A beat balance and cooperation control method and system of a crystal automatic processing production line
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YONGHAO OPTIC&ELECTRONIC CO LTD (CHINA)
- Filing Date
- 2026-03-26
- Publication Date
- 2026-06-26
AI Technical Summary
In the existing crystal processing production line, the individual processing cycle of each process in the manufacturing of zoom lenses is inconsistent, which leads to blockage of the entire line or idle running of equipment, making it difficult to achieve stable continuous automated processing.
Machine learning models are used to predict the processing cycle time baseline, and the cycle time is corrected by combining real-time operating status data. Key processes are identified and the timing of workpiece input and release intervals are adjusted in a coordinated manner to achieve process cycle time balance.
It achieves consistent control of process cycle time, avoids line blockage and equipment idling, and improves the continuous operation capability and overall efficiency of the production line.
Smart Images
Figure CN122284533A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of production line control technology, specifically to a cycle balance and collaborative control method and system for an automated crystal processing production line. Background Technology
[0002] The manufacturing process of zoom lenses and optical crystal components typically involves multiple steps, including rough grinding, fine grinding, polishing, and centering correction. Due to significant differences in material removal rates, machining accuracy requirements, and tool conditions among these steps, the processing cycle time for a single piece varies naturally between different pieces of equipment.
[0003] In existing technologies, most crystal processing production lines still use one of the following methods:
[0004] One approach is to passively coordinate by manually setting a uniform tempo or simply adding buffer bits.
[0005] Second, it relies on a fixed ratio of equipment quantity to approximate production capacity;
[0006] Third, it focuses on the efficiency of single-machine processing, ignoring the dynamic coupling relationship between processes.
[0007] The above method has the following shortcomings:
[0008] The production line cannot dynamically adjust the cycle time according to the different specifications, curvature and material differences of the zoom lens. When a certain process is affected by the wear of the grinding disc, the compensation of detection accuracy or the insertion of abnormal workpieces, it is easy to cause the entire line to be blocked or the equipment to run idle. The overall control strategy of the production line is centered on the equipment rather than the consistency of the workpiece cycle time, making it difficult to achieve stable continuous automated processing. Summary of the Invention
[0009] To address the aforementioned problems, this invention provides a cycle balance and collaborative control method and system for an automated crystal processing production line.
[0010] This invention adopts the following technical solution: a cycle time balance and collaborative control method for an automated crystal processing production line, comprising:
[0011] Step S01: Pre-build a machine learning model for predicting processing cycle time, obtain the structural parameters, accuracy constraints and processing steps of the crystal to be processed, input them into the pre-built machine learning model for predicting processing cycle time, and predict the processing cycle time benchmark of the crystal to be processed, which is recorded as the target processing cycle time.
[0012] Step S02: Collect real-time operating status data of each processing step in the production line, and correct the processing cycle benchmark based on the acquired real-time operating status data to obtain the current effective processing cycle of each step;
[0013] Step S03: Based on the current effective processing cycle time of each process, calculate the process cycle time deviation relative to the target processing cycle time, and combine the position weight of the process in the overall production line to classify and identify the process cycle time deviation, thereby determining the key process with the highest impact on the continuous processing rhythm of the entire line.
[0014] The method for determining the key processes that have the greatest impact on the continuous processing rhythm of the entire production line includes:
[0015] Step S031: Calculate the process cycle deviation value based on the current effective processing cycle time and the target processing cycle time for each processing step;
[0016] Step S032: Based on the sequence of each processing step in the production line, normalize the relative position of each processing step to obtain a relative position factor that reflects the structural importance of the process.
[0017] Step S033: Within a preset time window, obtain the transmission relationship of the cycle time change of each processing step in its downstream process, so as to obtain the coupling influence factor reflecting the amplification effect of the processing step on the upstream and downstream cycle time.
[0018] Step S034: Based on the relative position factor and the coupling influence factor, process position weights are generated. The process position weights are multiplied with the corresponding process cycle deviations to obtain the comprehensive influence of process cycle. A set of comprehensive influence of process cycle is constructed, and the cycle corresponding to the maximum value of the comprehensive influence of process cycle is obtained and marked as a key process.
