A system for predicting wrinkles in an automotive bearing stamping

Abstract

The application discloses a kind of automobile bearing stamping part wrinkle prediction control systems, it is related to model prediction technical field, including: data acquisition unit, by fluctuation acquisition matrix and variable acquisition matrix, obtain the fluctuation change data of plate in stamping process by fluctuation acquisition matrix, and obtain the variable data of plate in stamping process by variable acquisition matrix;Stamping model is established based on historical stamping data, obtains fluctuation change data and carries out differential analysis, determines the abnormal data section of historical stamping data, determines plate abnormal sound wave and vibration wave characteristics, in combination with variable data, the wrinkle of stamping part is predicted by wrinkle prediction algorithm, generates prediction result.This application is in the process of plate stamping, by monitoring the vibration wave and sound wave change of plate, determine the state change of plate in mould, to achieve the purpose of wrinkle prediction, further improve the quality of plate stamping processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of model prediction, in particular to a prediction control system for wrinkles of automobile bearing stamping parts. BACKGROUND

[0002] Stamping bearings are widely used in engines, gearboxes and steering systems to support and guide rotating parts, ensuring smooth operation of the transmission system at high speed and high load. Lightweight and high-precision design helps to reduce the weight of the vehicle and improve fuel economy, while meeting the requirements of high speed and wear resistance.

[0003] During the stamping process, the physical changes of the sheet metal include elastic and plastic deformation. When external stamping force is applied to the sheet metal, the material first undergoes elastic deformation, which is characterized by a linear relationship between stress and strain. Beyond the yield point, the sheet metal enters the plastic deformation stage and undergoes permanent deformation. During the plastic deformation process, the internal lattice structure of the material adjusts, resulting in irreversible changes, and this process is often accompanied by local work hardening.

[0004] Publication No. CN113221402B discloses a stamping part springback prediction and monitoring method, system and storage medium. The stamping part springback prediction and monitoring method includes obtaining first data corresponding to the target die surface and second data corresponding to the single stamping surface. The target compensation value is determined according to the first data and the second data. The springback amount prediction model and / or the monitoring plate state are constructed based on the target compensation value. The second data of the single stamping surface in the actual stamping process of the stamping part is obtained. The target compensation value is determined by combining the first data corresponding to the target die surface with the second data. The processing method of the stamping part is guided according to the springback amount prediction model, so that the processing method of the stamping part is more in line with the actual processing requirements, and / or the monitoring of the plate state is also conducive to ensuring the processing quality of the stamping part.

[0005] However, due to the complex shape and typical thin-walled structure of the bearing stamping part, the stamping process is prone to problems such as local overpressure and uneven stress, which can cause sheet metal wrinkling. Therefore, it is difficult to solve the wrinkling problem by only predicting springback.

[0006] When the sheet metal is subjected to local overpressure or stamping force during the stamping process, the sheet metal locally bears a large compressive stress, which exceeds the yield limit of the material, and wrinkles are easily generated. If the friction between the sheet metal and the die is not effectively controlled, the local deformation will be limited, stress concentration will occur, and the risk of wrinkling will increase.

[0007] Therefore, the present application proposes a prediction control system for real-time monitoring of the stress distribution and deformation of the sheet metal during the stamping process. SUMMARY

[0008] One of the purposes of the present application is to provide a wrinkling prediction control system for automobile bearing stamping parts, which can determine the state change of the plate in the mold during the plate stamping process by monitoring the vibration wave and sound wave change of the plate, so as to achieve the purpose of wrinkling prediction and further improve the quality of plate stamping.

