Feedback control method and system for heavy truck transmission system sand casting shell finishing process

Abstract

This application relates to the field of feedback control technology in precision machining, specifically to a feedback control method and system for the precision machining process of sand casting housings in heavy-duty truck transmission systems. The method includes: analyzing the overall numerical distribution and fluctuation characteristics of reference data at each machining position of the sand casting housing, and classifying the vibration levels of each machining position; analyzing the observed data of each group of housings based on the control limits determined by the reference data to obtain the runaway frequency of each machining position in each group of housings; determining the vibration deviation of each machining position in each group of housings; training a nonlinear characteristic calculation model to obtain the nonlinear characteristic values ​​of each machining position; determining the instability characteristics of each machining position; determining the initial PID parameters using the Ziegler-Nichols method; and dynamically adjusting the PID parameters based on the characteristic deviation of the machining position during real-time precision machining. This application aims to improve the efficiency of precision machining of sand casting housings.

Description

Technical Field

[0001] This application relates to the field of precision machining feedback control technology, specifically to a feedback control method and system for the precision machining process of sand casting housings in heavy-duty truck transmission systems. Background Technology

[0002] As a core component of new energy heavy-duty trucks, the machining quality of the parts in the heavy-duty truck transmission system directly affects the vehicle's power transmission efficiency, reliability, and service life. The heavy-duty truck transmission system includes various precision components with complex curved surface structures, such as the transmission housing, transmission bearings, and gears. Currently, the machining processes for heavy-duty truck transmission system components are not yet perfect, resulting in insufficient machining accuracy, low production efficiency, and high costs, making it difficult to meet the demands of the industry's rapid development. Therefore, the development and research of machining processes for new energy heavy-duty truck transmission system components is urgently needed.

[0003] The housings in heavy-duty truck transmission systems are typically cast using sand casting. During this process, after demolding, the sand-cast housing requires precision machining to ensure its final dimensions and geometric tolerances strictly conform to the CAD drawings. This is usually achieved using a PID control algorithm with preset fixed parameters for feedback control of the machining process. However, in actual sand casting housing precision machining, the housing is not a regular structure, and the material allowance at different machining locations is random. Using fixed PID parameters for control will lead to a mismatch between the controller response and real-time process dynamics, resulting in the sand casting housing precision machining failing to meet the CAD drawing requirements and impacting production yield. Summary of the Invention

[0004] In view of the above, it is necessary to provide a feedback control method and system for the precision machining process of sand casting housings in heavy-duty truck transmission systems to solve the above problems.

[0005] The first aspect of this application provides a feedback control method for the finishing process of sand casting housings in heavy-duty truck transmission systems, the method comprising: The first and second historical vibration data of each group of shells in the sand casting were used as the benchmark data and observation data, respectively, during the precision machining of the shells. The overall numerical distribution and fluctuation characteristics of the reference data for each machining position of the sand casting shell are analyzed, and the vibration level of each machining position of the shell is classified. Based on the control limits determined by the reference data, the observation data of each group of shells are analyzed to obtain the runaway frequency of each machining position of each group of shells. The degree of difference between the comprehensive characteristics of the observation data of each machining position and the reference data is calculated to determine the vibration deviation of each machining position of each group of shells. The vibration deviation is used as input and the runaway frequency is used as output. A regression model is used to obtain the nonlinear characteristic value of each machining position. Based on the overall distribution differences of nonlinear characteristic values ​​of different vibration levels, and combined with the nonlinear characteristic values ​​of each processing position, the instability characteristics of each processing position are determined. Based on the instability characteristics of each processing position, the initial PID parameters are determined by the Ziegler-Nichols method. During real-time finishing, the characteristic deviation of each processing position within a preset time window is calculated, and the PID parameters are dynamically adjusted in combination with the average characteristic deviation determined by the previous processing position.

