An Abnormal Monitoring Method and System for an Aluminum Plate Cutting Device

Abstract

The present invention relates to the technical field of data processing, and specifically discloses a method and system for abnormal monitoring of an aluminum plate cutting device. The method includes the following steps: obtaining a training set, where the training set includes multiple cutting pressure values; training an isolation forest model using the training set to obtain a trained isolation forest model; the improvement threshold of the isolation forest model is inversely correlated with the mean of the contamination ratios of all isolation trees in the isolation forest model; the contamination ratio is inversely correlated with the degree of dispersion of each cutting pressure value in each isolation tree and the degree of correlation with the diversity of the aluminum plates; the diversity is the variance of the weighted average of the normalized values of each parameter of all aluminum plates; inputting the measured cutting pressure value into the trained isolation forest model, and when the output abnormal score is greater than the improvement threshold, it is determined that there is an abnormality, and an early warning is given and the operation of the aluminum plate cutting device is stopped. The solution of the present invention improves the accuracy of abnormal detection and reduces false alarms and missed alarms.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly relates to a method and system for abnormal monitoring of an aluminum plate cutting device. Background Art

[0002] Aluminum plate cutting devices play a crucial role in modern industrial production, and the stability of their performance and cutting accuracy directly affect the quality and production efficiency of products. However, in the actual production process, the performance of aluminum plate cutting devices is often affected by various factors, such as material characteristics, cutting parameters, equipment aging, improper operation, etc. These factors may all cause the device to malfunction. Traditional device monitoring methods mainly rely on manual regular inspections or experience-based judgments. This method is not only time-consuming and laborious but also often difficult to detect device abnormalities in a timely and accurate manner.

[0003] With the rapid development of computer technology and anomaly detection algorithms, it has become increasingly common to apply anomaly detection algorithms to device anomaly detection. It is not only time-saving and labor-saving but also has high detection accuracy and can detect device abnormalities in a timely manner. One of the important parameters of an aluminum plate cutting device is the cutting pressure value, that is, the pressure value of the cutting tool head. The pressure value of normal data indicates that the aluminum plate cutting device is in a normal state, while a pressure value different from the normal data indicates that the aluminum plate cutting device has an abnormal risk. As a simple and efficient anomaly detection algorithm, the Isolation Forest algorithm can monitor the operation of an aluminum plate cutting device by detecting the pressure value of the device.

[0004] The Isolation Forest algorithm is based on the following theoretical basis: The proportion of abnormal data in the total sample data is very small, normal data is densely distributed, while abnormal data is often isolated, and abnormal data usually differs greatly from normal data. Therefore, abnormal data is more easily separated and identified. The Isolation Forest algorithm constructs multiple isolation trees based on multiple sample data, calculates the average path length of each sample data in all isolation trees, obtains the anomaly score of the sample data, and uses the sample data with an anomaly score greater than the threshold as abnormal data.

[0005] The threshold is set according to the contamination ratio, which is the proportion of the estimated abnormal data samples in the total data samples and is an empirical value set manually. The higher the contamination ratio is set, the more estimated abnormal data there are, and the smaller the threshold is, so that more sample data is determined to be abnormal data. The lower the contamination ratio is set, the fewer estimated abnormal data there are, and the larger the threshold is, so that fewer sample data is determined to be abnormal data. However, since the contamination ratio is set manually, there may be unreasonable settings. If the contamination ratio is set too high, normal sample data will be considered abnormal data, resulting in false alarms of the device. If the contamination ratio is set too low, real abnormal data will be ignored, missing potential abnormalities of the device. Summary of the Invention

[0006] The present invention provides a method and system for abnormal monitoring of an aluminum plate cutting device, aiming to solve the technical problem that in the prior art, due to the unreasonable setting of the contamination ratio of the isolation forest model, the accuracy of the model is low, resulting in false alarms or missed alarms.