[0009] To achieve the above purpose, the present application is realized by the following technical scheme: a wrinkling prediction control system for automobile bearing stamping parts, comprising:

[0010] The data acquisition unit is composed of a fluctuation acquisition matrix and a variable acquisition matrix, which acquires the control parameters of the control system for the stamping equipment, acquires the fluctuation change data of the plate during the stamping process through the fluctuation acquisition matrix, and acquires the variable data of the plate during the stamping process through the variable acquisition matrix;

[0011] The preprocessing unit splits the fluctuation change data and the variable data into data segments after noise suppression by noise filtering;

[0012] The stamping model is established based on historical stamping data, differentially analyzes the fluctuation change data, determines the abnormal data segment of the historical stamping data, determines the abnormal sound wave and vibration wave characteristics of the plate, combines the variable data, and predicts the wrinkling of the stamping part through a wrinkling prediction algorithm to generate a prediction result;

[0013] The feedback control unit receives the prediction result, feeds back the abnormal state of the plate, and adjusts the stamping parameters according to the feedback result.

[0014] In one or more embodiments of the present application, the fluctuation acquisition matrix is composed of a plurality of acoustic emission sensors and vibration sensors, which are distributed at different positions of the mold to capture the sound wave and vibration wave signals of the plate during the stamping process with the middle part of the plate as the reference point;

[0015] A three-dimensional coordinate space is established and the coordinate point of each sensor in the coordinate space is determined, the data collected by the sensor is mapped to the coordinate point, a three-dimensional distribution map of the sound wave and the vibration wave is constructed, and the wrinkling position of the plate is determined by comparing and analyzing the signal intensity, frequency change and time sequence change at different positions.

[0016] In one or more embodiments of the present application, the variable acquisition matrix includes a plurality of pressure sensors and displacement sensors, the pressure sensors are located at the high stress or weak area position in the mold, and the displacement sensors are arranged outside the mold to monitor the displacement distance of the mold during the stamping process;

[0017] The pressure sensor and the displacement sensor both have a unified clock synchronization function, a pressure distribution surface graph is constructed based on the pressure data with the synchronous time as the reference, and is fused with the displacement data to generate a dynamic stress displacement model, and the plate state under pressure is mapped.

[0018] In one or more embodiments of the present application, the variable collection matrix includes a plurality of temperature sensors arranged at key hot areas of the mold and a plurality of strain sensors attached to the surface of the plate to monitor temperature changes and strain distribution in real time, and the coordinate points of the temperature sensors and the strain sensors are mapped in a three-dimensional coordinate space to construct a temperature and strain distribution map.

[0019] In one or more embodiments of the present application, the vibration sensor acquires vibration change data, and the acoustic emission sensor acquires acoustic wave change data, the vibration change data and the acoustic wave change data are subjected to noise suppression by a preprocessing unit, the vibration change data and the acoustic wave change data are split into data segments based on a complete vibration cycle in the vibration change data, and the periodicity of the signals is retained.

[0020] In one or more embodiments of the present application, the determination of the abnormal acoustic wave and vibration wave characteristics of the plate is as follows:

[0021] Each vibration change data segment and acoustic wave change data segment is subjected to feature extraction through time series and frequency domain analysis;

[0022] A feature distribution model in a normal stamping process is established based on historical stamping data to form a feature baseline, which describes the feature range, distribution and evolution trend in a normal stamping state;

[0023] The features of the vibration and acoustic wave change data segments are compared with the feature distribution model to determine whether the features of the vibration and acoustic wave change data segments are within the normal range;

[0024] Whether the data segments are abnormal is determined by setting a threshold or an anomaly detection algorithm, and the abnormal data segments are labeled.

[0025] In one or more embodiments of the present application, the abnormal data segments are combined with the variable data to establish a wrinkle prediction algorithm:

[0026] The timestamp of the abnormal data segment is determined, the variable data range is determined based on the timestamp, each group of corresponding abnormal data segments and variable data forms a multi-dimensional feature vector, and a state label is generated;

[0027] The signal features in the abnormal data segments and the variable features in the variable data are extracted;

[0028] The signal features and the variable features are constructed into a complete feature vector X;

[0029] A learning model is established according to the feature vector X and the label using historical stamping data, and the learning model is further optimized through cross-validation and feature importance analysis to screen sensitive features.

[0030] In one or more embodiments of the present application, in the stamping process, the collected vibration, sound wave change data and corresponding process parameter data are preprocessed and feature extracted through the fluctuation acquisition matrix and the variable acquisition matrix, and a new feature vector is generated, which is input into the prediction model to determine the prediction result and risk probability.

