Production management method and system based on sheet metal punching quality trend prediction

By configuring a monitoring sliding window and linkage sensors, sheet metal punching data is collected and standardized, and trend prediction is performed. This solves the problem of lagging quality monitoring in existing technologies and achieves a predictive improvement in forward-looking quality monitoring and production management.

CN120875695BActive Publication Date: 2025-12-09XUZHOU JIATENG PRECISION MASCH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511400728.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-09-28
Publication Date
2025-12-09
Estimated Expiration
2045-09-28

AI Technical Summary

Technical Problem

Current sheet metal punching quality monitoring relies on manual inspection or real-time online monitoring, which suffers from lagging quality monitoring, lack of predictability, and inability to effectively prevent punching defects.

Method used

Configure a monitoring sliding window, collect data through image sensors and coordinate measuring machines, establish burr and size datasets, activate linkage sensors for synchronous monitoring, perform feature standardization processing, configure a trend prediction network to predict change trends, generate quality monitoring results, and manage production.

Benefits of technology

It enables proactive quality monitoring, improves the efficiency of quality monitoring, can promptly detect and prevent punching defects, and enhances the predictability of production management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120875695B_ABST
    Figure CN120875695B_ABST
Patent Text Reader

Abstract

The application discloses a production management method and system based on sheet metal punching quality trend prediction, and relates to the technical field of detection devices.The method comprises the following steps: configuring a monitoring sliding window, performing punching data collection, and establishing a punching data set.Activate the linkage sensor, perform synchronous monitoring in the monitoring sliding window, and establish a linkage data set.Based on the punching data set and the linkage data set, the standardization processing of burr characteristics, size characteristics and linkage characteristics in the monitoring sliding window is performed, and the feature change rate is calculated.A trend prediction network is configured, the change trend is predicted based on the feature change rate calculation result, the change trend prediction result is established, and the first quality monitoring result is generated.The punching data set is evaluated, and the second quality monitoring result is established.The production management of punching is performed according to the first quality monitoring result and the second quality monitoring result.Further, the technical effect of providing prospective quality monitoring and improving the quality monitoring efficiency is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of detection devices, in particular to a production management method and system based on sheet metal punching quality trend prediction. BACKGROUND

[0002] Sheet metal processing is a key process in manufacturing industry, and is widely used in automobile, aviation, construction, electrical appliances and other fields.

[0003] Among them, the punching processing of sheet metal parts is an important link in sheet metal processing, and the punching quality directly affects the performance and service life of the product. The existing punching quality monitoring mainly relies on manual inspection or real-time online monitoring. Manual inspection includes visual inspection and manual measurement, which depends on the experience and skills of the operator, and is subjective, low in efficiency and high in misjudgment rate. Real-time online monitoring collects force, vibration, sound and other signals in the punching process through sensors, or detects the shape, size and other parameters of the hole through image processing technology to realize monitoring, which cannot effectively predict and prevent punching defects, and has the technical problems of quality monitoring lag and lack of predictability. SUMMARY

[0004] The present application provides a production management method and system based on sheet metal punching quality trend prediction to solve the technical problems of quality monitoring lag and lack of predictability in the prior art, and to achieve the technical effects of providing forward-looking quality monitoring and improving quality monitoring efficiency.

[0005] In a first aspect, the present application provides a production management method based on sheet metal punching quality trend prediction, wherein the method comprises:

[0006] A monitoring sliding window is configured, and punching data collection is performed in the monitoring sliding window to establish a punching data set, including a burr data set collected by an image sensor and a size data set collected by a three-coordinate device; a linkage sensor is activated, and synchronous monitoring is performed in the monitoring sliding window based on the linkage sensor to establish a linkage data set; burr features, size features and linkage features in the monitoring sliding window are standardized based on the punching data set and the linkage data set, feature change rate calculation is performed on the standardized processing results, and feature change rate calculation results are generated; a trend prediction network is configured, change trend prediction is performed based on the feature change rate calculation results, and change trend prediction results are established; a first quality monitoring result is generated according to the change trend prediction results; data evaluation is performed on the punching data set to establish a second quality monitoring result; and production management of punching is performed according to the first quality monitoring result and the second quality monitoring result.

[0007] In a second aspect, the present application further provides a production management system based on sheet metal punching quality trend prediction, wherein the system comprises:

[0008] a data collection module configured to configure a monitoring sliding window and perform piercing data collection in the monitoring sliding window to establish a piercing data set, the piercing data set including a burr data set collected by an image sensor and a dimension data set collected by a three-coordinate device.

[0009] an auxiliary monitoring module configured to activate a linkage sensor and perform synchronous monitoring in the monitoring sliding window based on the linkage sensor to establish a linkage data set.