[0006] Preferably, the vibration level classification for each processing position of the housing is specifically as follows: For each processing position of the shell, the mean value of each group of first historical vibration data is recorded as the first time feature, and the coefficient of variation of each group of first historical vibration data is recorded as the second time feature. The dispersion of the first temporal feature of the first set of historical vibration data in a preset number of sets is used as the first spatial feature; the dispersion of the second temporal feature of the preset number of historical vibration data is used as the second spatial feature. The first spatial feature and the second spatial feature of each processing position are normalized and then averaged to obtain the reference fluctuation feature of each processing position. Based on the baseline wave characteristics, the first historical vibration data of all processing positions of the shell are clustered to obtain a preset number of clusters; Calculate the mean value of the baseline fluctuation characteristics within each cluster, and mark the vibration level of the shell processing location within the cluster from high to low in descending order.

[0007] Preferably, when the precision machining position of the housing is less than the preset value, the vibration level of the housing machining position is marked from high to low by the value of the reference fluctuation characteristic of each machining position from large to small.

[0008] Preferably, the process of obtaining the runaway frequency at each processing position of each group of shells specifically involves: SPC control limits are established based on the first historical vibration data of each processing position using a process control algorithm; vibration data outside the SPC control limits of the second historical vibration data of each processing position are taken as out-of-control data, and the proportion of out-of-control data of all second historical vibration data of each processing position is taken as the out-of-control frequency of each processing position.

[0009] Preferably, the step of calculating the degree of difference between the comprehensive characteristics of the observed data at each processing position and the reference data to determine the vibration deviation at each processing position of each group of shells is as follows: The mean and coefficient of variation of the second historical vibration data of each processing position are calculated, and the mean is calculated after normalization to obtain the first feature of each processing position. The difference between the first characteristic and the reference fluctuation characteristic at each processing position is calculated as the vibration deviation at each processing position.

[0010] Preferably, determining the instability characteristics of each processing position specifically involves: The absolute value of the difference between the nonlinear characteristic value of each processing position and the mean value of the nonlinear characteristic value of all processing positions under the adjacent vibration level is calculated to obtain the nonlinear difference characteristic value of each processing position. The instability characteristics of each processing position are obtained by geometrically averaging the nonlinear eigenvalues ​​and nonlinear difference eigenvalues ​​at each processing position.

[0011] Preferably, the initial PID parameters are determined using the Ziegler-Nichols method, and the specific calculation rules are as follows: in, 、 、 These represent the initial PID parameters for the i-th processing position calculated by the Ziegler-Nichols method; , , These represent the gain coefficient, period coefficient, and correction factor, respectively. This represents the normalized result of the instability characteristics at the i-th processing position; This indicates the preset first scaling factor; , These represent the preset reference gain and the preset reference period, respectively.

[0012] Preferably, the calculation of the feature deviation within a preset time window for each processing position specifically involves: Based on the nonlinear feature model trained by the regression model, the baseline nonlinear feature value of each processing position is calculated based on the first historical vibration data, and each nonlinear feature value is calculated based on each group of second historical vibration data. The maximum difference between the baseline nonlinear eigenvalue and all obtained nonlinear eigenvalues ​​is taken as the historical maximum tolerance for each processing position. The difference between the nonlinear eigenvalue of each processing position in each time window and the historical nonlinear eigenvalue of the corresponding processing position is calculated, and then positively fused with the negative correlation mapping result of the historical maximum tolerance of the processing position to obtain the eigenvalue deviation.

[0013] Preferably, the dynamic adjustment of PID parameters specifically includes: in, , , These are the dynamically adjusted PID parameters for the i-th processing position; This is the preset second reduction factor; This represents the feature deviation of the i-th processing position in the a-th time window; This represents the average feature deviation of all finishing positions before the i-th machining position.