[0031] In one or more embodiments of the present application, the determination of the wrinkling position of the plate based on the three-dimensional coordinate space is as follows:

[0032] Each acoustic emission sensor and vibration sensor collects acoustic wave and vibration signals in the stamping process, which are associated with the corresponding coordinate points;

[0033] The data of each discrete coordinate point is expanded to the entire plate surface area by a spatial interpolation method to form a three-dimensional distribution map of sound and vibration signals;

[0034] For the abnormal features of each point in the entire area, the coordinate points with similar abnormal features are grouped through cluster analysis to determine whether there is a locally aggregated abnormal area;

[0035] In combination with the spatial distribution map and the clustering result, the position of the abnormal area is identified in the three-dimensional coordinate system.

[0036] In one or more embodiments of the present application, for each sensor coordinate point, the preprocessed fluctuation intensity, timing information and frequency characteristics are integrated into a multi-dimensional feature vector F i where the multi-dimensional feature vector F i represents the feature vector of sensor i.

[0037] Through the above technical solution, the present application has the following beneficial effects:

[0038] 1. In the process of stamping the plate, the present application determines the state change of the plate in the mold by monitoring the vibration wave and sound wave change of the plate, thereby achieving the purpose of wrinkling prediction and further improving the quality of the plate stamping process.

[0039] 2. The sound waves and vibration waves generated by the normal state of plate stamping and the wrinkling state after stamping are inconsistent. The obtained fluctuation change data is segmented to form data segments, each data segment is analyzed, and the data segment corresponding to the time stamp can facilitate the combination of variable data to verify the wrinkling state.

[0040] 3. During the stamping process, the different locations of deformation and stress distribution of the sheet metal lead to different vibration locations and sound wave points, thus forming a unique acoustic vibration characteristic spectrum. By analyzing the location of abnormal points in coordinate space, the location of wrinkling can be determined.

[0041] 4. By comprehensively analyzing the wave intensity, temporal changes and frequency characteristics collected by each sensor in three-dimensional coordinate space, the sound wave and vibration distribution map of the sheet metal during the stamping process can be accurately constructed. By comparing the data at different locations, when an abnormal cluster appears in a local area, it can be preliminarily determined that there is a risk of wrinkling in that area.

[0042] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the predictive control system of the present invention. Detailed Implementation

[0044] The following describes several embodiments of the present invention with reference to the accompanying drawings. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the invention. That is, in some embodiments of the invention, these practical details are not essential. And features of different embodiments may be interchanged if feasible.

[0045] Unless otherwise defined, all terms used herein (including technical and scientific terms) have their ordinary meanings, which are understandable to those skilled in the art. Furthermore, the definitions of the foregoing terms in commonly used dictionaries should be interpreted in the context of this specification as having the meaning consistent with the relevant field of this invention. Unless specifically defined, these terms will not be construed as having idealized or overly formal meanings.

[0046] When the sheet metal is subjected to excessively high blank holder force or stamping force during the stamping process, the sheet metal is subjected to large compressive stress in a local area, which exceeds the yield limit of the material and is prone to wrinkling. If the friction between the sheet metal and the die is not effectively controlled, it will also lead to local deformation restriction, stress concentration, and increased risk of wrinkling.

[0047] See Figure 1As shown, the present application provides a prediction control system for automobile bearing stamping part wrinkling, which predicts the wrinkling of the stamping part based on the changes of the plate during the stamping process, determines the changes of the plate after stamping, and adjusts the stamping based on the change state of the stamping part.

[0048] The prediction control system comprises:

[0049] The data acquisition unit is composed of a fluctuation acquisition matrix and a variable acquisition matrix, which acquires the control parameters of the control system for the stamping equipment, acquires the fluctuation change data of the plate during the stamping process through the fluctuation acquisition matrix, and acquires the variable data of the plate during the stamping process through the variable acquisition matrix;

[0050] The preprocessing unit splits the fluctuation change data and the variable data into data segments after noise suppression through noise filtering;

[0051] The stamping model is established based on historical stamping data, performs differential analysis on the fluctuation change data, determines the abnormal data segments of the historical stamping data, determines the abnormal sound wave and vibration wave characteristics of the plate, combines the variable data, predicts the wrinkling of the stamping part through a wrinkling prediction algorithm, and generates a prediction result;

[0052] The feedback control unit receives the prediction result, feeds back the abnormal state of the plate, and adjusts the stamping parameters according to the feedback result.