[0010] a feature variation calculation module configured to perform standardization processing on burr features, dimension features, and linkage features in the monitoring sliding window based on the piercing data set and the linkage data set, calculate feature variation rates based on the standardization processing results, and generate feature variation rate calculation results.

[0011] a trend prediction module configured to configure a trend prediction network and perform variation trend prediction based on the feature variation rate calculation results to establish variation trend prediction results.

[0012] a trend evaluation module configured to generate a first quality monitoring result based on the variation trend prediction results.

[0013] a monitoring evaluation module configured to perform data evaluation on the piercing data set to establish a second quality monitoring result.

[0014] a management module configured to perform production management of piercing based on the first quality monitoring result and the second quality monitoring result.

[0015] The application discloses a production management method and system based on sheet metal punching quality trend prediction, and relates to the technical field of production management. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 It is a flowchart of the production management method based on sheet metal punching quality trend prediction of the application.

[0017] Figure 2 It is a structural schematic diagram of the production management system based on sheet metal punching quality trend prediction of the application.

[0018] The reference signs are explained as follows: data acquisition module 11, auxiliary monitoring module 12, feature change calculation module 13, trend prediction module 14, trend evaluation module 15, monitoring evaluation module 16 and management module 17. DETAILED DESCRIPTION

[0019] The technical scheme provided in the embodiment of the application is used to solve the technical problems of quality monitoring lag and lack of predictability in the prior art, and the overall idea is as follows:

[0020] First, a monitoring sliding window is configured, and punching data acquisition is performed in the monitoring sliding window to establish a punching data set, wherein the punching data set includes a burr data set collected by an image sensor and a size data set collected by a three-coordinate device.

[0021] Then, a linkage sensor is activated, and synchronous monitoring is performed in the monitoring sliding window based on the linkage sensor to establish a linkage data set.

[0022] Then, based on the punching data set and the linkage data set, standardization processing of burr features, size features and linkage features in a monitoring sliding window is performed, feature change rate calculation is performed on the standardization processing results, and feature change rate calculation results are generated.

[0023] Next, a trend prediction network is configured, a change trend is predicted based on the feature change rate calculation results, and change trend prediction results are established.

[0024] Subsequently, a first quality monitoring result is generated according to the change trend prediction results.

[0025] Then, data evaluation is performed on the punching data set, and a second quality monitoring result is established.

[0026] Finally, production management of punching is performed according to the first quality monitoring result and the second quality monitoring result.

[0027] The above technical solutions will be described in detail below in combination with the drawings and specific embodiments, so that the above technical solutions can be better understood. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments for explaining the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. In addition, it should be noted that, for convenience of description, only parts related to the present application are shown in the drawings, not all. Embodiment one

[0028] Figure 1 The flowchart of the production management method of the present application based on sheet metal punching quality trend prediction is shown, wherein the method comprises:

[0029] A monitoring sliding window is configured, and punching data acquisition is performed in the monitoring sliding window to establish a punching data set, wherein the punching data set comprises a burr data set acquired by an image sensor and a size data set acquired by a three-coordinate device.

[0030] Optionally, the monitoring sliding window refers to a period of time or a series of continuous production steps, and the properties of the monitoring sliding window determine the range of data acquisition. Specifically, the monitoring sliding window is configured, including defining the size of the window and the sliding step of the window, wherein the size of the window can be the length of time or the number of production steps, and the sliding step of the window is the unit of movement of the window each time it slides, including time unit, distance unit or step unit.

[0031] Optionally, the punch data collection is performed based on a monitoring sliding window, including time series data collection (using a timer or timestamp to record data at each time point) and event-driven data collection (data collection after a certain number of punch operations or a certain production step is completed).

[0032] Optionally, key data during the punching process is collected within the detection sliding window for analysis and optimization of the punching quality. First, a high-resolution image sensor is installed at a suitable position of the punching equipment to capture images of the punching surface in real time. A three-coordinate measuring device is installed near the punching equipment to accurately measure the geometric dimensions of the punched parts. At the same time, the image sensor and the three-coordinate measuring device are aligned in time to ensure that the data of the image sensor and the three-coordinate device are synchronized, facilitating subsequent data analysis. Exemplarily, the image sensor includes an industrial camera, a laser scanner, etc.

[0033] Further, the collected burr data set and the size data set are sorted and stored to form a structured punch data set. Specifically, the punch data set includes time-synchronized and one-to-one burr data and size data, and the punch data set is stored based on a preset data format structure.

[0034] In some embodiments, the establishing a punch data set further includes:

[0035] Obtaining an image acquisition result of the image sensor; configuring a preset punch contour based on basic data of the punching; performing pre-contour recognition of the image acquisition result, and performing positioning-based similarity matching based on the pre-contour recognition result and the preset punch contour to generate a similarity matching result; and establishing a burr data set based on the similarity matching result.

