Method and system for monitoring the operation of a flexible circuit board production device based on artificial intelligence

Abstract

The present invention discloses a method and system for monitoring the operation of a flexible circuit board production device based on artificial intelligence, which relates to the field of flexible circuit board production. The method includes: obtaining data such as material parameters, control motor parameters, high-temperature lamination pressure and temperature, etc., processing and feature extraction, calculating the multi-feature coupling correlation coefficient, establishing an equipment monitoring prediction model based on the random forest of dynamic parameters and training it, outputting an equipment monitoring prediction result based on the equipment monitoring prediction model, and obtaining the control parameters to be adjusted. It realizes the monitoring of the operation of the flexible circuit board production device, determines the selection of the data set in the training process of the random forest algorithm based on the multi-feature coupling correlation coefficient, and improves the voting method of the decision tree in the random forest algorithm, realizing more objective and accurate monitoring and prediction of the flexible circuit board production device, and improving the accuracy and operation efficiency of the intelligent algorithm.

Description

Technical Field

[0001] The present invention relates to the field of flexible circuit board production, and in particular to a method and system for monitoring the operation of flexible circuit board production equipment based on artificial intelligence. Background Art

[0002] A flexible printed circuit board (Flexible Printed Circuit Board), commonly known as FPC in the industry, is a printed circuit board made of a flexible insulating substrate and has many advantages that rigid printed circuit boards do not have. For example, it can be freely bent, wound, and folded. Using FPC can greatly reduce the volume of electronic products and meet the needs of the development of electronic products towards high density, miniaturization, and high reliability. Therefore, FPC has been widely used in fields or products such as aerospace, military, mobile communications, laptop computers, computer peripherals, PDAs, digital cameras, etc.

[0003] In the prior art, the equipment monitoring during the production process of flexible circuit boards is usually carried out through manual or machine vision inspection of the finished products after production, and the problems such as the operation of the equipment in the previous production steps are deduced by inspecting the finished products. The above method has a certain lag, and often when the problems are detected, a part of the products with quality problems have already been produced, and it is impossible to achieve early prediction. During the production process of flexible circuit boards, problems such as unstable equipment performance, inaccurate positioning, dirty surface of the coil, and deformation of the coil often occur in the common conveying process and high-temperature pressing process, and it is necessary to prevent the occurrence of production quality problems in advance. Summary of the Invention

[0004] In order to solve the technical problems of monitoring the operation of flexible circuit board production equipment in the prior art, the present invention provides a method and system for monitoring the operation of flexible circuit board production equipment based on artificial intelligence.

[0005] The present invention is realized by the following technical solutions:

[0006] A method for monitoring the operation of flexible circuit board production equipment based on artificial intelligence, characterized by including:

[0007] S1: Data acquisition, including obtaining material parameters, transmission roller control motor parameters, high-temperature pressing pressure and temperature, vibration sensor data, tension sensor data, and laser displacement sensor data;

[0008] S2: Data processing and feature extraction; including processing and feature extraction of transmission roller control motor parameters, vibration sensors, tension sensors, laser displacement sensors, pressing temperature, and pressure data;

[0009] S3: Calculate the multi-feature coupling correlation coefficients, including the autocorrelation coefficient and the cross-correlation coefficient; the autocorrelation coefficient includes the autocorrelation coefficients of different time series of the vibration sensor and the tension sensor, which are the first autocorrelation coefficient and the second autocorrelation coefficient respectively; the cross-correlation coefficient includes the first cross-correlation coefficient between the vibration sensor and the tension sensor, the second cross-correlation coefficient between the vibration sensor and the lamination temperature, the third cross-correlation coefficient between the vibration sensor and the lamination pressure, the fourth cross-correlation coefficient between the tension sensor and the lamination temperature, and the fifth cross-correlation coefficient between the tension sensor and the lamination pressure;

[0010] S4: Establish and train an equipment monitoring and prediction model; the input of the equipment monitoring and prediction model is the data features obtained in step S2, and the output is the equipment monitoring and prediction results, including the motor operating state, the position offset, the coil state, and the high-temperature lamination state; the equipment monitoring and prediction model is a random forest model based on dynamic parameters, and the composition of the sample data set in the model is dynamically determined by the multi-feature coupling correlation coefficients obtained in step S3;

[0011] S5: Based on the equipment monitoring and prediction model, output the equipment monitoring and prediction results to obtain the control parameters to be adjusted.

[0012] Further, step S2 includes processing and feature extraction of the transmission roller control motor parameter data, including performing current spectrum analysis on the motor current and extracting spectrum features, and extracting spectrum features based on the spectrum signal, including the average frequency, the center frequency, the root mean square frequency, and the frequency standard deviation; denoising and normalizing the speed and torque data of the motor.

[0013] Further, perform denoising processing on the vibration sensor based on the independent component analysis method.