[0014] Secondly, embodiments of this application also provide a feedback control system for the finishing process of sand casting housing of heavy truck transmission system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0015] This application has at least the following beneficial effects: This application first uses the first and second historical vibration data of the finishing of each group of shells in sand casting as benchmark data and observation data, respectively. It analyzes the overall numerical distribution and fluctuation characteristics of the benchmark data. Through the analysis of the benchmark data, the normal vibration range and characteristics of each machining position can be clearly understood, providing a basis for subsequent runaway judgment and control limit setting. Vibration level classification based on the distribution characteristics of the benchmark data helps identify normal and abnormal states, thereby improving the accuracy of fault detection. Based on the control limits determined by the benchmark data, the observation data of each group of shells is analyzed to obtain the runaway frequency of each machining position of each group of shells. Monitoring the runaway frequency can promptly detect potential problems during machining, ensuring machining quality and improving production efficiency. The degree of difference in comprehensive characteristics between the observation data of each machining position and the benchmark data is calculated to determine the vibration deviation of each machining position of each group of shells. The quantification of vibration deviation provides intuitive feedback information for the control system, enabling the controller to adjust according to the actual situation. Linear characteristic models can capture the complex relationship between vibration deviation and runaway frequency, providing support for subsequent control parameter optimization and helping to improve the predictive ability of future vibration states. Based on the overall distribution differences of nonlinear characteristic values ​​of different vibration levels, combined with the nonlinear characteristic values ​​of each processing position, the instability characteristics of each processing position can be determined. Corresponding control measures can be designed for different processing positions and their characteristics, improving the stability of the entire processing process. By setting reasonable PID parameters, the system's response speed and steady-state performance can be significantly improved, reducing overshoot and oscillation phenomena. Adjusting PID parameters enables the system to better adapt to environmental changes and process fluctuations, ensuring the stability and reliability of the processing process. By calculating the characteristic deviation in real time, the control system can not only adapt to the fluctuations of different processing at the same position, but also respond to the influence of different processing states at different positions on the same workpiece. This solves the problem of slow response caused by fixed parameters or oscillation instability caused by aggressive parameters in traditional methods, improving the efficiency of shell finishing. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the steps of a feedback control method for the finishing process of a sand casting housing of a heavy-duty truck transmission system, provided in one embodiment of this application. Detailed Implementation

[0017] The following, in conjunction with the accompanying drawings, details the feedback control method and system scheme for the precision machining process of the sand casting housing of the heavy-duty truck transmission system provided in this application.

[0018] Please see Figure 1 The diagram illustrates a flowchart of a feedback control method for the finishing process of a sand casting housing in a heavy-duty truck transmission system according to an embodiment of this application. The method includes the following steps: Step 1: Use the first and second historical vibration data of each group of sand casting shells for finishing as the reference data and observation data, respectively.

[0019] This application first collects vibration data from historical finishing processes, then collects runaway data calculated using process control algorithms from the historical finishing vibration data, and finally collects vibration and runaway data in real time during the finishing process.

[0020] In this embodiment, historical finishing data is acquired using a vibration acceleration sensor in a CNC machine tool. The acquisition frequency is set to 2000Hz, and the first historical vibration data of all machining positions of 50 sets of shells during the finishing process is acquired as benchmark data to establish statistical control limits under normal machining conditions. The first historical vibration data is used as input, and the upper and lower control limits and their differences of the first historical vibration data of all machining positions are obtained using a process control algorithm (statistical process control is used in this embodiment). Then, the second historical vibration data and the number of out-of-control data of all machining positions of 50 sets of shells during the finishing process are acquired. The second historical vibration data is used as observation data to represent the sample of subsequent actual machining and to verify its actual fluctuation. Among them, out-of-control data refers to the vibration data at each machining position where the second historical vibration data fails to meet the process control rules defined by the out-of-control diagram generated based on the first batch of historical vibration data. The vibration data of each machining position during each finishing process corresponds to a number of out-of-control data.

