A device and method for real-time monitoring of a die wear state

Abstract

The application provides a device and method for monitoring the wear state of a mold in real time, and belongs to the field of equipment safety monitoring.The device comprises a detection unit and a control unit, wherein the abrasive particle sensor of the detection unit is used to distinguish ferromagnetic particles and non-ferromagnetic particles in lubricating oil and obtain the particle size distribution curve of the ferromagnetic particles; the particle analyzer is used to obtain the particle size distribution curve of particles of different materials in the lubricating oil, and then the particle size distribution curve of the ferromagnetic particles is used for distinguishing and identifying optimization, so as to obtain the particle size distribution curve of the abrasive particles of the mold, and the wear state of the stamping mold is monitored in real time according to the particle size distribution curve; and the control unit is used to control the detection unit and display the monitoring result.The application considers that the source of the abrasive particles in the stamping mold operation is not only one kind, and the abrasive particle sensor cannot be simply used for monitoring to judge the wear condition, and further puts forward a monitoring device comprising the abrasive particle sensor and the particle analyzer, so as to improve the monitoring precision of the wear state of the stamping mold.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of equipment safety monitoring, and more particularly, to a device and method for real-time monitoring of die wear state. BACKGROUND

[0002] The stamping die needs to be in high-frequency contact with the workpiece during operation, and each set of stamping die needs to produce up to thousands of workpieces, so the wear of the stamping die is very serious, especially the blade edge part. The wear condition of the stamping die needs to be monitored in real time during production, otherwise not only a large number of production pieces will not meet the standards, but also the machine will be shut down due to unprepared die, thereby causing huge economic losses.

[0003] CN205887899U discloses a stamping die wear state monitoring device based on image processing technology, which scans the stamping die after each stamping is completed, and compares it with the original image for analysis, thereby realizing real-time monitoring of the die wear state. However, this method has the problems of difficulty in judgment, low precision, and the risk of huge economic losses caused by shutdown due to misjudgment for high-precision parts. SUMMARY

[0004] In view of the defects of the prior art, the purpose of the present application is to provide a device and method for real-time monitoring of die wear state, aiming to solve the problems of complex process and low judgment precision of the existing stamping die wear monitoring method.

[0005] To achieve the above-mentioned purpose, the present application provides a device for real-time monitoring of die wear state, which comprises a detection unit and a control unit, wherein the detection unit is connected with the lubricating oil outlet of the stamping module, and is used for detecting the lubricating oil to judge the wear state of the stamping die, which comprises a wear particle sensor and a particle analyzer. The wear particle sensor is used to distinguish ferromagnetic particles and non-ferromagnetic particles in the lubricating oil, and obtain the particle size distribution curve of the ferromagnetic particles. The particle analyzer is used to obtain the particle size distribution curve of different material particles in the lubricating oil, and then distinguish and identify the different material particles by using the particle size distribution curve of the ferromagnetic particles, so as to obtain the particle size distribution curve of the die wear particles, and monitor the wear state of the stamping die in real time according to it. The control unit is used to control the detection unit and display the monitoring result.

[0006] As a further preferred, the device for real-time monitoring of die wear state further comprises a sound sensor, which is arranged in the interior of the stamping module, and is used to judge whether the stamping die appears the phenomenon of blade collapse.

[0007] As a further preferred embodiment, the device for real-time monitoring of mold wear also includes a moisture sensor and a temperature sensor. The moisture sensor is located after the abrasive sensor and particle analyzer, so that the lubricating oil enters the moisture sensor last, for monitoring the moisture content of the lubricating oil. The temperature sensor is located inside the stamping module for monitoring the temperature of the lubricating oil.

[0008] As a further preferred embodiment, the device for real-time monitoring of mold wear status further includes a first liquid level sensor and a second liquid level sensor. The first liquid level sensor is disposed inside the stamping module and is used to monitor the level of lubricating oil in the stamping module. The second liquid level sensor is disposed inside the oil tank and is used to monitor the level of lubricating oil in the oil tank.