[0014] Further, the calculation method of the autocorrelation coefficient in step S3 is as follows:

[0015] Respectively obtain the data from the i-th to the i+m-th in the time series of the parameter to be calculated as the first set and the data from the i+m+1-th to the i+2m+1-th as the second set, then the correlation coefficient between the two sets is the autocorrelation coefficient, and the calculation formula is as follows:

[0016]

[0017] where r i is the autocorrelation coefficient at time i, x p is the value of the parameter to be calculated at time p, x p+m+1 is the value of the parameter to be calculated at time p+m+1, is the mean value of the first set, is the mean value of the second set, and m is the set value of the sequence length;

[0018] The method for calculating the cross - correlation coefficient is as follows:

[0019] Obtain the data from the \(i\) - th to the \((i + m)\) - th in the time series of the two parameters to be calculated respectively. The calculation formula is as follows:

[0020]

[0021] where \(h\) i is the cross - correlation coefficient at time \(i\), \(x\) q is the value of the first parameter to be calculated at time \(q\), \(y\) q is the value of the second parameter to be calculated at time \(q\), is the mean value of the data of the first parameter to be calculated from the \(i\) - th to the \((i + m)\) - th, is the mean value of the data of the second parameter to be calculated from the \(i\) - th to the \((i + m)\) - th.

[0022] Furthermore, in the step S4, the operating states of the motor include normal, slightly abnormal, moderately abnormal, and severely abnormal;

[0023] The states of the coil include normal, slightly abnormal, moderately abnormal, and severely abnormal;

[0024] The states of the high - temperature pressing include normal temperature and pressure, abnormal temperature and normal pressure, normal temperature and abnormal pressure, and both abnormal temperature and pressure.

[0025] Furthermore, the method for determining the composition of the sample data set in the model is as follows:

[0026] Let the first autocorrelation coefficient at time \(i\) be \(r\) 11 i , the second autocorrelation coefficient be \(r\) 22 i , the first cross - correlation coefficient be \(r\) 12 i , the second cross - correlation coefficient be \(r\) 13 i , the third cross - correlation coefficient be \(r\) 14 i , the fourth cross - correlation coefficient be \(r\) 23 i , the fifth cross - correlation coefficient be \(r\) 24 i , then the determination of the model training sample data set is carried out according to the following method:

[0027] S41: Based on the clustering algorithm, cluster the sample data set according to the clustering properties of the vibration sensor, tension sensor, pressing temperature, and pressing pressure data respectively, and obtain four clustering results, namely the clustering result of the vibration sensor data, the clustering result of the tension sensor, the clustering result of the pressing temperature, and the clustering result of the pressing pressure. Each clustering result includes \(k1\), \(k2\), \(k3\), and \(k4\) groups respectively;

[0028] S42: Randomly sample k1 groups respectively in the clustering results of the vibration sensor data, and the sampling ratios are shown as follows: 11 i where the first autocorrelation coefficient is r 12 i and the first cross-correlation coefficient is r 13 i and the second cross-correlation coefficient is r 14 i and the third cross-correlation coefficient is r

[0029]

[0030] where v1, v2, v3, and v4 are weight adjustment coefficients respectively;

[0031] S43: Randomly sample k2 groups respectively in the clustering results of the tension sensor data, and the sampling ratios are shown as follows: 22 i where the second autocorrelation coefficient is r 23 i and the fourth cross-correlation coefficient is r 24 i and the fifth cross-correlation coefficient is r

[0032]

[0033] where v5, v6, and v7 are weight adjustment coefficients respectively;

[0034] S44: Randomly sample k3 groups respectively in the clustering results of the lamination temperature, and the sampling ratios are shown as follows: 13 i where the second cross-correlation coefficient is r 23 i and the fourth cross-correlation coefficient is r

[0035]

[0036] where v8 and v9 are weight adjustment coefficients respectively;

[0037] S45: Randomly sample k4 groups respectively in the clustering results of the lamination temperature, and the sampling ratios are shown as follows: 14 i where the third cross-correlation coefficient is r 24 i and the fifth cross-correlation coefficient is r

[0038]

[0039] where v10 , v 11 are respectively weight adjustment coefficients;

[0040] S46: Combine the sample data sets sampled in steps S42 - S45, remove duplicate samples, and obtain a training sample data set.

[0041] Furthermore, the voting method of the random forest model is as follows:

[0042] According to the imbalance coefficient B of the sample data set corresponding to the i-th decision tree i , if B i > B', then remove the result of this decision tree; if B'' < B i < B', then determine the weight coefficient occupied by the decision tree when voting according to the following weights:

[0043]

[0044] where B i is the imbalance coefficient of the sample data set corresponding to the i-th decision tree, n i is the number of minority samples correctly predicted by this decision tree, Long is the length of the data samples for constructing this decision tree, acc i is the decision accuracy of the i-th decision tree, and B', B'' are set imbalance coefficient thresholds.

[0045] Furthermore, the control parameters in step S5 include control parameters of the transmission roller control motor, high-temperature pressing pressure and temperature, and position offset correction.