[0019] Step S04: Based on the process cycle deviation value of the key process, obtain the adjacent processes that participate in the coordinated control, and adjust the workpiece input sequence of the adjacent processes in a coordinated manner so that the cycle time of the key process gradually returns to the target processing cycle time.
[0020] Step S05: Adjust the workpiece release interval between the critical process and adjacent processes based on the workpiece release interval adjustment value.
[0021] As a further description of the above technical solution: the operating status data includes the actual processing time of a single piece, the operating status parameters of the cutting tool, and the material flow data before and after the corresponding process.
[0022] As a further description of the above technical solution: the method for obtaining the effective processing cycle time includes:
[0023] The processing time for each workpiece in each process is calibrated, and the actual processing time is calculated.
[0024] The operating status parameters of the tool during the machining process are collected, including load, rotational speed and wear degree, and the operating status coefficient of the tool is obtained based on the operating status parameters of the tool.
[0025] Obtain the number of workpieces transferred in the buffer zones before and after each process, and calculate the material flow status coefficient.
[0026] The process cycle correction factor is calculated based on the actual processing time, operating status coefficient and material flow status coefficient. The correction factor is used to correct the target processing cycle in real time to obtain the current effective processing cycle.
[0027] As a further description of the above technical solution: the method for obtaining adjacent processes participating in coordinated control based on the process cycle deviation value of key processes includes:
[0028] Obtain the cycle time deviation value corresponding to the key process, obtain the cycle time deviation type based on the cycle time deviation value, and determine the adjacent processes that participate in the coordinated control. The cycle time deviation type includes lag deviation and lead deviation.
[0029] When the cycle time deviation value of the process is greater than zero, it indicates that the key process is a cycle time lag type deviation. The upstream process is selected as the main collaborative control object and marked as the adjacent process.
[0030] When the cycle time deviation value of a process is less than zero, it indicates that the critical process has a cycle time lead type deviation. The downstream process is selected as the main object of coordinated control and marked as the adjacent process.
[0031] As a further description of the above technical solution: the method of gradually returning the cycle time of key processes to the target processing cycle time by adjusting the workpiece input sequence of adjacent processes in a coordinated manner includes:
[0032] Based on the determined adjacent processes, the timing of workpiece input between the key process and adjacent processes is adjusted in a coordinated manner. Based on the process cycle deviation value between the current effective processing cycle and the target processing cycle of the key process, the workpiece release time of adjacent processes is dynamically corrected so that the arrival rhythm of workpieces entering the key process matches the processing capacity of the key process, thereby eliminating cycle fluctuations caused by concentrated or intermittent input of workpieces.
[0033] As a further description of the above technical solution: the method for dynamically correcting the workpiece release time of adjacent processes based on the process cycle deviation value between the current effective processing cycle and the target processing cycle of the key process includes: inputting the obtained process cycle deviation value, the effective processing cycle of the key process, the comprehensive influence degree of the process cycle of the key process, and the current workpiece release interval into a pre-constructed workpiece release interval adjustment model, and outputting the predicted workpiece release interval adjustment value.
[0034] As a further description of the above technical solution: the training method of the workpiece release interval adjustment model includes:
[0035] Q sets of training data are collected in advance, where Q is a positive integer greater than 0. The training data includes process cycle deviation value, effective processing cycle of key processes, comprehensive influence of process cycle of key processes, current workpiece release interval, and corresponding workpiece release interval adjustment value.
[0036] Gradient boosting regression tree was selected as the prediction model, and initial hyperparameters were set.
[0037] Mean squared error is used as the loss function to measure the deviation between the model's predicted values and the true labels. Model training is carried out based on training data. The construction of each new tree is aimed at fitting the regression residual of the training set loss function. The optimal splitting feature is selected through the mean squared error criterion to divide the samples into different child nodes until the preset stopping condition is met.