[0021] This application does not impose special restrictions on the acquisition frequency or the amount of historical data for shell finishing, which can be adjusted according to actual conditions; all historical vibration data in this application are only used for offline calculation, and there is no need to consider the computational complexity during real-time finishing; the process control algorithm for obtaining the upper and lower control limits and their differences, as well as the runaway data, are all well-known technologies in the field, and the specific calculation methods will not be elaborated here.

[0022] In the actual finishing process, vibration data and runaway data are collected in real time. During the finishing process, the sensor may be affected by the splashing of shell material, which may cause noise in the collected vibration data and affect subsequent analysis and judgment. Therefore, in this embodiment, the mean filtering algorithm is used to preprocess all historical vibration data to eliminate external interference. The mean filtering algorithm is a well-known technology in the field, and the specific calculation method will not be described in detail.

[0023] Step 2: Analyze the overall numerical distribution and fluctuation characteristics of the reference data for each machining position of the sand casting shell, and classify the vibration level of each machining position of the shell; based on the control limits determined by the reference data, analyze the observation data of each group of shells to obtain the runaway frequency of each machining position of each group of shells; calculate the degree of difference between the comprehensive characteristics of the observation data of each machining position and the reference data to determine the vibration deviation of each machining position of each group of shells; use the vibration deviation as input and the runaway frequency as output, and use a regression model to obtain the nonlinear characteristic value of each machining position.

[0024] The housing is an irregularly shaped part, and its different machining positions need to be processed separately during its finishing process. The material allowance at different machining positions during the same finishing process and at the same machining position during different finishing processes are both random. The more material allowance, the greater the degree of finishing required, and the greater the amplitude and fluctuation of the tool vibration data. The randomness of the vibration data at different machining positions directly affects the rationality of the runaway rules of the process control algorithm and the accuracy of the feedback control of the PID algorithm. Therefore, in this embodiment, by analyzing the vibration data at different machining positions during the housing finishing process, the upper and lower limits of control and runaway data of each machining position are determined, thereby providing a basis for the initial parameter setting of the PID control algorithm. It should be noted that the PID parameters in this application specifically refer to the parameters of the process compensation PID controller and do not directly interfere with the closed-loop servo PID of the underlying motor driver of the CNC machine tool.

[0025] Based on the above analysis, the first historical vibration data is processed to obtain the upper and lower control limits of each processing position of the shell. Specifically, for each processing position of the shell, the mean of each group of first historical vibration data is recorded as the first time feature, and the coefficient of variation of each group of first historical vibration data is recorded as the second time feature. The dispersion of the first time feature of a preset number of groups of first historical vibration data is used as the first spatial feature. The dispersion of the second time feature of a preset number of groups of historical vibration data is used as the second spatial feature, and the min-max normalization method is used for processing to reflect the spatial characteristics of the processing position. In this embodiment, the preset number is 50, and the dispersion is calculated using the coefficient of variation. When performing normalization calculation, a very small positive number (0.0001 in this embodiment) is added to the denominator to prevent division by zero errors caused by the complete consistency of data in the same batch.

[0026] At this point, each processing position of the shell has two spatial features. The average of the two spatial features of each processing position after normalization is used as the reference fluctuation feature of each processing position. The Euclidean distance between the reference fluctuation features of different processing positions of the shell is used as the metric distance. Using a clustering algorithm (K-means algorithm in this embodiment, K is set to a preset value of 3), the first historical vibration data of all processing positions of the shell are clustered to obtain three clusters. The mean of the reference fluctuation feature in each cluster is calculated, and the shell processing positions corresponding to the historical vibration data in the cluster are marked as strong vibration, medium vibration and low vibration in descending order.

[0027] It should be noted that for shells with fewer than three finishing positions, clustering is not necessary; vibration can be directly labeled based on the first feature of each finishing position. For example, if there are only three finishing positions, they are labeled as strong vibration, medium vibration, and low vibration in descending order of the first feature; if there are only two finishing positions, they are labeled as strong vibration and low vibration in descending order of the first feature.