[0036] Optionally, the image sensor is used to take an image of the punched part, and the image data is processed through photoelectric conversion and ISP to obtain the image acquisition result. The image acquisition result is a real-time processing image of the workpiece on the target production line, and can be a white light image, a grayscale image, an infrared image, etc. according to the type of the image sensor.

[0037] Optionally, the preset punch contour is configured based on the basic data of the punching. First, the basic data of the punched part is collected and sorted, wherein the basic data includes design drawings, size specifications, and process parameters, etc. Then, the preset punch contour is generated based on the basic data as a reference contour in an ideal state. This includes using computer-aided design (CAD) software to generate an accurate preset punch contour or using parameterized modeling technology to convert the design drawings and process parameters into an adjustable preset contour model.

[0038] Optionally, a pre-contour recognition of the image acquisition result is performed, and image processing algorithms are used to pre-process the image acquisition result, including denoising, edge detection and contour extraction, and then the actual contour of the punching part is recognized, wherein the pre-contour recognition is a lower-precision rough recognition, which is used to determine the approximate position of the punching image in the image acquisition result, and has a higher recognition efficiency.

[0039] Further, based on the pre-contour recognition result and the preset punching contour, a positioning-based similarity matching is performed to generate a similarity matching result. First, an image registration algorithm (such as an ICP algorithm) is used to adjust the position and attitude of the recognition result and the preset contour, so as to realize the above registration and ensure that the two are compared in the same coordinate system, including image scaling and rotation. Then,

[0040] A similarity matching algorithm is used to calculate the similarity between the two, and a similarity matching result is generated. Specifically, algorithms such as Hausdorff distance and shape context are used to match the similarity of the contours.

[0041] In some implementations, the establishing of the burr data set based on the similarity matching result further includes:

[0042] Based on the similarity matching result, a perspective comparison of the preset punching contour and the pre-contour recognition result is performed to establish a region of interest; a convolution kernel-based traversal recognition is performed on the region of interest to generate a traversal recognition result; and a burr data set is established based on the traversal recognition result.

[0043] Optionally, a perspective transformation algorithm (such as Homography) is used to perform perspective transformation on the preset punching contour and the pre-contour recognition result, so as to ensure that the two are compared on the same perspective plane. Then, methods such as structural similarity index, mean square error, histogram and shape matching algorithm are used to determine the regions where the actual contour and the preset contour deviate, which are the regions of interest. In other words, the parts where the preset punching contour and the pre-contour recognition result differ are the positions where burrs or other quality defects may exist, and the corresponding regions are set as the regions of interest.

[0044] Optionally, a convolution operation is performed on the images in the region of interest to extract burr features. First, a convolution kernel for burr recognition is designed, which can be customized according to the features of the burr, such as edge detection convolution kernels (such as Sobel operator, Scharr operator, Laplacian operator, etc.). Then, all pixels in the region of interest are traversed, and the convolution kernel is applied for feature extraction to generate a traversal recognition result. Preferably, a convolutional neural network (CNN) is used to extract and recognize features in the region of interest.

[0045] Optionally, a burr data set is established by the traversal identification result, and the burr data set contains image features of a possible burr image part in the multiple time-series images in the monitoring sliding window, such as edge features, corner features, and corresponding chrominance or grayscale features.

[0046] The method steps described above can more accurately identify the difference position through perspective comparison and traversal identification based on the convolution kernel, so as to determine the attention region. Fine identification and analysis are performed on the attention region, which helps to improve the monitoring accuracy of the punching quality problem.

[0047] The linkage sensor is activated, and synchronous monitoring is performed in the monitoring sliding window based on the linkage sensor, and a linkage data set is established.

[0048] The linkage sensor includes multiple different types of sensors that work cooperatively to simultaneously monitor multiple characteristic parameters such as vibration, deformation, and damage. For example, the linkage sensor includes a vibration sensor, a distance sensor, a temperature sensor, a strain gauge, and the like.

[0049] Optionally, the linkage sensor is installed at key positions of the punching device or the workpiece based on the collection characteristics of different sensors to ensure that relevant data can be accurately captured. For example, the strain gauge is installed on the surface of the workpiece, and the vibration sensor is installed on or near the punching device.

[0050] Optionally, the linkage sensor realizes communication control and data transmission through a wireless sensor network.

[0051] Optionally, in the monitoring sliding window, the linkage sensor works synchronously to perform real-time monitoring and data collection on multiple characteristic parameters in the punching process, and data of characteristic parameters such as vibration, deformation, and damage are collected. Then, the collected multiple characteristic parameter data are synchronously processed to ensure that the data time stamps are consistent.