[0038] Gradient descent is used to optimize the weights of the leaf nodes of each new tree. Bayesian optimization is used to search for the optimal combination of hyperparameters within a preset optimization range. The optimization objective is to minimize the root mean square error of the validation set. An early stopping mechanism is introduced during training: the root mean square error of the validation set is calculated every 20 trees. When the root mean square error of the validation set decreases by less than 0.001 after 3 consecutive iterations (60 trees in total), model training is stopped to avoid overfitting the training data. After training, the model parameters with the lowest root mean square error of the validation set are saved.
[0039] As a further description of the above technical solution: the structural parameters of the crystal to be processed include length, width, height, radius of curvature, volume and weight; the precision constraints include surface roughness, angular tolerance and dimensional tolerance.
[0040] A cycle time balance and collaborative control system for an automated crystal processing production line, used to implement the cycle time balance and collaborative control method for the aforementioned automated crystal processing production line, the system comprising:
[0041] The target cycle generation module pre-builds a processing cycle prediction machine learning model, obtains the structural parameters, accuracy constraints and processing steps of the crystal to be processed, inputs them into the pre-built processing cycle prediction machine learning model, and predicts the processing cycle benchmark of the crystal to be processed, which is denoted as the target processing cycle.
[0042] The effective cycle time correction module collects real-time operating status data of each processing step in the production line, corrects the processing cycle time benchmark based on the acquired real-time operating status data, and obtains the current effective processing cycle time of each step.
[0043] The critical process identification module calculates the process cycle deviation relative to the target processing cycle based on the current effective processing cycle of each process, and combines the position weight of the process in the overall production line to classify and identify the process cycle deviation, thereby determining the critical process with the highest impact on the continuous processing rhythm of the entire line.
[0044] The collaborative cycle control module obtains the adjacent processes involved in collaborative control based on the cycle deviation value of the key processes. By adjusting the workpiece input sequence of the adjacent processes in a coordinated manner, the cycle time of the key processes is gradually brought back to the target processing cycle time.
[0045] The workpiece timing execution module adjusts the workpiece release interval between key processes and adjacent processes based on the workpiece release interval adjustment value.
[0046] Beneficial effects:
[0047] This invention compares the effective processing cycle time of each process with the target processing cycle time, and combines this with the structural position of the process in the overall production line and the transmission relationship of cycle time changes in upstream and downstream processes to classify and identify process cycle time deviations, thereby accurately locating the key processes that have the greatest impact on the continuous processing rhythm of the entire line. This method breaks through the traditional control approach centered on single-machine efficiency, achieving accurate identification of the source of cycle time imbalance from the perspective of the entire line, and avoiding secondary imbalances caused by blind adjustments or synchronous intervention across the entire line.
[0048] Furthermore, based on the identification of critical processes, this invention selectively chooses adjacent processes directly related to the critical processes as collaborative control targets according to the type of cycle time deviation. By adjusting the timing of workpiece input and release interval in a coordinated manner, the arrival rhythm of workpieces entering the critical processes matches their actual processing capacity. This control method does not directly change the processing parameters, reducing the impact on processing quality and equipment stability, while effectively eliminating cycle time fluctuations caused by concentrated or intermittent workpiece input. Attached Figure Description
[0049] The present invention will be further explained below with reference to the accompanying drawings and embodiments:
[0050] Figure 1 This is a flowchart of a cycle balance and collaborative control method for an automated crystal processing production line provided in Embodiment 1 of the present invention;
[0051] Figure 2 This is a flowchart of the method for obtaining an effective processing cycle time provided in Embodiment 1 of the present invention;
[0052] Figure 3 This is a flowchart of a method for determining the key process that has the greatest impact on the continuous processing rhythm of the entire production line, as provided in Embodiment 1 of the present invention.
[0053] Figure 4 This is a module connection diagram of a cycle balance and collaborative control system for an automated crystal processing production line provided in Embodiment 2 of the present invention. Detailed Implementation
[0054] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.
[0055] Example 1
[0056] Please see Figures 1-3 This invention provides a technical solution: a method for cycle time balance and coordinated control of an automated crystal processing production line, comprising:
[0057] Step S01: Pre-build a machine learning model for predicting processing cycle time, obtain the structural parameters, accuracy constraints and processing steps of the crystal to be processed, input them into the pre-built machine learning model for predicting processing cycle time, predict the processing cycle time benchmark of the crystal to be processed, and record it as the target processing cycle time, thereby realizing the establishment of a unified time calibration benchmark for all processing steps in the production line.