[0007] The method for abnormal monitoring of an aluminum plate cutting device according to the present invention includes the following steps:

[0008] Obtain a training set, where the training set includes multiple cutting pressure values;

[0009] Train the isolation forest model using the training set to obtain a trained isolation forest model; wherein, the threshold of the isolation forest model is an improved threshold; the improved threshold is inversely correlated with the mean value of the contamination ratios of all isolation trees in the isolation forest model; the contamination ratio is inversely correlated with the degree of correlation between the dispersion degree of each cutting pressure value in each isolation tree and the diversity of the aluminum plates; the degree of correlation is the arithmetic square root of the absolute value of the square difference between the dispersion degree and the diversity; the diversity is the variance of the weighted average value of the values obtained by normalizing each parameter corresponding to all aluminum plates in each isolation tree;

[0010] Wherein, when obtaining the weighted average value, the weight of the value obtained by normalizing each parameter is the value obtained by normalizing the correlation between the corresponding parameter and the cutting pressure value; the parameters include aluminum plate hardness, aluminum plate thickness, and aluminum plate surface friction;

[0011] Input the to-be-tested cutting pressure value into the trained isolation forest model and output the corresponding anomaly score; when the anomaly score is greater than the improved threshold, it is determined that there is an anomaly in the aluminum plate cutting device, and an alarm is given and the operation of the aluminum plate cutting device is stopped.

[0012] In the above solution, by calculating the dispersion degree of each cutting pressure value in each isolation tree and the diversity of the aluminum plates, the true contamination ratio in each isolation tree is obtained, and then the improved threshold is obtained. Using the isolation forest model with the improved threshold for anomaly detection can improve the accuracy of anomaly detection and prevent false alarms or missed alarms.

[0013] Preferably, the improved threshold is:

[0014]

[0015] In the formula, is the contamination ratio of the th isolation tree, is the total number of isolation trees, is the total number of cutting pressure values in the training set, is the standard normalization function.

[0016] Preferably, the Pollution ratio of an isolated tree is:

[0017] ;

[0018] In the formula, is the correlation degree between the dispersion degree of each cutting pressure value in the th isolated tree and the diversity of the aluminum plate, is an exponential function with the natural constant as the base.

[0019] In the above solution, characterizing the pollution ratio of the isolated tree by the correlation degree of the isolated tree can truly reflect the pollution situation of the isolated tree and make the calculation result more accurate.

[0020] Preferably, the dispersion degree of each cutting pressure value in the th isolated tree is:

[0021] ;

[0022] In the formula, is the total number of cutting pressure values in the th isolated tree, is the th cutting pressure value in the th isolated tree, is the mean value of all cutting pressure values in the th isolated tree.

[0023] In the above solution, by calculating the dispersion degree of the cutting pressure values in each isolated tree, the deviation degree of any cutting pressure value from the average cutting pressure value is reflected, which is convenient for the subsequent calculation of the pollution ratio.

[0024] Preferably, the dispersion degree of each cutting pressure value in the th isolated tree is:

[0025] ;

[0026] In the formula, is the total number of cutting pressure values in the th isolated tree, is the th cutting pressure value in the th isolated tree, is the th mean value of all cutting pressure values in the

[0027] Preferably, the weight of the th parameter of the aluminum plate is:

[0028] ;

[0029] In the formula, is the correlation between the th parameter of the aluminum plate and the corresponding cutting pressure value, is the total number of parameters of the aluminum plate.

[0030] Preferably, the anomaly score of the cutting pressure value to be measured is:

[0031]

[0032] In the formula, is the average path length of the cutting pressure value to be measured in all isolation trees, is the normalization factor of the expected path length, is the total number of cutting pressure values in the training set.

[0033] Preferably, the correlation between the parameters of the aluminum plate and the cutting pressure value is represented by the DTW correlation function or the Pearson correlation coefficient.

[0034] In the above solution, the DTW correlation function has the advantages of robustness and accurate calculation results. The Pearson correlation coefficient has the advantages of intuitiveness and simple calculation.

[0035] The present invention also provides an anomaly monitoring system for an aluminum plate cutting device, including a memory and a processor. The processor executes the computer program stored in the memory to implement the anomaly monitoring method for the aluminum plate cutting device described in any one of the above.