[0009] As a further preferred embodiment, the control unit includes an electrical signal connection processing module, a pressure control module, and a terminal display module. The electrical signal connection processing module is used to receive and process monitoring information. The pressure control module and the terminal display module are connected to the electrical signal connection processing module and are used to adjust the stamping parameters and display the monitoring results, respectively.

[0010] As a further preferred embodiment, the particle analyzer uses the K-means clustering algorithm to identify particles of different materials in the lubricating oil.

[0011] According to another aspect of the present invention, a method for real-time monitoring of mold wear condition is provided, the method comprising the following steps:

[0012] S1. Lubricating oil is introduced into the abrasive sensor and the particle analyzer respectively. The abrasive sensor is used to distinguish between ferromagnetic particles and non-ferromagnetic particles in the lubricating oil, thereby obtaining the particle size distribution curve of the ferromagnetic particles.

[0013] S2 uses the particle analyzer to identify particles of different materials in the lubricating oil to obtain two particle size distribution curves. Then, based on the changing trend of the particle size distribution curve of the ferromagnetic particles, it distinguishes them and selects the particle size distribution curve with the same changing trend as the particle size distribution curve of the mold abrasive.

[0014] S3 calculates the absolute error between the particle size distribution curve of the die abrasive and the particle size distribution curve of the ferromagnetic particles. If the absolute error is greater than a threshold, return to step S2 to optimize the identification algorithm of the particle analyzer; if the absolute error is less than or equal to the threshold, monitor the wear state of the stamping die in real time according to the particle size distribution curve of the die abrasive.

[0015] As a further preferred embodiment, in step S3, when the maximum particle size of the die abrasive grains is greater than 15 to 100 μm and the average number of die abrasive grains increases to 3 to 5 times the average number during the stable period, it is determined that the stamping die is severely worn and needs to be replaced.

[0016] As a further preferred embodiment, the method for real-time monitoring of die wear status further includes step S4, which specifically involves: establishing a combined neural network model, using stamping speed, stamping load, and die clearance as inputs, and stamping die life as output, to predict the life of stamping dies.

[0017] As a further preferred embodiment, in step S4, the mean squared error loss function is used to train the combined neural network model using three parameters: stamping speed, stamping load, and die clearance. The weight matrix and bias vector of the combined neural network model are obtained through mini-batch stochastic gradient descent, thereby optimizing the combined neural network model.

[0018] In summary, compared with the prior art, the above-described technical solutions conceived by this invention have the following advantages:

[0019] Beneficial effects:

[0020] 1. This invention takes into account that the sources of abrasive particles in stamping die operations are not singular, and that abrasive particle sensors cannot be simply used to monitor and determine the wear condition. Therefore, a monitoring device including an abrasive particle sensor and a particle analyzer is proposed. The abrasive particle sensor is used to distinguish between ferromagnetic particles and non-ferromagnetic particles to eliminate the influence of impurities. The particle analyzer is used to obtain the particle size distribution curves of die abrasive particles and sheet metal abrasive particles. The changing trends of the particle size distribution curves of ferromagnetic particles are used to distinguish, identify, and optimize them. Finally, the particle size distribution curves of die abrasive particles are used for monitoring, thereby improving the monitoring accuracy of the wear condition of stamping dies.

[0021] 2. At the same time, by setting a first liquid level sensor, a second liquid level sensor, a moisture sensor, a sound sensor, and a temperature sensor, the present invention achieves comprehensive monitoring and improves the reliability of judgment by improving the monitoring data of the wear state of the stamping die;

[0022] 3. In addition, the present invention also proposes a method for real-time monitoring of die wear. This method establishes a combined neural network model and uses stamping speed, stamping load, and die clearance as input and training parameters, which can effectively improve the accuracy of the combined neural network model and achieve accurate prediction of stamping die life. Attached Figure Description

[0023] Figure 1 This is an application diagram of the device for real-time monitoring of mold wear status provided in an embodiment of the present invention;

[0024] Figure 2 This is a flowchart illustrating the method for real-time monitoring of mold wear status provided in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the neural network topology used in a preferred embodiment of the present invention.