[0053] In an implementable manner, when the plate is insufficiently lubricated or the contact between the plate and the die is uneven during stamping, abnormal high-frequency friction sound or sudden impact sound will be generated, and local stress concentration and local sudden deformation will cause irregular vibration of the equipment or the plate itself. These abnormal sound waves and vibration waves are captured by the data acquisition unit for analysis to obtain the deformation state of the plate inside the die.

[0054] Among them, the sound wave and vibration wave characteristics are combined with the variable data to further verify the sound wave and vibration wave to determine the change state of the plate, the wrinkling risk is accurately identified through multi-dimensional data cross verification, the stamping process is adjusted in real time, the plate forming quality is ensured, and the production efficiency is improved.

[0055] In one embodiment, the fluctuation acquisition matrix is composed of a plurality of acoustic emission sensors and vibration sensors, which are distributed at different positions of the die to capture the sound wave and vibration wave signals of the plate during the stamping process with the middle part of the plate as the reference point;

[0056] A three-dimensional coordinate space is established and the coordinate point of each sensor in the coordinate space is determined, the data collected by the sensor is mapped to the coordinate point, a three-dimensional distribution map of the sound wave and vibration wave is constructed, and the wrinkling position of the plate is determined by comparing and analyzing the signal intensity, frequency change and time sequence change at different positions.

[0057] In an implementable manner, during the stamping process of the plate, due to different positions of deformation and stress distribution, the positions of vibration generation and sound wave points are different, thereby forming unique acoustic vibration characteristic maps. By monitoring the changes in the characteristics in real time, the wrinkling trend can be accurately judged.

[0058] In the embodiment, the wave transmission strength, frequency, time sequence change and other parameters are recorded in real time, and when the data is mapped in the spatial coordinates, the plate in the mold can be positioned, and the wrinkling area of the plate can be determined.

[0059] In an embodiment, the variable acquisition matrix includes a plurality of pressure sensors and displacement sensors, the pressure sensors are located at the high stress or weak area position in the mold, and the displacement sensors are arranged outside the mold to monitor the displacement distance of the mold during the stamping process.

[0060] The pressure sensor and the displacement sensor both have a unified clock synchronization function, a pressure distribution surface graph is constructed based on the pressure data with the synchronous time as the reference, and the dynamic stress displacement model is generated by fusing the displacement data, and the plate under pressure is mapped.

[0061] In an implementable manner, by comparing the acoustic vibration characteristics and the stress displacement model in real time, the wrinkling risk point can be accurately positioned, and during the stamping process of the plate, the pressure will change with the degree of deformation, and when the plate is normally pressed, the pressure change of the plate is normal pressure change, and when the plate shows signs of wrinkling, the pressure change will significantly deviate from the normal range.

[0062] By monitoring the deviation degree of the pressure change and the acoustic vibration characteristics in real time, the system can quickly identify and warn the wrinkling risk, ensure the stability of the plate forming quality, and effectively improve the production efficiency.

[0063] In an embodiment, the variable acquisition matrix includes a plurality of temperature sensors and strain sensors, the temperature sensors are arranged in the key heat area of the mold, and the strain sensors are attached to the surface of the plate to monitor the temperature change and strain distribution in real time. The temperature sensor and the strain sensor coordinate points are mapped in the three-dimensional coordinate space to construct a temperature strain distribution graph.

[0064] In an implementable manner, the temperature change of the plate during the stamping process is obtained by the temperature sensor and the strain sensor, and the temperature strain distribution graph is combined in the three-dimensional distribution graph. Through the three-dimensional coordinate space, the various change states of the plate from stamping to deformation, including temperature, sound wave and vibration wave, can be more directly determined, and the analysis of the plate can be more comprehensive.