[0028] Furthermore, for the second historical vibration data, each processing position can obtain the SPC control limit based on the first historical vibration data through the process control algorithm, and obtain the out-of-control data of each processing position based on the out-of-control rules of the SPC control limit; the proportion of out-of-control data in all the second historical vibration data of that processing position is used as the out-of-control frequency of each processing position (each processing position of each group of data has one out-of-control frequency).

[0029] For each group of second historical vibration data, the mean and coefficient of variation of the second historical vibration data at each processing position are calculated, and the mean is calculated after normalization to obtain the first feature of each processing position; at this time, each processing position of each group of second historical vibration data includes the first feature and the runaway frequency.

[0030] The difference between the first characteristic and the reference fluctuation characteristic of each processing position is calculated as the vibration deviation of each processing position; at this time, the vibration deviation of the second historical vibration data of each group of processing positions corresponds to a runaway frequency.

[0031] To analyze the nonlinear relationship between vibration deviation and runaway frequency (runaway characteristic), in this embodiment, a binary tuple consisting of all vibration deviations and runaway frequencies at each processing position is used as input. The kernel function in the Gaussian process regression model is set as a radial basis function, and the length parameter (range is...) is... In this embodiment, the value is 1). The Gaussian process regression (GPR) algorithm is used for training to obtain the nonlinear eigenvalues ​​and nonlinear eigenvalue calculation model between the vibration deviation and runaway frequency at each processing position. The min-max normalization method is used to normalize the nonlinear eigenvalues ​​of all processing positions. In this embodiment, the nonlinear eigenvalue is specifically the mean value predicted by the Gaussian process regression model for the current input data.

[0032] Nonlinear eigenvalues ​​are used to reflect the nonlinear relationship between the differences in vibration data at various processing positions of the shell and the number of runaway data calculated by the process control algorithm, reflecting the profound impact of the nonlinear distribution of material allowance on the stability of the process system in this scenario.

[0033] Step 3: Based on the overall distribution differences of nonlinear characteristic values ​​of different vibration levels, and combined with the nonlinear characteristic values ​​of each processing position, determine the instability characteristics of each processing position.

[0034] In theory, the vibration deviation range of a high-vibration processing position is relatively large, making the nonlinear characteristic value of the high-vibration processing position greater than that of the medium and low-vibration processing positions. Therefore, the larger the absolute value of the difference between the nonlinear characteristic value of a certain vibration level and the nonlinear characteristic value of the previous or next vibration level, the more obvious the nonlinearity of the vibration data of a certain processing position under that vibration level.

[0035] Specifically, if the vibration level category includes three categories and the i-th processing position belongs to the strong vibration level, the absolute value of the difference between the nonlinear characteristic value of the i-th processing position and the mean of the nonlinear characteristic values ​​of all processing positions within the medium vibration level is calculated, and this value is taken as the nonlinear difference characteristic value of the i-th processing position. If the vibration level category includes three categories and the i-th processing position belongs to the medium vibration level, the average of the absolute values ​​of the differences between the nonlinear characteristic value of the i-th processing position and the mean of the nonlinear characteristic values ​​of all processing positions within the strong and low vibration levels is calculated, and this value is taken as the nonlinear difference characteristic value of the i-th processing position. If the vibration level category includes two categories and the i-th processing position belongs to the strong vibration level, the absolute value of the difference between the nonlinear characteristic value of the i-th processing position and the mean of the nonlinear characteristic values ​​of all processing positions within the low vibration level is calculated, and this value is taken as the nonlinear difference characteristic value of the i-th processing position. Otherwise, the natural number 1 is taken as the nonlinear difference characteristic value of the i-th processing position. The nonlinear difference characteristic value characterizes the degree of specificity of the current processing position relative to the overall casting vibration characteristics. The stronger the specificity, the more independent control parameters are required. It should be noted that if the set of adjacent vibration levels is empty, the characteristic mean value within the current vibration level is directly taken as the substitute.