[0026] In all the accompanying drawings, the same reference numerals are used to denote the same elements or structures, wherein:

[0027] 1-Oil tank, 2-Hydraulic pump, 3-Pressure regulating valve, 4-Second filter screen, 5-Pressing module, 6-Sound sensor, 7-First liquid level sensor, 8-Temperature sensor, 9-Abrasive sensor, 10-Particle analyzer, 11-Moisture sensor, 12-First filter screen, 13-Second liquid level sensor, 14-Electrical signal connection processing module, 15-Pressure control module, 16-Terminal display module. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0029] Considering that lubricating oil is needed at the cutting edge of stamping dies to reduce wear and cool down, monitoring the particle size and quantity of die abrasive particles in the lubricating oil can achieve the monitoring of the wear condition of stamping dies. This method also makes good use of the lubricating oil during the stamping process, greatly maximizing its practicality and significantly improving the convenience of monitoring. However, the abrasive environment in the lubricating oil of stamping dies is complex, containing die abrasive particles, sheet metal particles, and impurity particles. Abrasive sensors alone cannot accurately monitor the wear condition of stamping dies, easily leading to misjudgments and disrupting normal operations. Therefore, this invention proposes a dual detection system using an abrasive sensor and a particle analyzer. Ultimately, the pixel features of die abrasive particles are accurately identified through the image from the particle analyzer, thereby achieving accurate real-time monitoring of die abrasive particles and improving the reliability of judging the wear condition of stamping dies.

[0030] like Figure 1 As shown, the present invention provides a device for real-time monitoring of the wear state of a die. The device includes a detection unit and a control unit. The detection unit is used to connect to the lubricating oil outlet of the stamping module 5 and to detect the lubricating oil to determine the wear state of the stamping die. It includes an abrasive sensor 9 and a particle analyzer 10. The abrasive sensor 9 is used to distinguish between ferromagnetic particles and non-ferromagnetic particles in the lubricating oil and to obtain the particle size distribution curve of the ferromagnetic particles to eliminate the influence of impurities.

[0031] The particle analyzer 10 monitors the image of the lubricating oil. After noise reduction, it uses artificial intelligence algorithms for identification to obtain the particle size distribution curve of different materials in the lubricating oil. Then, it uses the particle size distribution curve of ferromagnetic particles to distinguish and identify particles of different materials, thereby obtaining the particle size distribution curve of die abrasives. Based on its real-time monitoring of the wear status of the stamping die, when the particle size of the die abrasives is greater than 15-100μm and the average number of die abrasives rises to 3-5 times the average number during the stable period, it is determined that the stamping die is severely worn and needs to be replaced.

[0032] The control unit is used to control the detection unit and display the monitoring results. It includes an electrical signal connection processing module 14, a pressure control module 15, and a terminal display module 16. The electrical signal connection processing module 14 is connected to the abrasive sensor 9 and the particle analyzer 10 to receive and process the monitoring information. The pressure control module 15 and the terminal display module 16 are connected to the electrical signal connection processing module 14. The pressure control module 15 is used to adjust the stamping speed and stamping load of the press according to the monitoring results. The electrical signal connection processing module 14 is used to display the monitoring results to realize the human-machine interaction function.

[0033] Furthermore, the device for real-time monitoring of mold wear also includes a first liquid level sensor 7, a second liquid level sensor 13, a moisture sensor 11, a temperature sensor 8, and a sound sensor 6. All of these sensors are connected to the electrical signal processing module 14. Specifically: the first liquid level sensor 7 is located inside the stamping module 5 to monitor the level of lubricating oil in the stamping module 5, preventing blockage of the oil outlet pipe of the stamping module 5; the second liquid level sensor 13 is located inside the oil tank 1 to monitor the level of lubricating oil in the oil tank 1, ensuring that the lubricating oil level remains within a certain range; the moisture sensor 11 is located after the abrasive sensor 9 and the particle analyzer 10, ensuring that the lubricating oil... Finally, the moisture sensor 11 is used to monitor the moisture content of the lubricating oil. The temperature sensor 8 is located inside the stamping module 5 to monitor the temperature of the lubricating oil. The quality of the lubricating oil is ensured by the moisture sensor 11 and the temperature sensor 8. When the temperature T > 80℃ or the water content W > 1wt%, an alarm is triggered, indicating that the lubricating oil is abnormal and prompting the lubricating oil to be replaced. The sound sensor 6 and the temperature sensor 8 are respectively located inside the stamping module 5. When particles larger than 500μm appear in the die abrasive, or when the amplitude of the information monitored by the sound sensor 6 is greater than 90dB, it is determined that the stamping die has a chipping phenomenon, the machine stops running, an alarm is triggered, and the worker is prompted to check.