[0046] The present invention also provides an operation monitoring system for a flexible circuit board production device based on artificial intelligence, based on the above-mentioned operation monitoring method for a flexible circuit board production device based on artificial intelligence, which includes:

[0047] A data acquisition module, which is used to obtain material parameters, transmission roller control motor parameters, high-temperature pressing pressure and temperature, vibration sensor data, tension sensor data, and laser displacement sensor data; the vibration sensor is arranged on the roller shaft; the tension sensor is used to detect the tension received during the transmission of the coil; the laser displacement sensor is used to measure the offset of the coil;

[0048] A data processing and feature extraction module, which is used to process and extract features from the transmission roller control motor parameters, vibration sensor, tension sensor, laser displacement sensor, pressing temperature, and pressure data;

[0049] A multi - feature coupling correlation coefficient calculation module, which is used to calculate the multi - feature coupling correlation coefficient according to the vibration sensor, the tension sensor, and the temperature and pressure of the lamination, including the autocorrelation coefficient and the cross - correlation coefficient; the autocorrelation coefficient includes the autocorrelation coefficients of different time series of the vibration sensor and the tension sensor, which are the first autocorrelation coefficient and the second autocorrelation coefficient respectively; the cross - correlation coefficient includes the first cross - correlation coefficient between the vibration sensor and the tension sensor, the second cross - correlation coefficient between the vibration sensor and the lamination temperature, the third cross - correlation coefficient between the vibration sensor and the lamination pressure, the fourth cross - correlation coefficient between the tension sensor and the lamination temperature, and the fifth cross - correlation coefficient between the tension sensor and the lamination pressure;

[0050] A model establishment and training module, which is used to establish and train an equipment monitoring and prediction model; the equipment monitoring and prediction model is an improved random forest model;

[0051] The input of the equipment monitoring and prediction model is the data features collected, and the output is the equipment monitoring and prediction result, including the motor operation status, the position offset, the coil status, and the high - temperature lamination status;

[0052] A prediction output module, which is used to output the equipment monitoring and prediction result based on the equipment monitoring and prediction model to obtain the control parameters to be adjusted.

[0053] In addition, to achieve the above object, the present invention also provides a computer - readable storage medium, on which program instructions of the operation monitoring method for a flexible circuit board production equipment based on artificial intelligence are stored. The program instructions of the operation monitoring method for a flexible circuit board production equipment based on artificial intelligence can be executed by one or more processors to implement the steps of the operation monitoring method for a flexible circuit board production equipment based on artificial intelligence as described above.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] The present invention realizes the operation monitoring of the flexible circuit board production equipment, determines the selection of the data set in the training process of the random forest algorithm based on the multi - feature coupling correlation coefficient, and improves the voting method of the decision tree in the random forest algorithm, realizing a more objective and accurate monitoring and prediction of the flexible circuit board production equipment, and improving the accuracy and operation efficiency of the intelligent algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0057] Figure 1Schematic flow chart of an operation monitoring method for a flexible circuit board production device based on artificial intelligence according to an embodiment of the present application;

[0058] Figure 2 Schematic diagram of an equipment monitoring prediction model according to an embodiment of the present application. Detailed implementation manners

[0059] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0060] The following specific examples illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0061] It should also be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. The diagrams only show the components related to the present invention, rather than being drawn according to the number, shape, and size of the components in actual implementation. The actual form, quantity, and ratio of each component during actual implementation can be arbitrarily changed, and the component layout form may also be more complex.

[0062] See Figure 1 , an operation monitoring method for a flexible circuit board production device based on artificial intelligence, including the following steps:

[0063] S1: Data collection, including obtaining material parameters, transmission roller control motor parameters, high-temperature lamination pressure and temperature, vibration sensor data, tension sensor data, and laser displacement sensor data;

[0064] The material parameters include substrate, cover film material type and thickness, copper foil thickness, adhesive material and dosage;

[0065] The substrate material types include polyimide, polyester, and polyethersiloxane;

[0066] Different material parameters result in significant differences in flexibility, strength, heat resistance, etc. of the flexible circuit board in each step of the production process. Therefore, appropriate processing parameters should be selected based on the material characteristics of the substrate during the production process;

[0067] The parameters of the transmission roller control motor include current, speed, and torque;

[0068] Among them, the vibration sensor is arranged on the roller shaft;

[0069] The tension sensor is used to detect the tension suffered during the coil transmission process;

[0070] The vibration sensor and the tension sensor are used to monitor the states of the roller and the transmitted material during the transmission process. For example, dirt on the coil surface and coil deformation will be reflected in the vibration sensor of the roller and the change of tension;

[0071] The laser displacement sensor is used to measure the offset of the coil.

[0072] S2: Data processing and feature extraction; among them, it includes,

[0073] S21: Data processing and feature extraction of the transmission roller control motor parameters.