[0065] In another embodiment, the strain sensor is arranged inside the mold. Since the strain sensor is thin and flexible, it is necessary to avoid the influence of the strain sensor on the plate. Therefore, the thickness and flexibility of the strain sensor are required to avoid the influence on the plate. Therefore, the strain sensor is installed inside the mold to monitor the temperature change of the position where the mold contacts the plate, thereby reducing the difficulty of selecting the strain sensor.

[0066] In one embodiment, the vibration sensor acquires vibration change data, and the acoustic emission sensor acquires acoustic wave change data. The vibration change data and the acoustic wave change data are subjected to noise suppression by the preprocessing unit, are split into data segments based on a complete vibration period in the vibration change data, and the periodicity of the signal is retained.

[0067] In an implementable manner, the vibration change data and the acoustic wave change data are split in a form that retains the periodicity, which retains the periodicity and helps to extract key information about the vibration mode, the peak and the valley distribution, and more accurately capture the vibration characteristics in each period, so as to facilitate comparison and prediction of abnormal vibration characteristics.

[0068] In another embodiment, the vibration change data and the acoustic wave change data are split based on a fixed time window. No additional detection of the period is required, so the segmentation is simpler and easier to implement. Based on the time window, a stable segmentation strategy is provided, so that the length of each data segment remains consistent. The relatively robust and uniform data division can be applied to various stamping states, and is used when there is less historical stamping data and the data fluctuates complexly.

[0069] In one embodiment, the steps of determining the abnormal acoustic wave and vibration wave characteristics of the plate are as follows:

[0070] The characteristics of each vibration change data segment and acoustic wave change data segment are extracted through time series and frequency domain analysis;

[0071] A feature distribution model in a normal stamping process is established based on historical stamping data to constitute a feature baseline, which describes the feature range, distribution and evolution trend in a normal stamping state.

[0072] The characteristics of the vibration and acoustic wave change data segments are compared with the feature distribution model to determine whether the characteristics of the vibration and acoustic wave change data segments are within the normal range.

[0073] Whether the data segment is abnormal is determined by setting a threshold or an abnormality detection algorithm, and the abnormal data segment is marked.

[0074] In this embodiment, the vibration data characteristics include but are not limited to: maximum amplitude, mean value, variance, spectral peak, spectral distribution, and energy distribution.

[0075] The acoustic data features include but are not limited to: signal strength, frequency component, frequency band energy, acoustic emission.

[0076] In an implementable manner, the historical stamping data records the vibration signal and the acoustic signal during the entire stamping process, and the vibration signal and the acoustic signal are preprocessed in each stamping cycle, and then the feature baseline in the normal stamping process is constructed through the historical stamping data.

[0077] The distance metric is used to determine whether the real-time data is within the normal range. If the amplitude, frequency spectrum distribution or acoustic energy in a certain data segment in the real-time data exceeds the normal interval defined by the baseline, the data segment has a significant difference from the historical stamping data.

[0078] Wherein, the data segment feature deviating from the normal range is marked as an abnormal data segment.

[0079] In an embodiment, the abnormal data segment establishes a wrinkle prediction algorithm in combination with variable data:

[0080] The timestamp of the abnormal data segment is determined, the variable data range is determined based on the timestamp, each group of corresponding abnormal data segments and variable data forms a multi-dimensional feature vector, and a state label is generated;

[0081] The signal features in the abnormal data segment and the variable features in the variable data are extracted.

[0082] The signal features and the variable features are constructed into a complete feature vector X.

[0083] The historical stamping data is used to establish a learning model according to the feature vector X and the label, and the learning model is further optimized through cross-validation and feature importance analysis to screen sensitive features.

[0084] In an implementable manner, the multi-dimensional feature vector label is 0 or 1, and the label 0 represents normal sheet without wrinkles, and the label 1 represents abnormal sheet with wrinkles.

[0085] For example, the feature vector X = (vibration RMS, vibration peak, acoustic energy, pressure change rate, stamping speed).