[0034] Furthermore, the particle analyzer 10 employs the K-means clustering algorithm to identify particles of different materials in the lubricating oil. Specifically, it first initializes k cluster centers, then summarizes the samples belonging to each cluster based on the calculated distance between the sample and the center point, iteratively minimizing the distance between the sample and its assigned cluster center (objective function) to complete the identification and segmentation of mold abrasive particles. Next, it uses an eight-connected region morphology method based on region growing to finely segment, label, and count each abrasive particle. The data detected by the abrasive particle sensor 9 and the data calculated by the K-means clustering algorithm are plotted as curves, and then fitted using a Weibull distribution. The number curves of particles with the same diameter range are compared. Based on the data detected by the abrasive particle sensor 9, the accuracy of the K-means clustering algorithm for image analysis is continuously improved. According to the trend that the particle size of mold abrasive particles increases and their quantity gradually increases, the pixel features of the mold abrasive particles are identified.

[0035] This invention uses a dual detection system of abrasive sensor 9 and particle analyzer 10 to measure the number, size, and material of abrasive particles in lubricating oil. Then, an artificial intelligence algorithm is used to improve the recognition accuracy of the images acquired by particle analyzer 10 by using the particle size distribution curve of ferromagnetic particles measured by the abrasive sensor as a basis. This allows for the identification of mold abrasive particles and real-time monitoring of mold abrasive particle data to determine the wear condition of the stamping die.

[0036] In practical applications, a lubricating oil collection assembly is built for the stamping module 5 to ensure that the wear area of ​​the stamping die is completely flushed, and that the lubricating oil is completely recovered into the oil outlet pipe after passing through the sheet metal and the lower die. The oil outlet pipe is then connected in sequence to the abrasive sensor 9, the particle analyzer 10, and the moisture sensor 11 to detect the lubricating oil. The lubricating oil is then filtered through the first filter screen 12 and sent back to the oil tank 1. At the same time, the oil tank 1, under the control of the hydraulic pump 2 and the pressure regulating valve 3, sends the lubricating oil into the second filter screen 4 for filtration, and then sends it back into the stamping module 5 to continue working.

[0037] like Figure 2 As shown, according to another aspect of the present invention, a method for real-time monitoring of mold wear condition is provided, the method comprising the following steps:

[0038] S1 introduces lubricating oil into the abrasive sensor and the particle analyzer respectively. The abrasive sensor distinguishes between ferromagnetic and non-ferromagnetic particles in the lubricating oil and counts the number of particles according to their particle size d (d<5μm, 5μm≤d≤15μm, d>15μm). The number of particles within the particle size range is counted as a curve and fitted with a Weibull distribution to obtain the particle size distribution curve of the ferromagnetic particles.

[0039] S2 uses a particle analyzer to obtain images of lubricating oil. After image noise reduction, a neural network model is used to identify particles of different materials (die abrasives and sheet metal abrasives). Then, the number of particles is counted according to particle size D (D<5μm, 5μm≤D≤15μm, D>15μm). The number of particles within the particle size range is counted as a curve, and a Weibull distribution is used to fit it to obtain two particle size distribution curves. Considering that the stamping die is in the break-in period in the early stage of operation, the particle size distribution of die abrasives is unstable, while the change trend of sheet metal abrasives is relatively stable. Therefore, by comparing the change trend of the number of particles of the same size, the particle size distribution curve that is consistent with the change trend of ferromagnetic particles can be selected to obtain the particle size distribution curve of die abrasives.