[0074] Specifically, it includes performing current spectrum analysis on the motor current and extracting spectrum features. The steps are as follows:

[0075] According to the speed of the motor, the motor current data within a set time is collected as a sample under different load conditions. For any sample, the time-domain current signal is converted into a frequency-domain signal based on FFT. The FFT expression is as follows:

[0076]

[0077] Among them, f(i) is the spectrum amplitude corresponding to x(k), x(k) = {x(0), x(1), …, x(K - 1)}, and K is the number of sequences.

[0078] Spectrum features are extracted based on the spectrum signal, including average frequency, center frequency, root mean square frequency, and frequency standard deviation.

[0079] Denoising and normalization processing are performed on the speed and torque data of the motor;

[0080] S22: Vibration sensor data processing and feature extraction.

[0081] In actual engineering, various noises are mixed in the structural vibration signal, and the useful signal may be masked by the noise. Moreover, due to relying on the detailed information of the signal or the acquisition system, traditional denoising processing methods are difficult to apply. Therefore, based on the independent component analysis method, denoising processing is performed on the vibration sensor. The steps are as follows:

[0082] a. Centralization and whitening preprocessing: Process the sensor data to make its mean value 0 and perform whitening processing, that is, eliminate the correlation between data, and obtain the whitening vector matrix Z;

[0083] b. Solve the unmixing matrix: By optimizing the objective function, the unmixing matrix W is solved from the whitened samples. Among them, based on negentropy as the criterion for maximizing the non-Gaussianity of the output signal, an initial weight vector with a random modulus of 1 is selected for iteration:

[0084]

[0085] Among them, G is a non-quadratic function, and W T is the transpose of the unmixing matrix W, G ’ is G the derivative of.

[0086] c. Normalization processing of W:

[0087]

[0088] Judge whether W converges. If it does not converge, return to step b to continue iteration. If it converges, enter step d;

[0089] d. Calculate the denoised signal: Use the unmixing matrix W to perform a linear transformation on the original data to obtain the output vector U, that is, the denoised signal:

[0090] U = WX

[0091] Among them, U is the denoised signal, and X is the vibration sensor data.

[0092] Furthermore, time-frequency domain feature extraction is performed based on the processed vibration sensor data;

[0093] Among them, time domain features include maximum value, minimum value, peak value, peak-to-peak value, average value, variance, root mean square value, waveform factor, pulse factor, peak factor, margin factor, kurtosis factor;

[0094] Frequency domain features include energy, center frequency, mean square frequency, root mean square frequency, frequency variance, frequency standard deviation.

[0095] S23: Denoise and normalize the data of the tension sensor, laser displacement sensor, pressing temperature, and pressure data.

[0096] S3: Calculate the multi-feature coupling correlation coefficient

[0097] Since the vibration sensor, the tension sensor, and the temperature and pressure of lamination each reflect the state of the coil material from their respective aspects, the four parameters are coupled with each other. How to determine their respective correlations affects the accuracy of subsequent prediction results. Based on this, the present invention calculates the multi-feature coupling correlation coefficients, including the autocorrelation coefficient and the cross-correlation coefficient; the autocorrelation coefficient includes the autocorrelation coefficients of different time series of the vibration sensor and the tension sensor, which are the first autocorrelation coefficient and the second autocorrelation coefficient respectively; the cross-correlation coefficient includes the first cross-correlation coefficient between the vibration sensor and the tension sensor, the second cross-correlation coefficient between the vibration sensor and the lamination temperature, the third cross-correlation coefficient between the vibration sensor and the lamination pressure, the fourth cross-correlation coefficient between the tension sensor and the lamination temperature, and the fifth cross-correlation coefficient between the tension sensor and the lamination pressure.

[0098] The calculation of the autocorrelation coefficient and the cross-correlation coefficient is as follows:

[0099] S31: Calculation of the autocorrelation coefficient:

[0100] Respectively obtain the data from the i-th to the (i + m)-th in the time series of the parameter to be calculated as the first set and the data from the (i + m + 1)-th to the (i + 2m + 1)-th as the second set. Then the correlation coefficient between the two sets is the autocorrelation coefficient, and the calculation formula is as follows:

[0101]

[0102] where r i is the autocorrelation coefficient at the i-th moment, x p is the value of the parameter to be calculated at the p-th moment, x p+m+1 is the value of the parameter to be calculated at the (p + m + 1)-th moment, is the mean value of the first set, is the mean value of the second set, and m is the set value of the sequence length.

[0103] Taking the calculation of the first autocorrelation coefficient as an example:

[0104] Respectively obtain the data from the i-th to the (i + m)-th in the time series of the vibration sensor to form the first set S1, and the data from the (i + m + 1)-th to the (i + 2m + 1)-th to form the second set S2. Then the calculation formula of the first autocorrelation coefficient is as follows:

[0105]

[0106] where r i ’ is the first autocorrelation coefficient at the i-th moment, x p ’ is the output value of the vibration sensor at the p-th moment, x p+m+1 ’ is the output value of the vibration sensor at the (p + m + 1)-th moment, is the mean value of the first set S1, Is the average value of the second set S2.

