occurs, b _i (O _t+1 ) represents the emission matrix where the characteristic O _t+1 is the i-th hidden state.

Furthermore, perform backward recursion on the initial backward probability according to time sequence, that is, recurse the backward probabilities at times T-1, T-2,...2 to obtain the target backward probability. The mathematical expression of the target backward probability is:

Among them, β _t (i) represents the target backward probability obtained by backward recursion, b _j (o _t+1 ) represents the feature O _t+1 is the emission matrix of the jth hidden state, β _t+1 (j) Represents the backward probability of the jth hidden state at time t+1.

S603: Obtain the recognition result based on the target forward probability and/or target backward probability.

Specifically, based on the target forward probability, the mathematical expression of obtaining the recognition result is:

N

P(O|λ)=Σα _T (i)

i=1

Among them, P(O|λ) represents the recognition result, that is, the observation sequence probability, and α _T (i) represents the target forward probability obtained by forward recursion, that is, the probability of the i-th hidden state at time T.

Furthermore, based on the target backward probability, the mathematical expression for obtaining the recognition result is:

Among them, π _i represents the initial state probability matrix corresponding to the i-th hidden state, b _i (O ₁ ) represents the emission matrix of feature O ₁ for the i-th hidden state, and β ₁ (i) represents the target obtained by backward recursion. Backward probability, that is, the probability that time 1 is the i-th hidden state.

In addition, the recognition results obtained based on the forward probability of the target and the backward probability of the target are usually the same. Therefore, when the forward probability calculation and the backward probability calculation are performed at the same time, the recognition result obtained by the forward probability calculation or the backward probability calculation is used. Any one of the recognition results can be used as the final recognition result.

Please refer to Figure 7. In some embodiments, the observation sequence is input into a pre-trained hidden Markov model to predict tool wear status. The steps of obtaining the prediction results include:

S701: Output the observation sequence to the hidden Markov model and obtain initialized local state data.

Specifically, the observation sequence O = {O ₁ , O ₂ ,...O _T } is input into the hidden Markov model λ = (A, B, Л), the local state is initialized, and the initialized local state data is obtained. The mathematics Expressed as:

δ ₁ (i)＝π _i b _i (O ₁ ),i＝1,2…N

Ψ ₁ (i)＝0,i＝1,2…N

Among them, δ ₁ (i) and Ψ ₁ (i) represent the two state parameters for initializing local state data, δ ₁ (i) focuses on the initial probability, and Ψ ₁ (i) focuses on the transition probability between hidden states.

S702: Based on the initialized local state data, perform local state dynamic planning to obtain predicted local states at multiple times.

Specifically, based on the initialized local state data, local state dynamic planning is performed to obtain the predicted local state at time t=2, 3,...T. The mathematical expression of the predicted local state is:

Among them, δ _t (i) and Ψ _t (i) represent the two state parameters for predicting the local state, δ _t-1 (j) represents the value of δ corresponding to the j-th hidden state at time t-1, b _i (O _t ) represents the feature O _t is the emission matrix of the i-th hidden state.

S703: Based on the predicted local state, obtain the target hidden state sequence probability and the target hidden state sequence at the target time.

In some embodiments, the mathematical expression of the target hidden state sequence probability is:

The mathematical expression of the target hidden state sequence is:

Among them, P ^* represents the target hidden state sequence probability, represents the target hidden state sequence, δ _T (i) represents the δ value corresponding to the i-th hidden state at time T.

S704: Based on the predicted local state and the target hidden state sequence, perform backtracking to obtain the final hidden state sequence; use the final hidden state sequence as the prediction result. That is, Ψ _t (i) is used to trace back the time t=T-1, T-2,...,1 to obtain the final hidden state sequence. The mathematical expression of the final hidden state sequence is:

in, Represents the final hidden state sequence.

In the above steps, by using the trained hidden Markov model to predict the tool wear status, it is easier for relevant personnel to better evaluate and control the future wear status and wear degree of the tool, and is highly implementable. Moreover, the accuracy of the final hidden state sequence predicted by the above method is relatively high.

In some embodiments, the step of feeding back the identification results and prediction results to the associated terminal device includes: issuing an alarm or generating warning information based on the identification results and prediction results; and feeding back the warning information to the associated terminal device. equipment.