[0040] S3 calculates the absolute error between the particle size distribution curve of the die abrasive and the particle size distribution curve of the ferromagnetic particles. If the absolute error is greater than the threshold, return to step S2 to optimize the recognition algorithm of the neural network model to improve the accuracy of image recognition. If the absolute error is less than or equal to the threshold, monitor the wear status of the stamping die in real time according to the particle size distribution curve of the die abrasive. When the maximum particle size of the die abrasive is greater than 15 to 100 μm and the average number of die abrasive particles rises to 3 to 5 times the average number during the stable period, it is determined that the stamping die is severely worn and needs to be replaced.

[0041] S4 establishes a combined neural network model, trains it using collected stamping speed, stamping load, and die clearance, and uses the trained combined neural network model to predict the life of stamping dies.

[0042] Furthermore, the neural network model is used to identify particles of different materials, specifically as follows:

[0043] The abrasive particle pixel image obtained from the particle analyzer is converted into grayscale values. K-means clustering is then performed on the grayscale values ​​at different locations in the abrasive particle pixel image. K-means is an iterative clustering analysis algorithm, belonging to the unsupervised learning algorithm in machine learning. It is a partitioning-based clustering algorithm. The principle is to first initialize k cluster centers, then summarize the samples belonging to each cluster based on the distance between the sample and the center point, iteratively minimizing the distance between the sample and its cluster center (objective function). Let the abrasive particle pixel image grayscale dataset be D = {x_1, x_2, ..., x_m}, and the clusters obtained by clustering using the k-means algorithm be C = {c_1, c_2, ..., c_k}, with the objective function being:

[0044]

[0045] In the formula, c j It is the j-th cluster, x i It is C j Each pixel value in the cluster, u jIt is C j The center value of all pixel values ​​in the cluster, i.e., the average value; Indicates the cth j The sum of the squared differences between all pixel values ​​in the cluster and the center value. The expression represents the sum of squared differences between the center value and the value of all pixels in all clusters; arg min J(c) is the expression that, after multiple iterations, makes the sum of squared differences between the center value and the value of all pixels in all clusters. When the distance between a sample and the center of its assigned cluster is minimized, the k clusters are formed; that is, the goal of the K-means clustering algorithm is to minimize the distance between a sample and the center of its assigned cluster.

[0046] The steps of the K-means clustering algorithm are as follows:

[0047] i. Randomly select k pixel grayscale value samples as the initial cluster centers (k is a hyperparameter representing the number of clusters. The value can be determined based on prior knowledge or verification methods);

[0048] ii. For each sample in the dataset, calculate its distance to the k cluster centers and assign it to the class corresponding to the cluster center with the smallest distance;

[0049] iii. For each cluster, recalculate its cluster center position using the following formula.

[0050]

[0051] In the formula, u j t+1 It is the next c after the iteration. j The center value (average) of all pixel values ​​in the cluster; |c j | is the cth j The number of all objects (pixels) in the cluster;

[0052] iv. Repeat steps ii and iii until the cluster center position remains unchanged.

[0053] Since the grayscale values ​​of mold abrasive particles in the abrasive particle pixel image are different from those of other impurities or plate abrasive particles, the K-means clustering algorithm can be used to identify and segment the mold abrasive particles.

[0054] After identifying and segmenting the abrasive particles in the mold using the K-means clustering algorithm, an octet-connected region morphology method based on region growing is then used to further segment, count, and count each abrasive particle, as detailed below:

[0055] I. Using an 8-connected region labeling method based on region growing, seed pixels are placed into the segmented abrasive grain region of the mold, serving as the starting point for growth. Pixels in the surrounding neighborhood of the seed pixel that have the same or similar properties are merged into the region where the seed pixel is located. New pixels continue to grow outwards as seeds until no more pixels meeting the conditions can be merged, at which point it is considered an independent region (i.e., a single abrasive grain). In this way, the abrasive grain region can be segmented using the K-means clustering algorithm, and a corresponding program can be used to count the data: from the start of seed placement to the end of seed merging, the count is incremented by 1.

[0056] II. Count the number of pixels in each abrasive region, obtain the abrasive particle size based on each pixel, and then calculate the number of abrasive particles within a certain range of particle sizes;

[0057] III uses a counting algorithm program to classify and statistically analyze the number of abrasive particles calculated in II according to different particle sizes, plots the variation curves, and then fits them using a Weibull distribution.