[0107] The second autocorrelation coefficient of the tension sensor is calculated by the same method above.

[0108] S32: Cross-correlation coefficient calculation

[0109] The data from the i-th to the (i + m)-th in the time series of the two parameters to be calculated are respectively obtained, and the calculation formula is as follows:

[0110]

[0111] Among them, h i Is the cross-correlation coefficient at time i, x q Is the value of the first parameter to be calculated at time q, y q Is the value of the second parameter to be calculated at time q, Is the average value of the data of the first parameter to be calculated from the i-th to the (i + m)-th, Is the average value of the data of the second parameter to be calculated from the i-th to the (i + m)-th.

[0112] Taking the calculation of the first cross-correlation coefficient as an example:

[0113] The data of the vibration sensor and the tension sensor at the i-th to the (i + m)-th moments are respectively obtained and calculated according to the following formula:

[0114]

[0115] Among them, h i ’ Is the first cross-correlation coefficient at time i, x q ’ Is the output value of the vibration sensor at time q, y q ’ Is the output value of the tension sensor at time q, Is the average value of the data of the vibration sensor from the i-th to the (i + m)-th moments, Is the average value of the data of the tension sensor from the i-th to the (i + m)-th moments.

[0116] The second cross-correlation coefficient, the third cross-correlation coefficient, the fourth cross-correlation coefficient, and the fifth cross-correlation coefficient are calculated by the same method above.

[0117] S4: Establish and train an equipment monitoring and prediction model; the input of the equipment monitoring and prediction model is the data features obtained in step S2, and the output is the equipment monitoring and prediction result, including the motor operation state, position offset, coil state, and hot pressing state;

[0118] Furthermore, the motor operation state includes normal, slightly abnormal, moderately abnormal, and severely abnormal;

[0119] The coil state includes normal, slightly abnormal, moderately abnormal, and severely abnormal;

[0120] The high-temperature pressing state includes normal temperature and pressure, abnormal temperature and normal pressure, normal temperature and abnormal pressure, and both abnormal temperature and pressure.

[0121] As Figure 2 shown, the device monitoring and prediction model is a random forest model based on dynamic parameters. The composition of the sample data set in the model is dynamically determined by the multi-feature coupling correlation coefficient at the i-th moment obtained in step S3. The specific determination method is as follows:

[0122] Let the first autocorrelation coefficient at the i-th moment be r 11 i , the second autocorrelation coefficient be r 22 i , the first cross-correlation coefficient be r 12 i , the second cross-correlation coefficient be r 13 i , the third cross-correlation coefficient be r 14 i , the fourth cross-correlation coefficient be r 23 i , the fifth cross-correlation coefficient be r 24 i , then the determination of the model training sample data set is carried out according to the following method:

[0123] S41: Based on the clustering algorithm, the sample data set is clustered according to the clustering properties of the vibration sensor, tension sensor, pressing temperature, and pressing pressure data respectively, and four clustering results are obtained, namely the vibration sensor data clustering result, the tension sensor clustering result, the pressing temperature clustering result, and the pressing pressure clustering result. Each clustering result includes k1, k2, k3, and k4 groups respectively;

[0124] S42: According to the first autocorrelation coefficient being r 11 i , the first cross-correlation coefficient being r 12 i , the second cross-correlation coefficient being r 13 i , the third cross-correlation coefficient being r 14 i random sampling is carried out on the k1 groups in the vibration sensor data clustering result respectively, and the sampling ratios are shown as follows:

[0125]

[0126] Among them, v1, v2, v3, and v4 are weight adjustment coefficients respectively.

[0127] S43: According to the second autocorrelation coefficient being r 22 i, the fourth cross - correlation coefficient is r 23 i , the fifth cross - correlation coefficient is r 24 i Random sampling is performed on k2 groups respectively in the clustering results of the tension sensor data, and the sampling ratios are shown as follows:

[0128]

[0129] Among them, v5, v6, and v7 are weight adjustment coefficients respectively.

[0130] S44: According to the second cross - correlation coefficient is r 13 i , and the fourth cross - correlation coefficient is r 23 i , random sampling is performed on k3 groups respectively in the clustering results of the lamination temperature, and the sampling ratios are shown as follows:

[0131]

[0132] Among them, v8 and v9 are weight adjustment coefficients respectively;

[0133] S45: According to the third cross - correlation coefficient is r 14 i , and the fifth cross - correlation coefficient is r 24 i , random sampling is performed on k4 groups respectively in the clustering results of the lamination temperature, and the sampling ratios are shown as follows:

[0134]

[0135] Among them, v 10 and v 11 are weight adjustment coefficients respectively.

[0136] Optionally, the weight adjustment coefficient is determined according to experience and can be adjusted.

[0137] S46: Combine the sample data sets sampled in steps S42 - S45, and eliminate duplicate samples to obtain the training sample data set.