[0036] The beneficial effects are:

[0037] In the solution of the present invention, first, according to the dispersion degree of the cutting pressure value and the diversity of the aluminum plate, the correlation degree between the two is calculated. Then, the contamination ratio of each isolation tree is obtained. Next, based on the mean value of the contamination ratios of all isolation trees, an improved threshold is obtained. Finally, the data is anomaly detected by the isolation forest model with the improved threshold. Compared with the contamination ratio set artificially in the prior art, the present invention calculates the real contamination ratio in the data to obtain a suitable improved threshold in the isolation forest model, so that the isolation forest model can more accurately distinguish normal data and anomaly data, improving the accuracy of anomaly detection and reducing false alarms and missed alarms in the process of anomaly monitoring of the aluminum plate cutting device. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is the step flow chart of the anomaly monitoring method for the aluminum plate cutting device according to the embodiment of the present invention;

[0039] Figure 2 It is a structural block diagram of an abnormal monitoring system for an aluminum plate cutting device according to an embodiment of the present invention. Specific embodiments

[0040] The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0041] As Figure 1 shown, according to a first aspect of the present invention, there is provided an abnormal monitoring method for an aluminum plate cutting device, including the following steps:

[0042] S1. Obtain a training set, where the training set includes a plurality of cutting pressure values.

[0043] In this step, the cutting pressure value is the pressure value of the cutting tool head. During the process of cutting the aluminum plate, the cutting tool head is in direct contact with the aluminum plate, and the cutting tool head needs to bear a large pressure. The cutting tool head is also the most easily damaged part of the aluminum plate cutting device. The pressure value of the cutting tool head can reflect the operating state of the aluminum plate cutting device. The normal cutting pressure values account for the majority and are concentratedly distributed, indicating that the aluminum plate cutting device is in a normal operating state. The abnormal cutting pressure values are fewer and different from the normal data, indicating that the aluminum plate cutting device is in an abnormal operating state, and it is necessary to check and process the aluminum plate cutting device in time. The cutting pressure value can be collected by a pressure sensor provided on the aluminum plate cutting device.

[0044] S2. Use the training set to train an isolation forest model to obtain a trained isolation forest model.

[0045] It should be noted that the isolation forest algorithm itself belongs to the prior art, and its core idea is to "isolate" data points by constructing a series of random isolation trees. The isolation forest randomly selects features and split values, and gradually divides the data set to form a tree structure. Abnormal data points are more likely to be isolated due to their rarity and characteristics different from normal data points, so they show shorter path lengths in the isolation tree. The isolation forest algorithm includes two stages. The first stage is to train the isolation forest model, and the second stage is to use the trained isolation forest model for abnormal detection of data.

[0046] Among them, training the isolation forest model includes the following steps:

[0047] S21. Select the model parameters of the isolation forest model.

[0048] The model parameters of the isolation forest model include the number of isolation trees, the maximum tree depth, the subsample size, the random seed, and the contamination ratio. The more isolation trees there are, the higher the stability and accuracy of the isolation forest model usually are, but the computational cost will also increase. The larger the maximum tree depth, the better it can capture the details of the data, but it may lead to overfitting. The subsample size is the number of samples used in each isolation tree. A smaller subsample can improve the efficiency of the model, but may reduce the accuracy. The random seed is used to control the random process to ensure the repeatability of the results. The contamination ratio is the proportion of the estimated number of abnormal data in the total number of data, which is used to adjust the threshold of the anomaly score.

[0049] S22. Construct isolation trees according to the model parameters of the isolation forest model. Specifically, it includes the following steps:

[0050] S221. Randomly select a preset number of cutting pressure value data from the training set as the subsample and put it into the root node of the isolation tree;

[0051] S222. Randomly select any dimension of the data and randomly select a splitting value within the range of this dimension. Since there is only one dimension of cutting pressure value in the present invention, the splitting value is a cutting pressure value randomly selected within the range of the cutting pressure value.

[0052] S223. Compare all the data within the subsample with the splitting value respectively and assign them to the left subtree and the right subtree;

[0053] S224. Recursively divide the data in the subtree into the next-level subtree until the maximum tree depth is reached; for example, the maximum tree depth is 10.