[0058] Specifically, by establishing a unified processing rhythm benchmark, all subsequent process rhythm adjustments are constrained by this target rhythm, thus avoiding the use of single equipment efficiency as the basis for control from the outset.
[0059] It should be noted that the processing cycle benchmark is the processing time benchmark for each processing step in the production line; the processing cycle refers to the time period corresponding to the completion of processing a single workpiece under continuous processing conditions, provided that the process accuracy and quality requirements are met. The processing cycle is used to characterize the actual contribution of the process to the overall operating rhythm of the production line.
[0060] The structural parameters of the crystal to be processed include length, width, height, radius of curvature, volume, and weight;
[0061] The precision constraints include surface roughness, angular tolerance, and dimensional tolerance.
[0062] Optionally, the processing steps include rough grinding, fine grinding, and polishing, and corresponding numerical labels can be set for each. Optionally, rough grinding, fine grinding, and polishing can be set to 1, 2, and 3 in sequence.
[0063] The processing cycle time references include rough grinding cycle time references, fine grinding cycle time references, polishing cycle time references, and inspection cycle time references.
[0064] The training method for the processing beat prediction machine learning model includes:
[0065] Select workpiece data records that have been normally processed on the production line, obtain the structural parameters, precision constraints, numerical labels of processing steps, and corresponding actual processing cycle benchmarks of the crystal to be processed, and construct a model training dataset.
[0066] The structural parameters, precision constraints, and processing steps of each group of crystals to be processed are integrated into a set of numerical feature vectors. At the same time, the actual value of the processing cycle benchmark corresponding to each set of feature vectors is clearly defined, forming a one-to-one correspondence sample pair between feature vectors and processing cycle benchmarks.
[0067] The dataset was divided using time-series stratified sampling. The samples were divided into training, validation and test sets according to a preset ratio (7:2:1). Preprocessing was performed on the divided datasets to standardize the structural parameters and precision constraints, eliminate differences in units, remove abnormal samples and duplicate records, and fill a small number of missing values using the K-nearest neighbor method to ensure data quality.
[0068] One of the following is selected as the training model: Support Vector Machine Regression, Random Forest Regression, or Neural Network Regression. The feature vectors of the training set are used as the model input, and the actual values of the corresponding processing beat benchmarks are used as the training objectives. The mean squared error is used as the loss function. The model parameters are optimized by minimizing the loss function value. The mean squared error is calculated as the sum of the squares of the differences between the actual processing beat benchmarks and the model predictions of all samples. During the training process, the model performance is monitored in real time using the validation set. After each round of training, the mean squared error and root mean square error of the validation set are calculated. Training is stopped when the root mean square error of the validation set does not decrease significantly for three consecutive iterations to avoid model overfitting. The optimal model parameters at this time are saved.
[0069] The optimal model is evaluated using a test set. The model's performance is judged by three indicators: root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (CCD). The RMSE must be controlled within 5% of the target processing cycle time, the MAE must not exceed 3% of the target processing cycle time, and the CCD must be no less than 0.9 to ensure that the model's prediction accuracy meets the production line scheduling requirements. If the evaluation indicators do not meet the standards, the feature transformation method must be re-optimized or the model parameters adjusted, and training and evaluation must be carried out again until the model performance meets the standards.
[0070] Step S02: Collect real-time operating status data of each processing step in the production line, and correct the processing cycle benchmark based on the acquired real-time operating status data to obtain the current effective processing cycle of each step.
[0071] The operational status data includes the actual processing time for a single piece, the operating status parameters of the cutting tool, and the material flow data before and after the corresponding process.
[0072] The method for obtaining the effective processing cycle time includes:
[0073] The processing time for each workpiece in each process is calibrated, and the actual processing time is calculated.
[0074] The operating status parameters of the tool during the machining process are collected, including load, rotational speed, and wear degree. The operating status coefficient of the tool is obtained based on the operating status parameters of the tool.