[0138] The present invention dynamically selects the training sample data set of the device monitoring and prediction model based on the dynamic multi - feature coupling correlation coefficient, realizes the reduction of training sample data redundancy and the real - time nature of data following, and effectively improves the accuracy and timeliness of the device monitoring and prediction model.

[0139] In addition, in the prior art, the method of using the average value is adopted to vote on the decision tree to generate the output result, and it is impossible to respond to the quality of the classification results of each decision tree in the random forest with different classification capabilities. Especially in the case of an imbalanced data set, the classification result often tends to the majority class. Therefore, the voting method of the random forest is improved based on the voting method with dynamic weights in this application.

[0140] The steps of the voting method are as follows:

[0141] According to the imbalance coefficient B of the sample data set corresponding to the i-th decision tree i , if B i > B', then the result of this decision tree is eliminated; if B'' < B i < B', then the weight coefficient when the decision tree votes is determined according to the following weight:

[0142]

[0143] where B i is the imbalance coefficient of the sample data set corresponding to the i-th decision tree, n i is the number of minority samples correctly predicted by this decision tree, Long is the length of the data sample for constructing this decision tree, acc i is the decision accuracy of the i-th decision tree, B' and B'' are the set imbalance coefficient thresholds, and the realization of the weight coefficient improves the result on the minority class data.

[0144] S5: Based on the device monitoring prediction model, the device monitoring prediction result is output, and the control parameters to be adjusted are obtained. The control parameters include the control parameters of the transmission roller control motor, the high-temperature pressing pressure and temperature, and the position offset correction. Specifically, the data features obtained in step S2 are input into the trained model, the device monitoring prediction result is output, and the control parameters to be adjusted are obtained.

[0145] In this embodiment, the operation monitoring of the flexible circuit board production equipment is realized. Based on the multi-feature coupling correlation coefficient, the selection of the data set in the training process of the random forest algorithm is determined, and the voting method of the decision tree in the random forest algorithm is improved, realizing more objective and accurate monitoring and prediction of the flexible circuit board production equipment, and improving the accuracy and operation efficiency of the intelligent algorithm.

[0146] The embodiment of the present invention also proposes an operation monitoring system for a flexible circuit board production equipment based on artificial intelligence. Based on the operation monitoring method for a flexible circuit board production equipment based on artificial intelligence as described above, it includes:

[0147] A data acquisition module, which is used to obtain material parameters, transmission roller control motor parameters, high-temperature pressing pressure and temperature, vibration sensor data, tension sensor data, and laser displacement sensor data; the vibration sensor is arranged on the roller shaft; the tension sensor is used to detect the tension suffered during the coil transmission process; the laser displacement sensor is used to measure the offset of the coil.

[0148] A data processing and feature extraction module, which is used to process and extract features from the transmission roller control motor parameters, vibration sensor, tension sensor, laser displacement sensor, pressing temperature, and pressure data.

[0149] A multi-feature coupling correlation coefficient calculation module, which is used to calculate the multi-feature coupling correlation coefficient according to the vibration sensor, tension sensor, and the temperature and pressure of pressing, including the autocorrelation coefficient and the cross-correlation coefficient; the autocorrelation coefficient includes the autocorrelation coefficients of different time series of the vibration sensor and the tension sensor, which are the first autocorrelation coefficient and the second autocorrelation coefficient respectively; the cross-correlation coefficient includes the first cross-correlation coefficient between the vibration sensor and the tension sensor, the second cross-correlation coefficient between the vibration sensor and the pressing temperature, the third cross-correlation coefficient between the vibration sensor and the pressing pressure, the fourth cross-correlation coefficient between the tension sensor and the pressing temperature, and the fifth cross-correlation coefficient between the tension sensor and the pressing pressure.

[0150] A model establishment and training module, which is used to establish and train an equipment monitoring and prediction model; the equipment monitoring and prediction model is an improved random forest model.

[0151] The input of the equipment monitoring and prediction model is the data features obtained by collection, and the output is the equipment monitoring and prediction result, including the motor running state, position offset, coil state, and high-temperature pressing state.

[0152] Further, the motor running state includes normal, slightly abnormal, moderately abnormal, and severely abnormal.

[0153] The coil state includes normal, slightly abnormal, moderately abnormal, and severely abnormal.

[0154] The high-temperature pressing state includes normal temperature and pressure, abnormal temperature and normal pressure, normal temperature and abnormal pressure, and both abnormal temperature and pressure.

[0155] The composition of the sample data set in the model is dynamically determined by the multi-feature coupling correlation coefficient at the i-th moment; the model votes based on dynamic weights.

[0156] A prediction output module, which is used to output the equipment monitoring and prediction result based on the equipment monitoring and prediction model to obtain the control parameters to be adjusted.