[0054] S225. Repeat steps S221 to S224 until the specified number of isolation trees are constructed. For example, the number of isolation trees is 100.

[0055] In step S21, since the contamination ratio is set artificially, it often cannot truly reflect the proportion of abnormal data in the total data. The threshold obtained with the artificially set contamination ratio will cause false alarms or missed alarms in the aluminum plate cutting equipment. The specific situation is as follows:

[0056] When the contamination ratio is set too high, the isolation forest model will think that the proportion of abnormal data is large. Therefore, the condition for determining abnormal data will be loose, that is, the threshold will be small, so that the anomaly scores of more data are greater than the threshold to increase the number of abnormal data. This will cause some normal data to be misidentified as abnormal data by the isolation forest model, resulting in false alarms in the aluminum plate cutting equipment.

[0057] When the pollution ratio is set too low, the Isolation Forest model will consider that the proportion of abnormal data is small. Therefore, the conditions for determining abnormal data will be strict, that is, the threshold will be large, so that the abnormal scores of less data are greater than the threshold to reduce the number of abnormal data. This will cause some abnormal data to be misjudged as normal data by the Isolation Forest model, resulting in false negatives of the aluminum plate cutting equipment.

[0058] From the above analysis, it can be seen that when the Isolation Forest model in the prior art is used to monitor the abnormality of the aluminum plate cutting equipment, false positives or false negatives will occur. The main reason is that the artificially set pollution ratio is inappropriate and cannot truly reflect the proportion of abnormal data, resulting in an inappropriate setting of the threshold. Furthermore, the Isolation Forest model cannot accurately distinguish normal data from abnormal data, leading to false positives or false negatives of the aluminum plate cutting equipment.

[0059] Therefore, the present invention improves the artificially set pollution ratio in the prior art to obtain an appropriate threshold, enabling the Isolation Forest model to accurately distinguish normal data from abnormal data and avoiding false positives or false negatives of the aluminum plate cutting equipment. The specific steps are as follows:

[0060] S23. Obtain the pollution ratio of each isolation tree.

[0061] Since when constructing the isolation tree, the data of the sub-samples in each isolation tree are randomly selected from the training set, the pollution ratio of each isolation tree is different, and it is necessary to calculate the pollution ratio of each isolation tree. The pollution ratio is inversely correlated with the degree of dispersion of each cutting pressure value in each isolation tree and the degree of correlation with the diversity of the aluminum plates. Among them, the diversity of the aluminum plates is reflected in the fact that the aluminum plates are composed of different parameters, and the parameters include aluminum plate hardness, aluminum plate thickness, and aluminum plate surface friction, and these parameters will all affect the cutting pressure value. For example, the greater the hardness of the aluminum plate, the thicker the thickness, and the greater the surface friction, the greater the cutting pressure value. The smaller the hardness of the aluminum plate, the thinner the thickness, and the smaller the surface friction, the smaller the cutting pressure value. The diversity of the aluminum plates will directly affect the degree of dispersion of the cutting pressure values. The greater the diversity of the aluminum plates, the greater the degree of dispersion of the cutting pressure values. These cutting pressure values that seem abnormal but are actually normal are caused by the diversity of the aluminum plates, rather than by the abnormality of the aluminum plate cutting equipment. And the higher the degree of correlation between the diversity of the aluminum plates and the degree of dispersion of the cutting pressure values, the greater the possibility that these data are normal data, that is, the actual pollution ratio is smaller. Therefore, by calculating the degree of correlation between the degree of dispersion of the cutting pressure values in each isolation tree and the diversity of the aluminum plates, it can be used to represent the pollution ratio of that isolation tree. Specifically, obtaining the pollution ratio of each isolation tree includes the following steps:

[0062] S231. Obtain the degree of dispersion of the cutting pressure values.

[0063] The Degree of dispersion of cutting pressure values in each isolated tree is as follows:

[0064] ;

[0065] In the formula, is the total number of cutting pressure values in the th isolated tree, is the th cutting pressure value in the th isolated tree, is the mean value of all cutting pressure values in the th isolated tree.