[0036] Furthermore, the nonlinear difference eigenvalues ​​obtained at each processing position are positively fused with the nonlinear eigenvalues ​​to obtain the instability characteristics of each processing position. In this embodiment, the geometric mean calculation method is used to positively fuse multiple variables.

[0037] It should be understood that the larger the nonlinear difference characteristic value of the i-th processing position, the greater the difference in nonlinear characteristics between the i-th processing position and the processing positions under adjacent vibration levels; the larger the nonlinear characteristic value of the i-th processing position, the greater the degree of nonlinearity between the vibration data deviation and the runaway data at the i-th processing position; the larger the instability characteristic of the i-th processing position, the greater the randomness of the vibration data fluctuation at the i-th processing position. When performing fine machining on this processing position, the P value of the PID algorithm should be reduced to stabilize the control and prevent insufficient or excessive machining at this processing position, which would cause the shell to fail to meet the drawing requirements.

[0038] Step 4: Based on the instability characteristics of each processing position, determine the initial PID parameters using the Ziegler-Nichols method; during real-time finishing, calculate the characteristic deviation within a preset time window for each processing position, and dynamically adjust the PID parameters by combining the average characteristic deviation determined by the previous processing position.

[0039] After processing the instability characteristics at different processing positions using the min-max normalization method, the results are used as input to the Ziegler-Nichols method, with the gain coefficient set (range: ...). The standard value is 1.2), and the periodic coefficient (range is...) (The normal value is 1), correction factor (the range is...) (The standard value is 1.05), to obtain the initial PID parameters for each processing position of the housing.

[0040] The initial PID parameter calculation rules for the Ziegler-Nichols method are as follows: 、 、 These represent the initial PID parameters for the i-th processing position calculated by the Ziegler-Nichols method; , , These represent the gain coefficient, period coefficient, and correction factor, respectively. This represents the normalized result of the instability characteristics at the i-th processing position; This indicates the preset first reduction factor to prevent... The adjustment is excessive; in this embodiment, the value is 0.3. This application does not impose special restrictions on the reduction factor, which can be adjusted according to the actual situation. , These represent the preset reference gain and the preset reference period, respectively, with value ranges of [value ranges missing]. , The usual values ​​are 30 and 0.3.

[0041] In actual shell finishing, taking the i-th machining position as an example, real-time vibration data of the i-th machining position is collected in real time, and a time window is preset (this application does not impose special restrictions on the time window. In this embodiment, it is assumed that the finishing time of the i-th machining position is 10 minutes. The time window is specifically a sliding time window with a length of 1 second and a step size of 20ms. The finishing time can be obtained from the historical finishing time of the i-th machining position).

[0042] When the real-time vibration data acquisition time is less than 20ms, no dynamic adjustment of the PID parameters is performed; feedback control is only performed using the adjusted initial PID parameters.

[0043] When the real-time vibration data acquisition time reaches 20ms, the first feature, runaway frequency, and vibration deviation within a time window are obtained. The vibration deviation is then input into the nonlinear feature calculation model to obtain the nonlinear feature value of the i-th processing position.

[0044] Calculate the baseline nonlinear characteristic value of the i-th processing position based on the first historical vibration data, and each nonlinear characteristic value obtained based on each group of second historical vibration data; take the maximum difference between the baseline nonlinear characteristic value and all obtained nonlinear characteristic values ​​as the historical maximum tolerance of the i-th processing position. Calculate the difference between the nonlinear characteristic value of the i-th processing position in the a-th time window and the historical nonlinear characteristic value of the corresponding processing position, and then perform positive fusion with the negative correlation mapping result of the historical maximum tolerance of the processing position to obtain the characteristic deviation of the i-th processing position in the a-th time window, denoted as . Wherein, the historical maximum tolerance of the i-th processing position is denoted as The negative correlation mapping result is obtained through the formula get, This represents a very small positive number. In this embodiment, the value is 0.01. In this embodiment, multiple variables are positively fused using a multiplication method.