[0086] The random forest algorithm process is as follows:

[0087] import numpy as np

[0088] from sklearn.ensemble import RandomForestClassifier

[0089] from sklearn.model_selection import train_test_split

[0090] from sklearn.metrics import classification_report

[0091] # Assuming we have already constructed the dataset X and the label y

[0092] # X.shape = (n_samples, n_features)

[0093] # y.shape = (n_samples) where 0 indicates no wrinkling, and 1 indicates a risk of wrinkling.

[0094] # Split the training set and the test set

[0095] X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

[0096] # Building a Random Forest Classifier

[0097] model = RandomForestClassifier(n_estimators=100, random_state=42)

[0098] model.fit(X_train, y_train)

[0099] # Make predictions on the test set

[0100] y_pred = model.predict(X_test)

[0101] print(classification_report(y_test, y_pred))

[0102] In the example above, the model was further optimized using cross-validation and feature importance analysis to select the features most sensitive to the risk of wrinkling.

[0103] In one embodiment, during the stamping process, the collected vibration and acoustic wave change data and corresponding process parameter data are preprocessed and feature extracted by the wave acquisition matrix and variable acquisition matrix, and a new feature vector is generated. The new feature vector is then input into the prediction model to determine the prediction result and risk probability.

[0104] For example,

[0105] # new_feature_vector is the one-dimensional feature vector generated in real time.

[0106] wrinkle_prob = model.predict_proba(new_feature_vector.reshape(1, -1))[0][1] # Get the probability of being predicted as wrinkled

[0107] threshold = 0.6 # Preset threshold (can be optimized using historical data)

[0108] if wrinkle_prob>threshold:

[0109] # Trigger feedback control, such as: reducing stamping speed, adjusting die pressure, or increasing lubrication.

[0110] trigger_feedback_control()

[0111] By inputting real-time data into the model, if the predicted risk exceeds a preset threshold, the process parameters are automatically or manually adjusted to achieve closed-loop control and prevent sheet metal wrinkling. The stamping process can utilize historical and real-time data for dynamic monitoring and prediction, thereby significantly reducing the defect rate caused by wrinkling and improving the level of intelligence in the production process.

[0112] In one embodiment, the steps for determining the wrinkling location of the board material based on three-dimensional coordinate space are as follows:

[0113] The acoustic emission sensor and vibration sensor collect sound waves and vibration signals during the stamping process and associate them with their corresponding coordinate points.

[0114] Using spatial interpolation, the data from each discrete coordinate point are extended to the entire surface area of ​​the board, forming a three-dimensional distribution map of sound and vibration signals.

[0115] For the abnormal characteristics of each point in the entire region, cluster analysis is used to group coordinate points with similar abnormal characteristics to determine whether there are locally clustered abnormal regions.

[0116] By combining the spatial distribution map with the clustering results, the location of the abnormal region is identified in the three-dimensional coordinate system.

[0117] In an implementable manner, by comprehensively analyzing the fluctuation intensity, timing change and frequency characteristics collected by each sensor in a three-dimensional coordinate space, the sound wave and vibration distribution map of the plate during the stamping process can be accurately constructed. By comparing the data at different positions, when there is an abnormal aggregation in a local area, it can be preliminarily judged that there is a risk of wrinkling in the area. Using spatial clustering method and multi-dimensional feature fusion technology, not only the wrinkling risk position can be predicted, but also the basis for real-time feedback control can be provided, so as to realize closed-loop regulation and ensure the stamping quality.

[0118] In an embodiment, for each sensor coordinate point, the pre-processed fluctuation intensity, timing information and frequency characteristics are integrated into a multi-dimensional feature vector F i , wherein the multi-dimensional feature vector F i represents the feature vector of sensor i.

[0119] For example, in a stamping cycle, the sensors collect the following pre-processed data characteristics (the example data unit is assumed value):

[0120] Signal intensity (amplitude / RMS): A(0,0): 1.0, B(10,0): 1.1, C(-10,0): 2.5, D(0,10): 1.0, E(0,-10): 1.2;

[0121] Main frequency (Hz): A(0,0): 2000, B(10,0): 2050, C(-10,0): 2500, D(0,10): 1980, E(0,-10): 2010;

[0122] Timing characteristics (such as signal start time or phase change rate): The phase change of most points is relatively stable, and the C point has a significant advance phenomenon and a large phase fluctuation.