[0157] In addition, an embodiment of the present invention further provides a computer-readable storage medium, on which program instructions for the operation monitoring method of a flexible circuit board production device based on artificial intelligence are stored. The program instructions for the operation monitoring method of the flexible circuit board production device based on artificial intelligence can be executed by one or more processors to implement the steps of the operation monitoring method of the flexible circuit board production device based on artificial intelligence as described above.

[0158] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for monitoring the operation of a flexible circuit board production device based on artificial intelligence, characterized in that, include: S1: Data acquisition, including obtaining material parameters, transmission roller control motor parameters, high temperature pressing pressure and temperature, vibration sensor data, tension sensor data, laser displacement sensor data; S2: Data processing and feature extraction; including processing and feature extraction of transmission roller control motor parameters, vibration sensor, tension sensor, laser displacement sensor, lamination temperature, and pressure data; S3: Calculate multi-feature coupling correlation coefficients, including autocorrelation coefficients and mutual correlation coefficients; the autocorrelation coefficients include autocorrelation coefficients of different time series of the vibration sensor and the tension sensor, which are respectively the first autocorrelation coefficient and the second autocorrelation coefficient; the mutual correlation coefficients include the first mutual correlation coefficient between the vibration sensor and the tension sensor, the second mutual correlation coefficient between the vibration sensor and the pressing temperature, the third mutual correlation coefficient between the vibration sensor and the pressing pressure, the fourth mutual correlation coefficient between the tension sensor and the pressing temperature, and the fifth mutual correlation coefficient between the tension sensor and the pressing pressure; S4: Establishing and training an equipment monitoring prediction model; the equipment monitoring prediction model inputs the data features obtained in step S2, and outputs the equipment monitoring prediction results, including motor operation status, position offset, coil status, and high-temperature pressing status; the equipment monitoring prediction model is a random forest model based on dynamic parameters, and the composition of the sample data set in the model is dynamically determined by the multi-feature coupling correlation coefficient obtained in step S3; The composition of the sample data set in the model is determined as follows: Let the first autocorrelation coefficient at the \(i\)-th moment be \(r\) 11 i , and the second autocorrelation coefficient be \(r\) 22 i , the first cross-correlation coefficient be \(r\) 12 i , the second cross-correlation coefficient be \(r\) 13 i , the third cross-correlation coefficient be \(r\) 14 i , the fourth cross-correlation coefficient be \(r\) 23 i , the fifth cross-correlation coefficient be \(r\) 24 i , then the determination of the model training sample data set is carried out according to the following method: S41: clustering the sample data sets according to the clustering properties of the vibration sensor, tension sensor, pressing temperature and pressing pressure data based on the clustering algorithm, and obtaining four clustering results, namely, the vibration sensor data clustering result, the tension sensor clustering result, the pressing temperature clustering result and the pressing pressure clustering result, each clustering result includes k1, k2, k3, k4 groups respectively; S42: According to the first autocorrelation coefficient being r 11 i , the first cross-correlation coefficient being r 12 i , the second cross-correlation coefficient being r 13 i , the third cross-correlation coefficient being r 14 i Random sampling is respectively performed on k1 groups in the clustering result of the vibration sensor data, and the sampling ratios are shown as follows wherein, v1, v2, v3, and v4 are respectively weight adjustment coefficients; S43: According to the second autocorrelation coefficient being r 22 i , the fourth cross-correlation coefficient being r 23 i , the fifth cross-correlation coefficient being r 24 i Random sampling is performed on k2 groups respectively in the clustering results of the tension sensor data, and the sampling ratios are shown as follows: Among them, v5, v6, and v7 are weight adjustment coefficients respectively; S44: According to the second cross-correlation coefficient being r 13 i , and the fourth cross-correlation coefficient being r 23 i , in the lamination temperature clustering results, random sampling is respectively performed on k3 groups, and the sampling ratios are shown as follows: Among them, v8 and v9 are weight adjustment coefficients respectively; S45: According to the third cross-correlation coefficient being r 14 i , and the fifth cross-correlation coefficient being r 24 i , in the lamination temperature clustering results, random sampling is performed on k4 groups respectively, and the sampling ratios are shown as follows: Among them, v 10 and v 11 are weight adjustment coefficients respectively; S46: Combined with the sample data set obtained by sampling in steps S42-S45, duplicate samples are removed to obtain a training sample data set; S5: Output the equipment monitoring prediction results based on the equipment monitoring prediction model to obtain the control parameters that need to be adjusted.

2. The operation monitoring method of the flexible circuit board production equipment based on artificial intelligence according to claim 1, wherein, The step S2 includes processing the transmission roller control motor parameter data and extracting features, including current spectrum analysis of the motor current and extracting spectrum features, extracting spectrum features based on spectrum signals, including average frequency, center of gravity frequency, root mean square frequency and frequency standard deviation; and denoising and normalizing the motor speed and torque data.

3. The operation monitoring method of the flexible circuit board production equipment based on artificial intelligence according to claim 2, wherein, The vibration sensor is denoised based on independent component analysis.