[0052] Optionally, the synchronous monitoring includes a hardware-based synchronous method and a software-based synchronous method, such as a synchronous trigger or a time synchronization protocol (such as NTP), to ensure that all sensors perform data collection at the same time point or that data of different sensors have a unified time reference.

[0053] Further, the collected vibration, deformation and damage characteristic parameter data are sorted and cleaned to remove noise and outliers, and the sorted data is classified according to the characteristic parameters to form a plurality of characteristic sets. Then, based on the timestamps or time sequence markers in the plurality of characteristic sets, a data mapping relationship of the plurality of characteristic sets is established, and is integrated and stored in a database to form a linkage data set. The linkage data set is auxiliary data in the punching quality monitoring, and contains a plurality of indirect index parameters to reflect the working state of the punching equipment or target production line and the development trend of the working state.

[0054] Through the cooperative work of the plurality of linkage sensors, a plurality of characteristic parameters such as vibration, deformation and damage in the punching process can be monitored at the same time, comprehensive monitoring data is provided, and more abundant information is provided for subsequent analysis to identify potential problems in the punching process and the development trend of the equipment operating state.

[0055] Based on the punching data set and the linkage data set, standardization processing of burr characteristics, size characteristics and linkage characteristics in a sliding window is performed, and a feature change rate calculation is performed on the standardization processing result to generate a feature change rate calculation result.

[0056] Optionally, the feature standardization processing includes time sequence standardization and scale standardization to improve the performance and stability of the model. Specifically, the scale standardization processing method includes z-score standardization, min-max standardization or Robust standardization. The Z-score standardization is suitable for data close to normal distribution and is sensitive to outliers. The Min-Max standardization is suitable for scaling data to a fixed range and is sensitive to outliers. Preferably, the Robust standardization is selected for the above-mentioned feature standardization processing. The Robust standardization is suitable for data with outliers and is not sensitive to outliers.

[0057] Optionally, the time sequence standardization is performed based on time sequence registration and interpolation method to align the data on different time axes to a common time axis. Specifically, first, the time sequence registration is performed to align different time sequences to a unified time axis for further analysis. The method includes dynamic time warping (DTW) and linear interpolation. After time sequence registration, some missing values or irregular sampling points may appear. Then, the interpolation method is used to complete the missing values and smooth the data.

[0058] Further, the feature change rate of the normalized processing result in the monitoring sliding window is calculated to evaluate the change of each feature over time. Optionally, the window change proportion of each feature is calculated corresponding to the feature change rate under the unified window, and the feature change proportion per unit time of each feature is calculated for the feature change rate under the non-unified window (such as the adaptive window).

[0059] For example, for the feature change rate under the unified window, the following formula is used for calculation:

[0060] ;

[0061] The above method steps eliminate the scale difference between different features through normalization processing, thereby improving the consistency and comparability of the data. By calculating the feature change rate, the feature change in the punching process can be monitored in real time, and abnormalities can be discovered and fed back in a timely manner.

[0062] The trend prediction network is configured to perform change trend prediction based on the feature change rate calculation result, and to establish a change trend prediction result.

[0063] In some embodiments, the trend prediction network further includes:

[0064] The trend prediction network is as follows:

[0065] ;

[0066] ;

[0067] wherein, is the change trend prediction result after the step, is the comprehensive feature value at the moment, is the weighted comprehensive change rate of each feature, characterizes the normalized burr size at the moment, characterizes the normalized size at the moment, characterizes the feature value of the normalized linkage feature at the moment, characterizes the normalized burr size at the moment, characterizes the normalized size at the moment, characterizes the feature value of the normalized linkage feature at the moment, , , , , , , , , , , , , The standard deviations of burrs, dimensions, and burr size are respectively characterized. standard deviation Characteristic change rate of burr features after standardization. Characteristic change rate of dimensional features after standardization Characterization of linkage features after standardization rate of change, , , These are the weighting coefficients for burr features, size features, and linkage features, respectively.

[0068] Optionally, a trend prediction network can be configured, which includes a weighted composite rate of change, expressed as:

[0069] ;

[0070] in, 、 、 These are, respectively, the feature change rate of burr features after standardization, the feature change rate of size features after standardization, and the linkage features after standardization, as described above. rate of change, This represents the weighted average of the rate of change of multiple standardized feature indicators in the linkage feature.

[0071] Optional, weighting coefficients for burr features Weighting coefficients for size features Weight coefficients of linkage features The degree of influence of the corresponding characteristics on the processing quality of the target production line is determined, for example, by means of correlation analysis or principal component analysis.

[0072] The trend prediction network described above, through standardization and weighted calculation, can effectively integrate the changes in different characteristics and provide accurate trend prediction results.