[0123] After digital signal processing, a multi-dimensional feature vector is formed, and the feature vector of sensor C is: F c =[intensity=2.5, main frequency=2500 Hz, phase fluctuation=large];

[0124] The data of the remaining sensors are relatively similar, all reflecting normal state.

[0125] By comparing the data baseline under the historical normal state, the signal intensity collected at each position during the normal stamping process is in the range of 0.9-1.2, the frequency is stable in the range of 1900-2100 Hz, and the timing signal is smooth. The data collected by sensor C is obviously abnormal-amplitude 2.5, main frequency 2500 Hz, and large phase fluctuation.

[0126] In combination with the abnormal signal of the position where the sensor C is located, it is judged that there is too high vibration energy and frequency offset in the local area, it is prompted that the plate in this place may be wrinkled due to abnormal local stress, then a warning is issued, and information is fed back to the stamping control unit through the feedback control unit, so as to automatically adjust the local die distribution pressure, reduce the stamping speed or improve the lubrication scheme to alleviate the stress concentration in this area.

[0127] The application has the following technical effects:

[0128] In the process of stamping the plate, the vibration wave and sound wave change of the plate are monitored to determine the state change of the plate in the die, so as to achieve the purpose of wrinkle prediction, and further improve the quality of stamping processing of the plate.

[0129] The sound waves and vibration waves generated by the normal state of plate stamping and the wrinkled state after stamping are inconsistent, the obtained fluctuation change data is segmented to form data segments, each data segment is analyzed, and the data segment corresponding to the time stamp can facilitate the combination of variable data to verify the wrinkled state.

[0130] During the stamping process of the plate, due to the different positions of deformation and stress distribution, the positions of vibration generation and sound wave points are different, thereby forming a unique sound vibration characteristic spectrum, the position of the abnormal point is analyzed through the coordinate space, and the position of the wrinkle generated can be determined.

[0131] Although the application is disclosed in combination with the above embodiments, it is not intended to limit the application, and any person skilled in the art can make various modifications and decorations without departing from the spirit and scope of the application, therefore the protection scope of the application should be defined by the appended claims.

Claims

1. A predictive control system for wrinkling of stamped automotive bearing parts, characterized in that, include: The data acquisition unit consists of a fluctuation acquisition matrix and a variable acquisition matrix. This data acquisition unit acquires the control parameters of the control system for the stamping equipment, acquires the fluctuation change data of the sheet metal during the stamping process through the fluctuation acquisition matrix, and acquires the variable data of the sheet metal during the stamping process through the variable acquisition matrix. The preprocessing unit performs noise filtering to suppress noise in fluctuating and variable data before splitting it into data segments. The stamping model is established based on historical stamping data. It acquires fluctuation data for differential analysis, identifies abnormal data segments compared with historical stamping data, determines the abnormal acoustic and vibration wave characteristics of the sheet metal, and combines variable data to predict wrinkling of stamped parts through a wrinkling prediction algorithm, generating prediction results. The feedback control unit receives the prediction results, provides feedback on abnormal conditions of the sheet metal, and adjusts the stamping parameters based on the feedback results. The wave acquisition matrix consists of multiple acoustic emission sensors and vibration sensors, which are distributed at different positions of the mold with the middle of the sheet as the reference point to capture the sound wave and vibration wave signals of the sheet during the stamping process. A three-dimensional coordinate space is established and the coordinate points of each sensor in this coordinate space are determined. The data collected by the sensors are mapped to the coordinate points to construct a three-dimensional distribution map of sound waves and vibration waves. By comparing and analyzing the signal intensity, frequency changes and time sequence changes at different locations, the wrinkling location of the board is determined.