[0045] To address the severe vibrations caused by surface hard spots or abrupt changes in allowance in sand castings, this solution employs a flexible avoidance control strategy: appropriately reducing the proportional gain. To reduce the rigid impact on the servo system while increasing the integral gain This eliminates steady-state errors caused by reduced rigidity, thereby suppressing vibration divergence while ensuring uninterrupted processing.

[0046] When the i-th machining position is the first machining position in the actual finishing process: When the i-th machining position is not the first machining position in the actual finishing process: in, , , These are the dynamically adjusted PID parameters for the i-th processing position; The second reduction factor is a preset value, which is 0.2 in this embodiment; This represents the feature deviation of the i-th processing position in the a-th time window; This represents the average feature deviation of all finishing positions before the i-th machining position. It characterizes the system cumulative error of the workpiece in the current clamping state (e.g., thermal deformation, tool wear accumulation), and is used to correct the reference parameters for subsequent machining positions. If the current position is the first machining position or the previous data is missing, The value is 0.

[0047] It should be noted that this application sets a safety lower limit threshold (the value in this embodiment is 0.1). When calculating the PID adjustment parameters, if the multiplier term in the formula... or If the calculated result is less than the safety lower limit threshold, then the multiplier term is directly taken as the safety lower limit threshold.

[0048] It should be understood that by reducing P (proportional gain) to decrease the rigidity of the servo system to suppress vibration and shock, and by increasing I (integral gain) to eliminate the steady-state error caused by the reduction in rigidity, machining accuracy can be guaranteed while protecting the tool.

[0049] Based on the same inventive concept as the above methods, this application also provides a feedback control system for the finishing process of sand casting housing of heavy truck transmission system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described feedback control methods for the finishing process of sand casting housing of heavy truck transmission system.

Claims

1. A feedback control method for the precision machining process of sand casting housings in heavy-duty truck transmission systems, characterized in that, The method includes the following steps: The first and second historical vibration data of each group of shells in the sand casting were used as the benchmark data and observation data, respectively, during the precision machining of the shells. The overall numerical distribution and fluctuation characteristics of the reference data for each machining position of the sand casting shell are analyzed, and the vibration level of each machining position of the shell is classified. Based on the control limits determined by the reference data, the observation data of each group of shells are analyzed to obtain the runaway frequency of each machining position of each group of shells. The degree of difference between the comprehensive characteristics of the observation data of each machining position and the reference data is calculated to determine the vibration deviation of each machining position of each group of shells. The vibration deviation is used as input and the runaway frequency is used as output. A regression model is used to obtain the nonlinear characteristic value of each machining position. Based on the overall distribution differences of nonlinear characteristic values ​​of different vibration levels, and combined with the nonlinear characteristic values ​​of each processing position, the instability characteristics of each processing position are determined. Based on the instability characteristics of each processing position, the initial PID parameters are determined by the Ziegler-Nichols method. During real-time finishing, the characteristic deviation of each processing position within a preset time window is calculated, and the PID parameters are dynamically adjusted in combination with the average characteristic deviation determined by the previous processing position.

2. The feedback control method for the finishing process of sand casting housing of heavy truck transmission system as described in claim 1, characterized in that, The vibration level classification for each machining location of the shell is as follows: For each processing position of the shell, the mean value of each group of first historical vibration data is recorded as the first time feature, and the coefficient of variation of each group of first historical vibration data is recorded as the second time feature. The dispersion of the first temporal feature of the first set of historical vibration data in a preset number of sets is used as the first spatial feature; the dispersion of the second temporal feature of the preset number of historical vibration data is used as the second spatial feature. The first spatial feature and the second spatial feature of each processing position are normalized and then averaged to obtain the reference fluctuation feature of each processing position. Based on the baseline wave characteristics, the first historical vibration data of all processing positions of the shell are clustered to obtain a preset number of clusters; Calculate the mean value of the baseline fluctuation characteristics within each cluster, and mark the vibration level of the shell processing location within the cluster from high to low in descending order.