[0073] In some embodiments, the method further includes:

[0074] Historical punching data is acquired, and features are extracted from this data to establish a steady-state prediction network. The previous monitoring sliding window is invoked, and a steady-state evaluation of the punching process is performed based on the invocation result and the steady-state prediction network, generating steady-state prediction coefficients. The trend prediction is then configured using these steady-state prediction coefficients and the size of the monitoring sliding window. value.

[0075] Optionally, the previously recorded piercing process data is extracted from the database or storage system to obtain historical piercing data, which contains detailed information of multiple piercing operations. Then, key characteristic parameters are extracted from the historical piercing data, which can reflect the stability and performance of the piercing process. For example, the characteristic parameters include piercing force characteristics, piercing speed characteristics, piercing depth characteristics, burr characteristics, size characteristics, vibration characteristics, strain characteristics, etc.

[0076] Optionally, using the extracted characteristic data, a piercing stability prediction network is constructed based on a neural network or other machine learning model to predict the stability level of the piercing process. Specifically, the model structure is first determined, including multilayer perceptron (MLP), convolutional neural network (CNN), or long short-term memory network (LSTM), etc.; then, based on the historical product quality, the historical piercing data is labeled for stability, and the historical piercing data with stable product quality is labeled with a higher stability index or index level; next, using the labeled historical piercing data, the data is divided into a training set and a validation set for model training and validation; finally, the performance of the model is evaluated by evaluation indicators (such as mean square error, accuracy, etc.), and hyperparameter tuning is performed.

[0077] Optionally, based on the current monitoring sliding window, the monitoring data of the previous time period is called, and the extracted monitoring data and the trained stability prediction network model are used to evaluate the stability of the current piercing process, and output the stability prediction coefficient, which quantifies the recent processing stability of the target production line or target piercing equipment.

[0078] Further, based on the stability prediction coefficient and the size of the monitoring sliding window, the parameter k value for change trend prediction is configured. This includes using mathematical modeling methods to construct a relationship model between the stability prediction coefficient and the k value. The k value can be considered as a correction coefficient for the monitoring sliding window, and the stability prediction coefficient is negatively correlated with the k value. The lower the recent processing stability of the target production line or target piercing equipment, the smaller the stability prediction coefficient, and the shorter the time interval required for change trend prediction to faithfully reflect the changes in the target production line or target piercing equipment. Accordingly, when the stability prediction coefficient is larger, it indicates that the processing stability is higher, and the prediction frequency can be reduced, i.e., the k value should be larger.

[0079] By dynamically adjusting the frequency of change trend prediction according to the changes in processing stability, it is helpful to more accurately reflect the changes in the target production line or target piercing equipment.

[0080] A first quality monitoring result is generated according to the change trend prediction result.

[0081] Optionally, the first quality monitoring result is used to reflect the processing stability of the target production line or the target piercing device. For example, the first quality monitoring result includes a trend fluctuation rate, a trend fluctuation frequency, a trend fluctuation amplitude, or the like, which are calculated based on a change trend prediction result.

[0082] The piercing data set is subjected to data evaluation to establish a second quality monitoring result.

[0083] Optionally, the data in the piercing data set is first evaluated based on data integrity, continuity, and the like to determine the quality and effectiveness of the data, and a data quality evaluation result is generated. Then, whether the quality of the current piercing process meets the requirements is determined based on the identified burr features, and a piercing quality evaluation result is determined.

[0084] Specifically, to determine the piercing quality evaluation result, first, a quality standard is defined to determine the evaluation standard of the piercing quality, such as the threshold values of burr height, length, and quantity. Then, the quality of the current piercing process is evaluated based on the identified burr features and the preset quality standard, including determining whether the current piercing quality meets the requirements. If not, the quality problem is recorded. If yes, the quality level of the current piercing quality is defined. Preferably, the quality evaluation is performed based on a quality evaluation model, such as a decision tree or a support vector machine (SVM).

[0085] Optionally, the data quality evaluation result and the piercing quality evaluation result are integrated to establish the second quality monitoring result. By evaluating the data quality and the product quality at the same time, the second quality monitoring result comprehensively integrates the data confidence and the processing quality, which helps to provide a more comprehensive evaluation.

[0086] The production management of piercing is performed based on the first quality monitoring result and the second quality monitoring result.

[0087] In some embodiments, the production management of piercing based on the first quality monitoring result and the second quality monitoring result includes:

[0088] It is determined whether the first quality monitoring result meets a preset quality threshold. If the first quality monitoring result does not meet the preset quality threshold, a prediction warning information is generated. The second quality monitoring result is subjected to a current situation deviation evaluation of piercing to generate additional warning information. The production management of piercing is performed based on the prediction warning information and the additional warning information.