[0075] Optionally, the formula for calculating the operating state coefficient of the cutting tool is:
[0076] ;
[0077] In the formula, For the first Operating state coefficient of the cutting tool in the process, For the first Real-time load of the cutting tool in the process, This is the minimum load required for the tool to operate normally. The maximum allowable load for the tool. For the first Current rotation speed of the process, For the ideal machining speed, For the first Current cumulative wear of the process This is the maximum allowable wear of the cutting tool. , and These are the weighting coefficients. , and All are greater than 0. + + =1;
[0078] It should be noted that the wear amount is measured using measuring instruments to measure changes in tool size, the rotational speed is obtained through an encoder, photoelectric speed sensor or spindle control system, and the real-time load is collected in real time by a tool spindle torque sensor.
[0079] Obtain the number of workpieces transferred in the buffer zones before and after each process, and calculate the material flow status coefficient.
[0080] The formula for calculating the material flow state coefficient is as follows:
[0081] ;
[0082] In the formula, This is the material flow state coefficient. For the first The number of workpieces in the pre-process buffer. For the first The number of workpieces in the post-process buffer. For the first The maximum number of workpieces that the buffer area corresponding to the process can hold;
[0083] >0 indicates a large backlog in the preceding process or a sluggish flow in the following process, which may lead to a longer cycle time; <0 indicates that the process is running idle and the cycle time is ahead of schedule.
[0084] The process cycle correction factor is calculated based on the actual processing time, operating status coefficient and material flow status coefficient. The correction factor is used to correct the target processing cycle in real time to obtain the current effective processing cycle.
[0085] Optionally, the formula for calculating the process cycle correction factor is:
[0086] ;
[0087] In the formula, For the first Process cycle time correction factor This is the ratio of actual processing time to the target processing cycle time. This represents the actual time spent. To achieve the target processing cycle, This is the material flow state coefficient. For the first Operating state coefficient of the cutting tool in the process, , and These are the weighting coefficients. , and All are greater than 0. + + =1;
[0088] It should be noted that, >1 indicates that the process cycle time is too slow. <1 indicates that the process cycle is too fast.
[0089] The formula for calculating the real-time correction of the target processing cycle using the correction factor is as follows:
[0090] ;
[0091] In the formula, For the first Effective processing cycle time of the process For the first Process target processing cycle time For the first Process cycle time correction factor.
[0092] Step S03: Based on the current effective processing cycle time of each process, calculate the process cycle time deviation relative to the target processing cycle time, and combine the position weight of the process in the overall production line to classify and identify the process cycle time deviation, thereby determining the key process with the highest impact on the continuous processing rhythm of the entire line, which can provide a clear direction for subsequent cycle time coordination and control.
[0093] The method for determining the key processes that have the greatest impact on the continuous processing rhythm of the entire production line includes:
[0094] Step S031: Calculate the process cycle deviation value based on the current effective processing cycle time and the target processing cycle time for each processing step;
[0095] The formula for calculating the cycle time deviation of the process is:
[0096] ;
[0097] In the formula, For the first Process cycle time deviation value of the process. For the first Effective processing cycle time of the process For the first The target processing cycle time for each process; it should be noted that when... A value greater than 0 indicates that the cycle time of this process is lagging; when... When the value is less than 0, it indicates that the cycle time of this process is ahead of schedule.
[0098] Step S032: Based on the sequence of each processing step in the production line, normalize the relative position of each processing step to obtain a relative position factor that reflects the structural importance of the process.
[0099] Optionally, the formula for calculating the relative position factor is: In the formula, N is the total number of processes on the production line. For the first The number of processes arranged in sequence For the first The relative position factor of the process in the production line.
[0100] Specifically, this method of obtaining the relative position factor, which reflects the structural importance of the process, indicates that the closer the process is to the end of the production line, the greater the direct impact of its rhythm abnormalities on the final continuous output. This is especially true in the processing of crystal / zoom lenses. For example, in the early stages of the process, such as rough grinding, even if it is a little slow, it can be digested through buffering and subsequent adjustments. In the middle stages, such as fine grinding, it begins to have a significant impact on the overall rhythm. In the later stages, such as polishing, once there is an abnormality, it will cause the finished product to be directly stuck.