The tool wear status identification system provided by the present invention is described below. The tool wear status identification system described below and the tool wear status identification method described above can be mutually referenced.

Please refer to Figure 8. The tool wear status identification system provided by this embodiment includes:

The collection module 801 is used to collect tool wear vibration signals.

The convolution module 802 is used to input the tool wear vibration signal into a preset convolutional neural network, perform feature extraction and one-time dimensionality reduction, and obtain vibration signal features.

The dimensionality reduction module 803 is used to perform a secondary dimensionality reduction on the vibration signal features based on preset local linear embedding rules to obtain an observation sequence.

The identification and prediction module 804 is used to input the observation sequence into a pre-trained hidden Markov model, perform tool wear status identification and tool wear status prediction, and obtain identification results and prediction results.

The communication module 805 is used to feed back the recognition results and prediction results to the associated terminal device. The acquisition module 801, convolution module 802, dimensionality reduction module 803, recognition prediction module 804 and communication module 805 are connected. The tool wear status identification system provided by this embodiment can automatically realize the feature extraction of tool wear vibration signals with high accuracy and strong real-time performance. Moreover, the tool wear status identification accuracy of this system is high, and the tool wear status identification system can realize the feature extraction of tool wear vibration signals. The prediction of wear status is more accurate, more implementable and lower cost.

In some embodiments, the convolution module 802 inputs the tool wear vibration signal into a preset convolutional neural network to perform feature extraction and primary dimensionality reduction. The steps of obtaining vibration signal features include:

The tool wear vibration signal is input into the convolution layer of the convolutional neural network, feature extraction is performed, and target features are obtained.

The target features are input into the pooling layer of the convolutional neural network, and dimensionality reduction is performed once to obtain the vibration signal features.

In some embodiments, the dimensionality reduction module 803 performs a secondary dimensionality reduction on the vibration signal features based on preset local linear embedding rules. The step of obtaining the observation sequence includes:

Based on the preset neighbor point discriminant sub-rules, the initial neighbor point of each sample point in the vibration signal feature is determined, and each sample point corresponds to one or more initial neighbor points.

According to the Euclidean distance between the sample point and the corresponding initial neighbor point, the initial neighbor points are sorted and filtered to obtain at least one target neighbor point.

By performing linear relationship fitting on the target neighbor point, a local reconstruction weight matrix of the sample point corresponding to the target neighbor point is obtained.

Based on the local reconstruction weight matrix of each sample point and its target neighbor point, the observation sequence after secondary dimensionality reduction is obtained.

In some embodiments, the step of obtaining the hidden Markov model includes:

Obtain an observation sequence sample set, where the observation sequence sample set includes one or more observation sequence samples.

Initialize the parameters of the preset original model and set the number of hidden states, which include: initial wear, mid-term wear and severe wear.

Input the observation sequence sample into the initialized original model, and use the preset forward-backward algorithm to obtain the first intermediate quantity and the second intermediate quantity. The first intermediate quantity is based on the hidden state of each observation sequence sample. The forward probability and backward probability are obtained, and the second intermediate quantity is obtained based on the forward probability, backward probability and state transition probability of each hidden state of the observation sequence sample.

Based on the first intermediate quantity and the second intermediate quantity, the model parameters of the original model are updated and iterated to obtain the trained hidden Markov model.

In some embodiments, the identification and prediction module 804 inputs the observation sequence into a pre-trained hidden Markov model to perform tool wear status identification and tool wear status prediction. The steps of obtaining identification results and prediction results include:

The observation sequence is input into the hidden Markov model, initial forward probability calculation and/or initial backward probability calculation is performed, and the initial forward probability and/or initial backward probability of the observation sequence is obtained.

Perform forward recursion on the initial forward probability according to time sequence to obtain the target forward probability, and/or perform backward recursion on the initial backward probability according to time sequence to obtain the target backward probability.

The recognition result is obtained based on the target forward probability and/or the target backward probability.

In some embodiments, the identification and prediction module 804 inputs the observation sequence into a pre-trained hidden Markov model to perform tool wear status identification and tool wear status prediction. The steps of obtaining identification results and prediction results include:

The observation sequence is output to the hidden Markov model to obtain initialized local state data.