[0089] Optionally, if the first quality monitoring result meets the preset quality threshold, it is considered that the piercing process is stable. If the first quality monitoring result does not meet the preset quality threshold, it indicates that the processing stability of the target production line or the target piercing device is low, the processing quality fluctuates greatly, or the processing quality at a future time will not meet the quality control standard, and a prediction warning information is generated to give a forward-looking warning on the processing quality.

[0090] Optionally, the present situation deviation evaluation of the second quality monitoring result is performed, and if the evaluation result shows that the present situation deviation is greater than a preset quality control standard, additional early warning information is generated to perform production management on the target production line or the target piercing device.

[0091] Optionally, the production management of piercing according to the prediction early warning information and the additional early warning information includes adjusting production parameters, i.e., adjusting the parameters (such as pressure, speed, etc.) of the piercing machine according to the early warning information to improve the quality; if the early warning information indicates that there is a problem with the device, maintenance and maintenance of the device are arranged; if the problem may be related to the operator, relevant training is arranged to improve the stability of the device and the quality of the product.

[0092] Further, the method further comprises:

[0093] The prediction early warning information and the additional early warning information are continuously recorded, and a periodical compensation is established; and the control optimization of the production line is performed through the periodical compensation.

[0094] Optionally, the periodical compensation is used for periodically performing control optimization on the target production line or the target piercing device, wherein the periodical compensation includes a time periodical compensation and a work amount periodical compensation. The time periodical compensation is a periodical compensation based on a natural day or a working day, and the work amount periodical compensation is a periodical compensation determined based on a working time length or a working cycle number.

[0095] Specifically, first, the early warning information (including the prediction early warning information and the additional early warning information) is stored in a database, and a time series analysis, a spectrum analysis, etc. are used to periodically (such as every week, every month) analyze the stored early warning information to identify the periodic characteristics of the early warning information. Then, based on the periodic characteristics of the obtained early warning information, corresponding periodical compensation parameters such as compensation nodes and compensation amounts are defined, and then at the corresponding compensation nodes, the pressure and speed of the piercing machine are automatically adjusted by the preset compensation amount to prevent the problem of burr characteristics exceeding the standard, thereby improving the control optimization automation level of the production line.

[0096] For example, if during the piercing production process, through the recording and analysis of the early warning information, it is found that the piercing quality fluctuates due to temperature changes in the first two hours after the device starts every morning. Then, a time periodical compensation model can be established. Before the device starts every morning, the temperature and pressure parameters of the piercing machine are adjusted in advance to prevent the problem of quality fluctuation. If the piercing head of the piercing machine is worn out after every 1000 products are produced, the work amount periodical compensation model can be established through the recording and analysis of the early warning information. Before every 1000 products are produced, the piercing head is checked and replaced in advance to prevent the quality problem.

[0097] In summary, the production management method based on sheet metal punching quality trend prediction provided by this invention has the following technical effects:

[0098] By configuring a monitoring sliding window and performing punching data acquisition within this window, a punching dataset is established. This dataset includes a burr dataset acquired via an image sensor and a dimensional dataset acquired via a coordinate measuring machine. A linkage sensor is activated, and synchronous monitoring is performed within the monitoring sliding window based on the linkage sensor, establishing a linkage dataset. Based on the punching dataset and the linkage dataset, burr features, dimensional features, and linkage features within the monitoring sliding window are standardized, and the feature change rate is calculated from the standardized results to generate a feature change rate calculation result. A trend prediction network is configured to predict the change trend based on the feature change rate calculation result, establishing a trend prediction result. A first quality monitoring result is generated based on the trend prediction result. The punching dataset is evaluated to establish a second quality monitoring result. Production management for punching is then conducted based on the first and second quality monitoring results. This achieves the technical effect of providing proactive quality monitoring and improving the efficiency of quality monitoring. Example 2

[0099] Figure 2 This is a schematic diagram of the production management system based on sheet metal punching quality trend prediction according to the present invention. For example, Figure 1 The flowchart of the production management method based on sheet metal punching quality trend prediction of the present invention can be illustrated as follows: Figure 2 The structure shown is implemented.

[0100] Based on the same concept as the production management method based on sheet metal punching quality trend prediction in the above embodiments, the present invention also provides a production management system based on sheet metal punching quality trend prediction, which includes:

[0101] The data acquisition module 11 is used to configure the monitoring sliding window and perform punching data acquisition in the monitoring sliding window to establish a punching dataset. The punching dataset includes a burr dataset acquired by an image sensor and a size dataset acquired by a coordinate measuring machine.

[0102] The auxiliary monitoring module 12 is used to activate the linkage sensor, perform synchronous monitoring in the monitoring sliding window based on the linkage sensor, and establish a linkage dataset.