4. The method for monitoring the operation of a flexible circuit board production device based on artificial intelligence according to claim 1, wherein The method for calculating the autocorrelation coefficient in step S3 is as follows: The data from the i~i+mth in the time series of the parameter to be calculated are obtained as the first set and the data from the i+m+1th to i+2m+1th are obtained as the second set. The correlation coefficient between the two sets is the autocorrelation coefficient, and the calculation formula is as follows: Among them, r i is the autocorrelation coefficient at time i, x p is the parameter value to be calculated at time p, x p+m+1 is the parameter value to be calculated at time p + m + 1, is the mean of the first set, is the mean of the second set, and m is the set value of the sequence length; The cross-correlation coefficient is calculated as follows: Get the data from the i~i+mth time series of the two parameters to be calculated respectively. The calculation formula is as follows: Among them, h i is the cross-correlation coefficient at time i, x q is the value of the first parameter to be calculated at time q, y q is the value of the second parameter to be calculated at time q, is the mean value of the data of the first parameter to be calculated from i to i + m, is the mean value of the data of the second parameter to be calculated from i to i + m.

5. The method for monitoring the operation of a flexible circuit board production device based on artificial intelligence according to claim 1, characterized in that, In step S4, the operating states of the motor include normal, slightly abnormal, moderately abnormal, and severely abnormal; The states of the coil include normal, slightly abnormal, moderately abnormal, and severely abnormal; The high-temperature pressing states include normal temperature and pressure, abnormal temperature and normal pressure, normal temperature and abnormal pressure, and both abnormal temperature and pressure.

6. The operation monitoring method of the flexible circuit board production equipment based on artificial intelligence according to claim 5, wherein The voting method steps of the random forest model are as follows: According to the imbalance coefficient B of the sample data set corresponding to the i-th decision tree i , if B i > B', then eliminate the result of this decision tree; if B'' < B i < B', then determine the weight coefficient of the decision tree in voting according to the following weights. The formula is as follows: Among them, B i is the imbalance coefficient of the sample data set corresponding to the i-th decision tree, n i is the number of minority samples correctly predicted by the decision tree, Long is the length of the data samples for constructing the decision tree, and acc i is the decision accuracy of the i-th decision tree, and B', B'' are the set thresholds for the imbalance coefficient.

7. The method for monitoring the operation of a flexible circuit board production device based on artificial intelligence according to claim 6, characterized in that, In step S5, the control parameters include the control parameters of the conveyor roller control motor, the high-temperature pressing pressure and temperature, and the position offset correction.

8. An operation monitoring system for a flexible circuit board production device based on artificial intelligence, based on the operation monitoring method for a flexible circuit board production device based on artificial intelligence according to any one of claims 1 to 7, includes: A data acquisition module, which is used to obtain material parameters, conveyor roller control motor parameters, high-temperature pressing pressure and temperature, vibration sensor data, tension sensor data, and laser displacement sensor data; the vibration sensor is arranged on the roller shaft; the tension sensor is used to detect the tension suffered during the coil transmission process; the laser displacement sensor is used to measure the offset of the coil; A data processing and feature extraction module, which is used to process and extract features from the conveyor roller control motor parameters, vibration sensor, tension sensor, laser displacement sensor, pressing temperature, and pressure data; A multi-feature coupling correlation coefficient calculation module, which is used to calculate the multi-feature coupling correlation coefficient according to the vibration sensor, tension sensor, and the temperature and pressure of pressing, including the autocorrelation coefficient and the cross-correlation coefficient; the autocorrelation coefficient includes the autocorrelation coefficients of different time series of the vibration sensor and the tension sensor, which are the first autocorrelation coefficient and the second autocorrelation coefficient respectively; the cross-correlation coefficient includes the first cross-correlation coefficient between the vibration sensor and the tension sensor, the second cross-correlation coefficient between the vibration sensor and the pressing temperature, the third cross-correlation coefficient between the vibration sensor and the pressing pressure, the fourth cross-correlation coefficient between the tension sensor and the pressing temperature, and the fifth cross-correlation coefficient between the tension sensor and the pressing pressure; A model establishment and training module, which is used to establish and train an equipment monitoring prediction model; the equipment monitoring prediction model is an improved random forest model; The input of the equipment monitoring prediction model is the data features collected, and the output is the equipment monitoring prediction result, including the operating state of the motor, the position offset, the state of the coil, and the high-temperature pressing state; A prediction output module, which is used to output the equipment monitoring prediction result based on the equipment monitoring prediction model to obtain the control parameters that need to be adjusted.

9. A computer-readable storage medium, characterized in that, The program instructions of the operation monitoring method for a flexible circuit board production device based on artificial intelligence are stored on the computer-readable storage medium, and the program instructions of the operation monitoring method for a flexible circuit board production device based on artificial intelligence can be executed by one or more processors to implement the steps of the operation monitoring method for a flexible circuit board production device based on artificial intelligence according to any one of claims 1 to 7.

Images