Based on the initialized local state data, local state dynamic planning is performed to obtain predicted local states at multiple times.

Based on the predicted local state, the target hidden state sequence probability and the target hidden state sequence at the target time are obtained.

Based on the predicted local state and the target hidden state sequence, backtracking is performed to obtain the final hidden state sequence; the final hidden state sequence is used as the prediction result.

In some embodiments, the step of the communication module 805 feeding back the identification results and prediction results to the associated terminal device includes:

Based on the recognition results and prediction results, an alarm is issued or warning information is generated.

Feed back the warning information to the associated terminal device.

Figure 9 illustrates a schematic diagram of the physical structure of an electronic device. As shown in Figure 9, the electronic device may include: a processor (processor) 910, a communications interface (Communications Interface) 920, a memory (memory) 930, and a communication bus 940. Among them, the processor 910, the communication interface 920, and the memory 930 complete communication with each other through the communication bus 940. The processor 910 can call the logic instructions in the memory 930 to execute the tool wear status identification method. The method includes: collecting the tool wear vibration signal; inputting the tool wear vibration signal into a preset convolutional neural network to perform feature extraction and Dimensionality reduction is performed once to obtain vibration signal features; based on preset local linear embedding rules, dimensionality reduction is performed twice on the vibration signal features to obtain an observation sequence; the observation sequence is input into a pre-trained hidden Markov model, Carry out tool wear status identification and tool wear status prediction, obtain identification results and prediction results, and feed back the identification results and prediction results to the associated terminal equipment.

In addition, the above-mentioned logical instructions in the memory 930 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention essentially or the part that contributes to the existing technology or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code. .

On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can Execute the tool wear status identification method provided by the above methods. The method includes: collecting tool wear vibration signals; inputting the tool wear vibration signals into a preset convolutional neural network, performing feature extraction and one-time dimensionality reduction to obtain the vibration signals Features; based on the preset local linear embedding rules, perform secondary dimensionality reduction on the vibration signal features to obtain the observation sequence; input the observation sequence into the pre-trained hidden Markov model to identify tool wear status and tool Wear status prediction: obtain identification results and prediction results; feed back the identification results and prediction results to the associated terminal equipment.

In another aspect, the present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. The computer program is implemented when executed by the processor to perform the tool wear status identification method provided by the above methods. The method It includes: collecting tool wear vibration signals; inputting the tool wear vibration signals into a preset convolutional neural network, performing feature extraction and one-time dimensionality reduction to obtain vibration signal characteristics; based on the preset local linear embedding rules, the vibration The signal features are subjected to secondary dimensionality reduction to obtain the observation sequence; the observation sequence is input into the pre-trained hidden Markov model, tool wear status identification and tool wear status prediction are performed, and the identification results and prediction results are obtained; the identification results are obtained The results and prediction results are fed back to the associated terminal equipment.

The device embodiments described above are only illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in One location, or it can be distributed across multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Persons of ordinary skill in the art can understand and implement the method without any creative effort.

Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the part of the above technical solution that essentially contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., including a number of instructions to cause a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that it can still be used Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent substitutions are made to some of the technical features; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. A method of identifying a tool wear state, comprising:

collecting cutter abrasion vibration signals;

inputting the cutter abrasion vibration signal into a preset convolutional neural network, performing feature extraction and one-time dimension reduction, and obtaining vibration signal features;

based on a preset local linear embedding rule, performing secondary dimension reduction on the vibration signal characteristics to obtain an observation sequence;

inputting the observation sequence into a pre-trained hidden Markov model, and carrying out cutter abrasion state identification and cutter abrasion state prediction to obtain an identification result and a prediction result;

and feeding the identification result and the prediction result back to the associated terminal equipment.

2. The method for recognizing the wear state of a tool according to claim 1, wherein the step of inputting the tool wear vibration signal into a predetermined convolutional neural network, performing feature extraction and one-time dimension reduction, and acquiring the vibration signal feature comprises:

inputting the cutter abrasion vibration signal into a convolution layer of the convolution neural network, and extracting features to obtain target features;

and inputting the target characteristics into a pooling layer of the convolutional neural network, and performing primary dimension reduction to acquire the vibration signal characteristics.