[0103] The feature variation calculation module 13 is used to perform standardization processing on the burr features, size features and linkage features within the monitoring sliding window based on the punching dataset and the linkage dataset, calculate the feature change rate on the standardization processing results, and generate the feature change rate calculation results.

[0104] The trend prediction module 14 is configured to configure a trend prediction network, perform change trend prediction based on the feature change rate calculation result, and establish a change trend prediction result.

[0105] The trend evaluation module 15 is configured to generate a first quality monitoring result according to the change trend prediction result.

[0106] The monitoring evaluation module 16 is configured to perform data evaluation on the punching data set to establish a second quality monitoring result.

[0107] The management module 17 is configured to perform production management of punching according to the first quality monitoring result and the second quality monitoring result.

[0108] The data acquisition module 11 includes:

[0109] An image acquisition unit configured to acquire an image acquisition result of an image sensor.

[0110] A contour definition unit configured to configure a preset punching contour based on basic data of punching.

[0111] A similarity matching unit configured to perform pre-contour identification of the image acquisition result, perform positioning-based similarity matching according to the pre-contour identification result and the preset punching contour, and generate a similarity matching result.

[0112] A burr identification unit configured to establish a burr data set based on the similarity matching result.

[0113] In some implementations, the burr identification unit in the data acquisition module 11 includes:

[0114] A perspective comparison unit configured to perform perspective comparison of the preset punching contour and the pre-contour identification result based on the similarity matching result, and establish a region of interest.

[0115] A convolution identification unit configured to perform traversal identification based on a convolution kernel on the region of interest, and generate a traversal identification result.

[0116] A burr integration unit configured to establish a burr data set through the traversal identification result.

[0117] In some embodiments, the trend prediction module 14 includes:

[0118] The trend prediction network is as follows:

[0119] .

[0120] .

[0121] Wherein, is the change trend prediction result after the step, is a comprehensive feature value of the moment, is a weighted comprehensive change rate of each feature, characterizes a normalized burr size of the moment, characterizes a normalized size of the moment, characterizes a normalized linkage feature of the moment , characterizes a normalized burr size of the moment, characterizes a normalized size of the moment, characterizes a normalized linkage feature of the moment , , , respectively characterizes the standard deviation of the burr, the standard deviation of the size, and the standard deviation of , characterizes a feature change rate of the burr feature after normalization, characterizes a feature change rate of the size feature after normalization, characterizes a change rate of the linkage feature after normalization, , , are weight coefficients of the burr feature, the size feature, and the linkage feature respectively.

[0122] In some embodiments, the trend prediction module 14 includes:

[0123] a steady state prediction unit configured to obtain historical piercing data, and after feature extraction on the historical piercing data, establish a piercing steady state prediction network.

[0124] a steady state evaluation unit configured to call a previous monitoring sliding window, perform steady state evaluation of piercing according to a calling result and the steady state prediction network, and generate a steady state prediction coefficient.

[0125] a parameter configuration unit configured to configure a value of the change trend prediction through the steady state prediction coefficient and a size of the monitoring sliding window.

[0126] In some embodiments, the management module includes:

[0127] a quality discrimination unit configured to determine whether the first quality monitoring result meets a preset quality threshold.

[0128] The early warning generation unit is configured to generate prediction early warning information if the first quality monitoring result fails to meet the preset quality threshold.

[0129] The additional early warning unit is configured to perform a present situation deviation evaluation on the second quality monitoring result to generate additional early warning information.

[0130] The joint early warning unit is configured to perform production management of punching according to the prediction early warning information and the additional early warning information.

[0131] Further, the system further comprises:

[0132] The compensation configuration unit is configured to continuously record the prediction early warning information and the additional early warning information, and establish a periodical compensation.

[0133] The periodical optimization unit is configured to perform control optimization of a production line through the periodical compensation.

[0134] It should be understood that the embodiments mentioned in the specification focus on the differences from other embodiments, and the specific embodiments in the first embodiment are also applicable to the production management system based on the sheet metal punching quality trend prediction in the second embodiment. For the sake of brevity of the specification, no further expansion is made here.