[0066] In some alternative embodiments, the degree of dispersion can also be represented by the standard deviation. Then, the degree of dispersion of cutting pressure values in the th isolated tree is as follows:

[0067] ;

[0068] In the formula, is the total number of cutting pressure values in the th isolated tree, is the th cutting pressure value in the th isolated tree, is the mean value of all cutting pressure values in the th isolated tree.

[0069] In this step, by calculating the degree of dispersion of cutting pressure values in each isolated tree, the deviation degree of any cutting pressure value from the average cutting pressure value is reflected, which is convenient for the subsequent calculation of the pollution ratio.

[0070] S232. Obtain the diversity of the aluminum plate.

[0071] The diversity of the aluminum plate is the variance of the weighted average of the parameters corresponding to all aluminum plates in each isolated tree. Obtaining the diversity of the aluminum plate includes the following steps:

[0072] S2321. Normalize the parameters of the aluminum plate.

[0073] The parameters of the aluminum plate include aluminum plate hardness, aluminum plate thickness, and aluminum plate surface friction. In order to eliminate the dimensional difference between different parameters and facilitate numerical calculation and analysis, it is necessary to first normalize the parameters of the collected aluminum plate so that each parameter falls within the range of 0 to 1.

[0074] S2322. Calculate the weights of the normalized values of each parameter.

[0075] The weight of the normalized value of each parameter is the normalized value of the correlation between the corresponding parameter and the cutting pressure value. Then, for the th parameter of the aluminum plate, the weight is:

[0076] ;

[0077] In the formula, is the correlation between the th parameter of the aluminum plate and the corresponding cutting pressure value, and is the total number of parameters of the aluminum plate.

[0078] Specifically, taking the three parameters of aluminum plate hardness, aluminum plate thickness, and aluminum plate surface friction as an example, the weight of aluminum plate hardness is set as , the weight of aluminum plate thickness is set as , and the weight of aluminum plate surface friction is set as .

[0079] The weight of aluminum plate hardness is:

[0080] ;

[0081] The weight of aluminum plate thickness is:

[0082] ;

[0083] The weight of aluminum plate surface friction is:

[0084] ;

[0085] In the formula, is the correlation between the hardness of the aluminum plate and the corresponding cutting pressure value, is the correlation between the thickness of the aluminum plate and the corresponding cutting pressure value, and is the correlation between the surface friction of the aluminum plate and the corresponding cutting pressure value.

[0086] The correlation between the parameters of the aluminum plate and the cutting pressure value is expressed by the DTW correlation function or the Pearson correlation coefficient.

[0087] S2323. Obtain the diversity of aluminum plates in each isolated tree.

[0088] The diversity is the variance of the weighted average of the normalized values of all parameters corresponding to the aluminum plates in each isolated tree. Then, the diversity of the aluminum plates in the th isolated tree is:

[0089] ;

[0090] In the formula, is the weight of the hardness of the aluminum plate, is the weight of the thickness of the aluminum plate, is the weight of the surface friction of the aluminum plate, is the value after normalization of the hardness of the aluminum plate, is the value after normalization of the thickness of the aluminum plate, is the value after normalization of the surface friction of the aluminum plate, is the variance function.

[0091] In some alternative embodiments, the diversity of the aluminum plates can also be the coefficient of variation of the weighted average of the values after normalization of each parameter corresponding to all the aluminum plates in each isolated tree.

[0092] S233. Calculate the degree of dispersion of each cutting pressure value in each isolated tree and the degree of correlation with the diversity of the aluminum plates.

[0093] The degree of correlation is the arithmetic square root of the absolute value of the square difference between the degree of dispersion and the diversity. Then, for the th isolated tree, the degree of correlation between the degree of dispersion of each cutting pressure value and the diversity of the aluminum plates is:

[0094] ;

[0095] In the formula, is the diversity of the aluminum plates in the th isolated tree, is the th isolated tree, and is the degree of dispersion of the cutting pressure value.

[0096] S234. Obtain the pollution ratio of each isolated tree.