[0058] Because the size and number of abrasive particles in a mold change according to certain patterns during use, while the abrasive particles in a sheet metal remain stable over a period of time, the fitting curve generated by the abrasive particle sensor (i.e., the particle size distribution curve of ferromagnetic particles) is compared with the fitting curve calculated and statistically analyzed by the algorithm to verify and correct the hyperparameter k of the K-means clustering algorithm, the number of iterations to calculate the new cluster centers (ultimately stabilizing the clustering), and the accuracy of the K-means clustering algorithm. This process continues until a fitting curve shows a consistent trend, at which point the abrasive particles can be identified as mold abrasive particles.

[0059] Further, step S4 specifically involves: inputting the stamping speed and stamping load to the connection processing module in real time, manually inputting the stamping die clearance, using these three variables as input variables for neural network 1, establishing two hidden layers, and outputting the quantity corresponding to each particle size range of the die abrasive. This output variable is then used as input variables for neural network 2, establishing two hidden layers, and finally outputting the stamping die life. That is, this method uses a combination of neural networks 1 and 2, with the following topology: Figure 3 ;

[0060] Neural network 1 was trained using experimental labeled data (X, Y), where X1 is the stamping speed, X2 is the stamping load, X3 is the stamping die clearance, and Y1~Y n The loss function is the root mean square error loss function, which represents the quantity corresponding to each particle size range of the abrasive grains in the mold.

[0061]

[0062] In the formula, E1 is the mean squared error after L2 regularization. Let f be the mean squared error of each batch of m data, where m is the number of data points in each batch used to optimize the weight parameters using gradient descent. 1k (x) represents the predicted value (output value) of neural network 1, Y k Let W be the true value of neural network 1, λ1∈(0,1) be the hyperparameter of the regularization term, and W be the true value of neural network 1. i Let i be the weight parameter;

[0063] Similarly, neural network 2 was trained using experimental labeled data (Y, Z), where Z represents the stamping die life, and the loss function is the mean squared error loss.

[0064]

[0065] In the formula, E2 is the mean squared error after L2 regularization. f is the mean squared error of each batch of n data points, where n is the number of data points in each batch used to optimize the weight parameters using gradient descent. 2k (y) represents the predicted value (output value) of neural network 2, Z k Let W be the true value of neural network 2, λ2∈(0,1) be the hyperparameter of the regularization term, and W be the true value of neural network 2. j Let j be the weight parameter;

[0066] The weight matrices and bias vectors of neural networks 1 and 2 are optimized using the mini-batch stochastic gradient descent method, thus learning and optimizing the combined neural network. By using this combined neural network, and inputting stamping speed, stamping load, and die clearance, the die wear state and die life can be predicted.

[0067] Using a genetic algorithm combined with the aforementioned combined neural network, the initial encoding, initialization of population, fitness evaluation, selection, crossover, and mutation of input variables (stamping speed, stamping load, and die clearance) are continuously iterated until the die life tends to stabilize. At this point, the die life is the maximum die life, and the input variables (stamping speed, stamping load, and die clearance) corresponding to this maximum die life are the optimal stamping parameters.

[0068] Furthermore, all replaceable parts of the stamping die are manually numbered. After all parts have been replaced, the service life of each part is calculated, thereby predicting the expected service life of the stamping die. By replacing the die multiple times, the prediction becomes more accurate. The service life progress bar of each numbered part is displayed in real time on the terminal display module 16, which makes it convenient for users to plan ahead and improve work efficiency.

[0069] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A device for real-time monitoring of mold wear status, characterized in that, The device includes a detection unit and a control unit. The detection unit is connected to the lubricating oil outlet of the stamping module (5) and is used to detect the lubricating oil to determine the wear state of the stamping die. It includes an abrasive sensor (9) and a particle analyzer (10). The abrasive sensor (9) is used to distinguish between ferromagnetic particles and non-ferromagnetic particles in the lubricating oil and obtain the particle size distribution curve of the ferromagnetic particles. The particle analyzer (10) is used to monitor the image of the lubricating oil. After noise reduction processing, it is used to identify the particle size distribution curve of particles of different materials in the lubricating oil. Then, the particle size distribution curve of the ferromagnetic particles is used to distinguish and identify particles of different materials to optimize the identification, thereby obtaining the particle size distribution curve of the die abrasive, and monitoring the wear state of the stamping die in real time. The control unit is used to control the detection unit and display the monitoring results.