[0135] It should be understood that the disclosed embodiments and the above description enable those skilled in the art to implement the present application. Meanwhile, the present application is not limited to the aforementioned part of the embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A production management method based on a punching quality trend prediction of sheet metal, characterized by, The method comprises: configuring a monitoring sliding window, and performing piercing data collection in the monitoring sliding window to establish a piercing data set, the piercing data set comprising a burr data set collected by an image sensor and a size data set collected by a three-coordinate device; activating a linkage sensor, performing synchronous monitoring in the monitoring sliding window based on the linkage sensor to establish a linkage data set, the linkage data set being auxiliary data in piercing quality monitoring and containing various indirect index parameters; performing standardization processing of burr features, size features and linkage features in the monitoring sliding window based on the piercing data set and the linkage data set, calculating feature change rates of the standardization processing results, and generating feature change rate calculation results; configuring a trend prediction network, performing change trend prediction based on the feature change rate calculation results, and establishing change trend prediction results; generating a first quality monitoring result according to the change trend prediction results; performing data evaluation on the piercing data set to establish a second quality monitoring result; performing piercing production management according to the first quality monitoring result and the second quality monitoring result; The piercing production management according to the first quality monitoring result and the second quality monitoring result further comprises: determining whether the first quality monitoring result meets a preset quality threshold; if the first quality monitoring result cannot meet the preset quality threshold, generating a prediction warning information; performing piercing current situation deviation evaluation on the second quality monitoring result to generate additional warning information; performing piercing production management according to the prediction warning information and the additional warning information; continuously recording the prediction warning information and the additional warning information, analyzing the warning information, identifying periodic characteristics, obtaining periodic characteristics of the warning information, defining corresponding periodic compensation parameters, and performing control optimization of a production line through the periodic compensation parameters.

2. The production management method based on the punching quality trend prediction of sheet metal according to claim 1, wherein, The configuration of the trend prediction network further comprises: The trend prediction network is as follows: ; ; wherein, is a change trend prediction result after the step, is a comprehensive feature value of the moment, is a weighted comprehensive change rate of each feature, characterizes a burr size after standardization of the moment, characterizes a size after standardization of the moment, characterizes a linkage feature value after standardization of the moment, characterizes a burr size after standardization of the moment, characterizes a size after standardization of the moment, characterizes a linkage feature value after standardization of the moment, characterizes a burr standard deviation, , , a size standard deviation, and a linkage standard deviation, respectively, characterizes a feature change rate of a burr feature after standardization, characterizes a feature change rate of a size feature after standardization, characterizes a change rate of a linkage feature after standardization, , , are weight coefficients of the burr feature, the size feature, and the linkage feature, respectively.

3. The production management method based on the punching quality trend prediction of sheet metal according to claim 2, wherein, The method further comprises: obtaining historical piercing data, performing feature extraction on the historical piercing data, and establishing a piercing steady-state prediction network; calling a previous monitoring sliding window, performing piercing steady-state evaluation according to a calling result and the steady-state prediction network, and generating a steady-state prediction coefficient; by the steady-state prediction coefficients and monitoring the size of the sliding window for configuration changes trends prediction values.

4. The production management method based on the punching quality trend prediction of sheet metal according to claim 1, wherein, The establishment of the piercing data set further comprises: obtaining an image collection result of the image sensor; configuring a preset piercing contour based on piercing basic data; performing pre-contour identification of the image collection result, performing positioning-based similarity matching based on a pre-contour identification result and the preset piercing contour, and generating a similarity matching result; establishing a burr data set based on the similarity matching result.

5. The production management method based on the punching quality trend prediction of sheet metal according to claim 4, wherein, The establishment of the burr data set based on the similarity matching result further comprises: performing perspective comparison of the preset piercing contour and the pre-contour identification result based on the similarity matching result, and establishing a region of interest; performing traversal identification based on a convolution kernel on the region of interest, and generating a traversal identification result; establishing a burr data set through the traversal identification result.

6. A production management system based on a punching quality trend prediction of sheet metal, characterized by, The system is used to implement the production management method based on sheet metal piercing quality trend prediction according to any one of claims 1-5, and the system comprises: The data acquisition module is configured to monitor a sliding window, perform piercing data acquisition in the sliding window, and establish a piercing data set, wherein the piercing data set comprises a burr data set acquired by an image sensor and a size data set acquired by a three-coordinate device; The auxiliary monitoring module is configured to activate a linkage sensor, perform synchronous monitoring in the sliding window based on the linkage sensor, and establish a linkage data set; The feature variation calculation module is configured to perform standardization processing on burr features, size features, and linkage features in the sliding window based on the piercing data set and the linkage data set, calculate a feature variation rate based on a result of the standardization processing, and generate a feature variation rate calculation result; The trend prediction module is configured to configure a trend prediction network, perform variation trend prediction based on the feature variation rate calculation result, and establish a variation trend prediction result; The trend evaluation module is configured to generate a first quality monitoring result according to the variation trend prediction result; The monitoring evaluation module is configured to perform data evaluation on the piercing data set, and establish a second quality monitoring result; The management module is configured to perform production management of piercing according to the first quality monitoring result and the second quality monitoring result.

Citation Information

Patent Citations

  • Pharmaceutical production line state monitoring and alarming method and system based on Internet of Things

    CN118114023A

  • Bus duct trend analysis early warning method and system and storage medium

    CN119492945A