3. The method for recognizing a tool wear state according to claim 1, wherein the step of secondarily reducing the dimension of the vibration signal characteristic based on a preset local linear embedding rule, and acquiring the observation sequence comprises:

determining initial adjacent points of each sample point in the vibration signal characteristics based on a preset adjacent point judging sub-rule, wherein each sample point corresponds to one or more initial adjacent points;

sorting and screening the initial adjacent points according to Euclidean distance between the sample point and the corresponding initial adjacent point to obtain at least one target adjacent point;

obtaining a local reconstruction weight matrix of a sample point corresponding to the target adjacent point by performing linear relation fitting on the target adjacent point;

and acquiring the observation sequence after secondary dimension reduction based on the local reconstruction weight matrix of each sample point and the target adjacent point thereof.

4. The tool wear state identification method according to claim 1, wherein the step of acquiring the hidden markov model includes:

obtaining an observation sequence sample set, the observation sequence sample set comprising one or more observation sequence samples;

initializing parameters of a preset original model, and setting the number of hidden states, wherein the hidden states comprise: initial wear, mid-wear, and severe wear;

Inputting the observation sequence sample into an initialized original model, and acquiring a first intermediate quantity and a second intermediate quantity by utilizing a preset forward and backward algorithm, wherein the first intermediate quantity is obtained based on the forward probability and backward probability of each hidden state of the observation sequence sample, and the second intermediate quantity is obtained based on the forward probability, backward probability and state transition probability of each hidden state of the observation sequence sample;

and updating and iterating model parameters of the original model based on the first intermediate quantity and the second intermediate quantity to obtain the trained hidden Markov model.

5. The method for recognizing the tool wear state according to claim 1, wherein the step of inputting the observation sequence into a pre-trained hidden markov model to recognize and predict the tool wear state and obtain the recognition result and the prediction result includes:

inputting the observation sequence into the hidden Markov model, and performing initial forward probability calculation and/or initial backward probability calculation to obtain initial forward probability and/or initial backward probability of the observation sequence;

forward recursion is carried out on the initial forward probability according to the time sequence to obtain a target forward probability, and/or backward recursion is carried out on the initial backward probability according to the time sequence to obtain a target backward probability;

And acquiring the identification result based on the target forward probability and/or the target backward probability.

6. The method for recognizing the tool wear state according to claim 1, wherein the step of inputting the observation sequence into a pre-trained hidden markov model to recognize and predict the tool wear state and obtain the recognition result and the prediction result includes:

outputting the observation sequence to the hidden Markov model to acquire initialized local state data;

based on the initialized local state data, carrying out local state dynamic planning to obtain predicted local states at a plurality of moments;

based on the predicted local state, acquiring a target hidden state sequence probability and a target hidden state sequence of a target moment;

backtracking is carried out based on the predicted local state and the target hidden state sequence, and a final hidden state sequence is obtained; and taking the final hidden state sequence as the prediction result.

7. The tool wear state identification method according to claim 1, wherein the step of feeding back the identification result and the prediction result to the associated terminal device includes:

based on the identification result and the prediction result, an alarm is sent out or warning information is generated;

And feeding the warning information back to the associated terminal equipment.

8. A tool wear state identification system, comprising:

the acquisition module is used for acquiring cutter abrasion vibration signals;

the convolution module is used for inputting the cutter abrasion vibration signal into a preset convolution neural network, extracting features and reducing dimensions once to obtain vibration signal features;

the dimension reduction module is used for carrying out secondary dimension reduction on the vibration signal characteristics based on a preset local linear embedding rule to obtain an observation sequence;

the identification prediction module is used for inputting the observation sequence into a pre-trained hidden Markov model, identifying the abrasion state of the cutter and predicting the abrasion state of the cutter, and acquiring an identification result and a prediction result;

and the communication module is used for feeding the identification result and the prediction result back to the associated terminal equipment.

9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the tool wear state identification method according to any one of claims 1 to 7 when executing the program.

10. A non-transitory computer readable storage medium having stored thereon a computer program, characterized in that the computer program, when executed by a processor, implements the tool wear state identification method according to any one of claims 1 to 7.