2. The device for real-time monitoring of mold wear status as described in claim 1, characterized in that, The device for real-time monitoring of mold wear also includes a sound sensor (6), which is located inside the stamping module (5) and is used to determine whether the stamping mold has chipped.

3. The device for real-time monitoring of mold wear status as described in claim 1, characterized in that, The device for real-time monitoring of mold wear also includes a moisture sensor (11) and a temperature sensor (8). The moisture sensor (11) is located after the abrasive sensor (9) and particle analyzer (10), so that the lubricating oil enters the moisture sensor (11) last to monitor the moisture content of the lubricating oil. The temperature sensor (8) is located inside the stamping module (5) to monitor the temperature of the lubricating oil.

4. The device for real-time monitoring of mold wear status as described in claim 1, characterized in that, The device for real-time monitoring of mold wear also includes a first liquid level sensor (7) and a second liquid level sensor (13). The first liquid level sensor (7) is located inside the stamping module (5) and is used to monitor the level of lubricating oil in the stamping module (5). The second liquid level sensor (13) is located inside the oil tank (1) and is used to monitor the level of lubricating oil in the oil tank (1).

5. The device for real-time monitoring of mold wear status as described in claim 1, characterized in that, The control unit includes an electrical signal connection processing module (14), a pressure control module (15), and a terminal display module (16). The electrical signal connection processing module (14) is used to receive and process monitoring information. The pressure control module (15) and the terminal display module (16) are connected to the electrical signal connection processing module (14) and are used to adjust the stamping parameters and display the monitoring results, respectively.

6. The device for real-time monitoring of mold wear status as described in any one of claims 1 to 5, characterized in that, The particle analyzer (10) uses the K-means clustering algorithm to identify particles of different materials in the lubricating oil.

7. A method for real-time monitoring of mold wear condition, utilizing the device for real-time monitoring of mold wear condition as described in any one of claims 1 to 6, characterized in that, The method includes the following steps: S1. Lubricating oil is introduced into the abrasive sensor and the particle analyzer respectively. The abrasive sensor is used to distinguish between ferromagnetic particles and non-ferromagnetic particles in the lubricating oil, thereby obtaining the particle size distribution curve of the ferromagnetic particles. S2 uses the particle analyzer to identify particles of different materials in the lubricating oil to obtain two particle size distribution curves. Then, based on the changing trend of the particle size distribution curve of the ferromagnetic particles, it distinguishes them and selects the particle size distribution curve with the same changing trend as the particle size distribution curve of the mold abrasive. S3 calculates the absolute error between the particle size distribution curve of the die abrasive and the particle size distribution curve of the ferromagnetic particles. If the absolute error is greater than a threshold, return to step S2 to optimize the identification algorithm of the particle analyzer; if the absolute error is less than or equal to the threshold, monitor the wear state of the stamping die in real time according to the particle size distribution curve of the die abrasive.

8. The method for real-time monitoring of mold wear status as described in claim 7, characterized in that, In step S3, when the maximum particle size of the die abrasive is greater than 15-100 μm and the average number of die abrasive particles increases to 3-5 times the average number during the stable period, it is determined that the stamping die is severely worn and needs to be replaced.

9. The method for real-time monitoring of mold wear status as described in claim 7 or 8, characterized in that, The method for real-time monitoring of die wear status further includes step S4, which is: establishing a combined neural network model, using stamping speed, stamping load, and die clearance as inputs, and stamping die life as output, to predict the life of stamping die.

10. The method for real-time monitoring of mold wear status as described in claim 9, characterized in that, In step S4, the mean squared error loss function is used to train the combined neural network model using three parameters: stamping speed, stamping load, and die clearance. The weight matrix and bias vector of the combined neural network model are obtained through mini-batch stochastic gradient descent, thereby optimizing the combined neural network model.

Images