[0097] The pollution ratio is inversely correlated with the degree of correlation of each isolated tree. Specifically, the pollution ratio of the th isolated tree is:

[0098] ;

[0099] In the formula, is the degree of correlation between the degree of dispersion of each cutting pressure value and the diversity of the aluminum plates in the th isolated tree, is the exponential function with the natural constant as the base.

[0100] S24. Calculate the improvement threshold based on the pollution ratio.

[0101] The improvement threshold is inversely correlated with the mean value of the pollution ratios of all isolated trees. Specifically, the improvement threshold is as follows:

[0102]

[0103] Wherein, is the pollution ratio of the th isolated tree, is the total number of isolated trees, is the total number of cutting pressure values in the training set, is the standard normalization function.

[0104] By calculating the degree of dispersion of the cutting pressure values and the degree of correlation of the diversity of the aluminum plates, the present invention obtains the pollution ratio of each isolated tree, and then obtains the improved threshold value with the average value of the pollution ratios of all isolated trees. Compared with the threshold value obtained by artificially setting the pollution ratio in the prior art, the setting of the improved threshold value is more reasonable, enabling the isolation forest model to accurately distinguish normal data and abnormal data, and avoiding false alarms or missed alarms of the aluminum plate cutting equipment.

[0105] S3. Input the to-be-tested cutting pressure value into the trained isolation forest model and output the corresponding anomaly score; when the anomaly score is greater than the improved threshold value, it is determined that there is an anomaly in the aluminum plate cutting equipment, and an alarm is given and the operation of the aluminum plate cutting equipment is stopped.

[0106] After the training of the isolation forest model is completed, the cutting methods of these isolated trees are fixed, that is, the splitting values of each node in the isolated trees are determined. The to-be-tested cutting pressure value is transmitted into the trained isolation forest model, so that the to-be-tested cutting pressure value traverses each isolated tree, and the path length of the to-be-tested cutting pressure value in all isolated trees is obtained. According to the anomaly score calculation formula, the anomaly score of the to-be-tested cutting pressure value can be obtained.

[0107] The anomaly score of the to-be-tested cutting pressure value is as follows: is as follows:

[0108]

[0109] Wherein, is the average path length of the to-be-tested cutting pressure value in all isolated trees, is the normalization factor of the expected path length, is the total number of cutting pressure values in the training set.

[0110] In some embodiments, The calculation formula of

[0111]

[0112] Wherein, , is the total number of cutting pressure values in the training set.

[0113] In some embodiments, when the anomaly score is greater than the improvement threshold, but the difference between the anomaly score and the improvement threshold is less than the preset value, only a warning signal is issued to remind the operator to check whether there is an abnormal risk in the equipment. When the difference between the anomaly score and the improvement threshold is greater than or equal to the preset value, the operation of the aluminum plate cutting equipment is stopped. Through these measures, the abnormal conditions of the aluminum plate cutting equipment can be detected in time, avoiding equipment damage and property losses.

[0114] In some embodiments, the detected abnormal data is further analyzed to determine the cause and impact of the anomaly. For example, visualization tools can be used to display the distribution and characteristics of the abnormal data to better analyze and understand the abnormal data.

[0115] In the abnormal monitoring method of the aluminum plate cutting equipment of the present invention, by calculating the degree of dispersion of the cutting pressure value and the degree of correlation of the diversity of the aluminum plate, the contamination ratio of each isolation tree is adaptively adjusted, and then according to the mean value of the contamination ratios of all isolation trees, an improvement threshold is obtained. Using the isolation forest model with the improvement threshold for anomaly detection can more accurately distinguish normal data from abnormal data, improve the accuracy of anomaly detection, reduce false alarms and missed alarms, and make the abnormal monitoring method of the aluminum plate cutting equipment of the present invention more reliable.

[0116] As Figure 2 shown, according to the second aspect of the present invention, an abnormal monitoring system for an aluminum plate cutting equipment is further provided. The system includes a memory and a processor, and the processor executes the computer program stored in the memory to implement the abnormal monitoring method for the aluminum plate cutting equipment described in the first aspect of the present invention.