3. The feedback control method for the finishing process of sand casting housing of heavy truck transmission system as described in claim 2, characterized in that, When the precision machining position of the housing is less than the preset value, the vibration level of the housing machining position is marked from high to low by the value of the reference fluctuation characteristic of each machining position from large to small.

4. The feedback control method for the finishing process of sand casting housing of heavy truck transmission system as described in claim 1, characterized in that, The process of obtaining the runaway frequency at each processing position of each housing group is as follows: SPC control limits are established based on the first historical vibration data of each processing position using a process control algorithm; vibration data outside the SPC control limits of the second historical vibration data of each processing position are taken as out-of-control data, and the proportion of out-of-control data of all second historical vibration data of each processing position is taken as the out-of-control frequency of each processing position.

5. The feedback control method for the finishing process of sand casting housing of heavy truck transmission system as described in claim 2, characterized in that, The step of calculating the degree of difference between the comprehensive characteristics of the observed data at each processing position and the reference data, in order to determine the vibration deviation at each processing position of each group of shells, is as follows: The mean and coefficient of variation of the second historical vibration data of each processing position are calculated, and the mean is calculated after normalization to obtain the first feature of each processing position. The difference between the first characteristic and the reference fluctuation characteristic at each processing position is calculated as the vibration deviation at each processing position.

6. The feedback control method for the finishing process of sand casting housing of heavy truck transmission system as described in claim 1, characterized in that, The determination of the instability characteristics at each processing position specifically includes: The absolute value of the difference between the nonlinear characteristic value of each processing position and the mean value of the nonlinear characteristic value of all processing positions under the adjacent vibration level is calculated to obtain the nonlinear difference characteristic value of each processing position. The instability characteristics of each processing position are obtained by geometrically averaging the nonlinear eigenvalues ​​and nonlinear difference eigenvalues ​​at each processing position.

7. The feedback control method for the finishing process of sand casting housing of heavy truck transmission system as described in claim 1, characterized in that, The initial PID parameters are determined using the Ziegler-Nichols method, and the specific calculation rules are as follows: in, 、 、 These represent the initial PID parameters for the i-th processing position calculated by the Ziegler-Nichols method; , , These represent the gain coefficient, period coefficient, and correction factor, respectively. This represents the normalized result of the instability characteristics at the i-th processing position; This indicates the preset first scaling factor; , These represent the preset reference gain and the preset reference period, respectively.

8. The feedback control method for the finishing process of sand casting housing of heavy truck transmission system as described in claim 1, characterized in that, The calculation of the feature deviation within a preset time window for each processing position is specifically as follows: Based on the nonlinear feature model trained by the regression model, the baseline nonlinear feature value of each processing position is calculated based on the first historical vibration data, and each nonlinear feature value is calculated based on each group of second historical vibration data. The maximum difference between the baseline nonlinear eigenvalue and all obtained nonlinear eigenvalues ​​is taken as the historical maximum tolerance for each processing position. The difference between the nonlinear eigenvalue of each processing position in each time window and the historical nonlinear eigenvalue of the corresponding processing position is calculated, and then positively fused with the negative correlation mapping result of the historical maximum tolerance of the processing position to obtain the eigenvalue deviation.

9. The feedback control method for the finishing process of sand casting housing of heavy truck transmission system as described in claim 7, characterized in that, The dynamic adjustment of PID parameters specifically refers to: in, , , These are the dynamically adjusted PID parameters for the i-th processing position; This is the preset second reduction factor; This represents the feature deviation of the i-th processing position in the a-th time window; This represents the average feature deviation of all finishing positions before the i-th machining position.

10. A feedback control system for the finishing process of sand casting housing of heavy truck transmission system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-9.

Images