2. The predictive control system for wrinkling of automotive bearing stamped parts according to claim 1, characterized in that, The variable acquisition matrix includes multiple pressure sensors and displacement sensors. The pressure sensors are located in high-stress or weak areas inside the mold, while the displacement sensors are set outside the mold to monitor the mold displacement distance during the stamping process. Both the pressure sensors and the displacement sensors have a unified clock synchronization function. Based on the pressure data, a pressure distribution surface map with the synchronization time as the reference is constructed and fused with the displacement data to generate a dynamic stress-displacement model that maps the pressure state of the sheet metal.

3. The predictive control system for wrinkling of automotive bearing stamped parts according to claim 2, characterized in that, The variable acquisition matrix includes multiple temperature sensors and strain sensors. Temperature sensors are arranged in the key hot areas of the mold, and strain sensors are attached to the surface of the plate to monitor temperature changes and strain distribution in real time. The coordinate points of the temperature sensors and strain sensors are mapped in three-dimensional coordinate space to construct a temperature-strain distribution map.

4. The predictive control system for wrinkling of automotive bearing stamped parts according to claim 3, characterized in that, Vibration sensors acquire vibration change data, and acoustic emission sensors acquire sound wave change data. A preprocessing unit performs noise suppression on the vibration change data and sound wave change data. Based on the complete vibration cycle in the vibration change data, the vibration change data and sound wave change data are split into data segments to retain the periodic characteristics of the signal.

5. A predictive control system for wrinkling of stamped automotive bearing parts according to claim 3 or 4, characterized in that, The steps to determine the characteristics of abnormal sound waves and vibration waves in the board material are as follows: Feature extraction was performed on each vibration variation data segment and sound wave variation data segment using time-series and frequency domain analysis. A feature baseline is established based on historical stamping data to construct a feature distribution model for the normal stamping process, which describes the feature range, distribution and evolution trend under normal stamping conditions. The characteristics of the vibration and sound wave change data segments are compared with the characteristic distribution model to determine whether the characteristics of the vibration and sound wave change data segments are within the normal range. The system determines whether a data segment is abnormal by setting a threshold or using an anomaly detection algorithm, and then marks the abnormal data segments.

6. The predictive control system for wrinkling of stamped automotive bearing parts according to claim 5, characterized in that, A wrinkle prediction algorithm is established by combining abnormal data segments with variable data: Determine the timestamp of the abnormal data segment, determine the range of variable data based on the timestamp, form a multi-dimensional feature vector for each group of corresponding abnormal data segments and variable data, and generate a status label. Extract signal features from abnormal data segments and variable features from variable data; Construct a complete feature vector X from the signal features and the variable features; Using historical stamping data, a learning model is built based on the feature vector X and the label. The learning model is further optimized and sensitive features are screened through cross-validation and feature importance analysis.

7. A predictive control system for wrinkling of stamped automotive bearing parts according to claim 6, characterized in that, During the stamping process, the vibration and acoustic wave change data and corresponding process parameter data collected by the wave acquisition matrix and variable acquisition matrix are preprocessed and feature extracted to generate new feature vectors. The new feature vectors are then input into the prediction model to determine the prediction results and risk probabilities.

8. A predictive control system for wrinkling of stamped automotive bearing parts according to claim 7, characterized in that, The steps for determining the wrinkling location of the board material based on three-dimensional coordinate space are as follows: The acoustic and vibration signals collected by each acoustic emission sensor and vibration sensor during the stamping process will be associated with their corresponding coordinate points. Using spatial interpolation, the data from each discrete coordinate point are extended to the entire surface area of ​​the board, forming a three-dimensional distribution map of sound and vibration signals. For the abnormal characteristics of each point in the entire region, cluster analysis is used to group coordinate points with similar abnormal characteristics to determine whether there are locally clustered abnormal regions. By combining the spatial distribution map with the clustering results, the location of the abnormal region is identified in the three-dimensional coordinate system.

9. A predictive control system for wrinkling of stamped automotive bearing parts according to claim 8, characterized in that, For each sensor coordinate point, the preprocessed wave intensity, time series information, and frequency characteristics are integrated into a multi-dimensional feature vector. F i Among them, multidimensional feature vectors F i This represents the feature vector of sensor i.

Images