[0117] The system further includes a communication bus, a communication interface and other components well known to those skilled in the art. Their settings and functions are known in the art, so they will not be described in detail here.

[0118] In the present invention, the foregoing memory may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as, for example, a resistive random access memory (RRAM), a dynamic random access memory (DRAM), a static random access memory (SRAM), an enhanced dynamic random access memory (EDRAM), a high-bandwidth memory (HBM), a hybrid memory cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device. Any application or module described in the present invention may be implemented by computer-readable / executable instructions stored or otherwise held by such a computer-readable medium.

[0119] Although this specification has shown and described several embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many variations, changes, and alternative approaches will occur to those skilled in the art without departing from the spirit and scope of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

Claims

1. A method for monitoring abnormality of aluminum plate cutting equipment, characterized in that: The steps include: Acquire a training set, wherein the training set includes a plurality of cutting pressure values; The isolation forest model is trained using the training set to obtain a trained isolation forest model; the threshold of the isolation forest model is the improved threshold, and the improved threshold The calculation formula is: In the formula, For the The pollution ratio of an isolated tree, is the total number of isolated trees, is the total number of cutting pressure values ​​in the training set, is the standard normalization function; The contamination ratio is inversely correlated with the discreteness of each cutting pressure value in each isolated tree and the correlation degree of the diversity of aluminum plates; the correlation degree is the arithmetic square root of the absolute value of the square difference between the discreteness and the diversity; the diversity is the variance of the weighted average of the normalized values ​​of each parameter corresponding to all aluminum plates in each isolated tree; Wherein, when obtaining the weighted average value, the weight of the normalized value of each parameter is the normalized value of the correlation between the corresponding parameter and the cutting pressure value; the parameters include the hardness of the aluminum plate, the thickness of the aluminum plate and the surface friction of the aluminum plate; The cutting pressure value to be measured is input into the trained isolation forest model, and the corresponding anomaly score is output; when the anomaly score is greater than the improved threshold, it is determined that the aluminum plate cutting equipment is abnormal, and an early warning and stop of the aluminum plate cutting equipment are issued.

2. The method for monitoring abnormality of aluminum plate cutting equipment according to claim 1, characterized in that: No. Pollution ratio of an isolated tree for: ； In the formula, For the The correlation between the discreteness of each cutting pressure value in an isolated tree and the diversity of aluminum plates, The natural constant The exponential function of base .

3. The method for monitoring abnormality of aluminum plate cutting equipment according to claim 1, characterized in that: No. The discrete degree of each cutting pressure value in an isolated tree for: ； In the formula, For the The total number of cutting pressure values ​​in an isolated tree, For the Isolated tree Cutting pressure value, For the The mean of all cutting pressure values ​​in an isolated tree.

4. The method for monitoring abnormality of aluminum plate cutting equipment according to claim 1, characterized in that: No. The discrete degree of each cutting pressure value in an isolated tree for: ； In the formula, For the The total number of cutting pressure values ​​in an isolated tree, For the Isolated tree Cutting pressure value, For the The mean of all cutting pressure values ​​in an isolated tree.

5. The method for monitoring abnormality of aluminum plate cutting equipment according to claim 1, characterized in that: Aluminum Plate The weight of the parameters for: ； In the formula, For aluminum plate The correlation between the parameters and the corresponding cutting pressure value, is the total number of parameters of the aluminum plate.

6. The method for monitoring abnormality of aluminum plate cutting equipment according to claim 1, characterized in that: Cutting pressure value to be measured Anomaly score for: In the formula, is the cutting pressure value to be measured The average path length among all isolated trees, is the normalization factor for the expected path length, is the total number of cutting pressure values ​​in the training set.

7. The method for monitoring abnormality of aluminum plate cutting equipment according to claim 1, characterized in that: The correlation between the parameters of the aluminum plate and the cutting pressure value is expressed by the DTW correlation function or the Pearson correlation coefficient.

8. An abnormality monitoring system for aluminum plate cutting equipment, comprising a memory and a processor, characterized in that: The processor executes the computer program stored in the memory to implement the abnormality monitoring method for aluminum plate cutting equipment according to any one of claims 1 to 7.

Images