[0101] Step S033: Within a preset time window, obtain the transmission relationship of the cycle time change of each processing step in its downstream process, so as to obtain the coupling influence factor reflecting the amplification effect of the processing step on the upstream and downstream cycle time.
[0102] Optionally, the method for obtaining the coupling influence factor includes:
[0103] Within the continuous running time window, the data were collected respectively. Process and downstream The current real-time processing cycle time of +1 process is calculated, and the change in cycle time is determined.
[0104] The coupling influence factor is calculated based on the change in beat rate. The formula for calculating the coupling influence factor is as follows: In the formula, For the first The coupling effect factor of the process on the downstream cycle time. For the first The change in cycle time of the process within a preset time window. For downstream +1 The change in cycle time within the same preset time window. To prevent extremely small positive numbers with a denominator of zero. Among them,
[0105] ;
[0106] ;
[0107] In the formula, For time Time of the first The current processing cycle time of the process, For time Time of the first The current processing cycle time of the process, For time Time of the first The current processing cycle time of the process, For time Time of the first The current processing cycle of the process.
[0108] Step S034: Based on the relative position factor and the coupling influence factor, process position weights are generated. The process position weights are multiplied by the corresponding cycle time deviations to obtain the comprehensive influence degree of the process cycle time. A set of comprehensive influence degrees of process cycle time is constructed, and the cycle time corresponding to the maximum value of the comprehensive influence degree of process cycle time is obtained and marked as a key process.
[0109] Optionally, the formula for calculating the process position weight is:
[0110] ;
[0111] In the formula, For the first Position weight of the process For the first The relative position factor of the process in the production line. For the first The coupling effect factor of the process on the downstream cycle time. and These are the weighting coefficients. and All are greater than 0. + =1.
[0112] Optionally, the formula for calculating the comprehensive impact of the process cycle time is: In the formula, For the first Overall impact of process cycle time For the first Position weight of the process For the first The process cycle time deviation value of the process.
[0113] Step S04: Based on the process cycle deviation value of the key process, obtain the adjacent processes that participate in the coordinated control, and adjust the workpiece input sequence of the adjacent processes in a coordinated manner so that the cycle time of the key process gradually returns to the target processing cycle time.
[0114] Methods for obtaining adjacent processes involved in coordinated control based on the process cycle time deviation value of key processes include:
[0115] Obtain the cycle time deviation value corresponding to the key process, obtain the cycle time deviation type based on the cycle time deviation value, and determine the adjacent processes that participate in the coordinated control. The cycle time deviation type includes lag deviation and lead deviation.
[0116] When the cycle time deviation value of the process is greater than zero, it indicates that the key process is a cycle time lag type deviation. The upstream process is selected as the main collaborative control object and marked as the adjacent process.
[0117] When the cycle time deviation value of a process is less than zero, it indicates that the critical process has a cycle time lead type deviation. The downstream process is selected as the main object of coordinated control and marked as the adjacent process.
[0118] Methods for gradually returning the cycle time of critical processes to the target processing cycle time by adjusting the workpiece input sequence of adjacent processes in a coordinated manner include:
[0119] Based on the determined adjacent processes, the timing of workpiece input between the key process and adjacent processes is adjusted in a coordinated manner. Based on the cycle deviation between the current effective processing cycle and the target processing cycle of the key process, the release time of workpieces in adjacent processes is dynamically corrected so that the arrival rhythm of workpieces entering the key process matches the processing capacity of the key process, thereby eliminating cycle fluctuations caused by concentrated or intermittent input of workpieces.
[0120] Specifically, the method for dynamically correcting the workpiece release time of adjacent processes based on the cycle deviation value between the current effective processing cycle and the target processing cycle of the key process includes: inputting the obtained cycle deviation value, the effective processing cycle of the key process, the comprehensive influence degree of the cycle of the key process, and the current workpiece release interval into a pre-constructed workpiece release interval adjustment model, and outputting the predicted workpiece release interval adjustment value.
[0121] The training method for the workpiece release interval adjustment model includes:
[0122] Q sets of training data are collected in advance, where Q is a positive integer greater than 0. The training data includes cycle time deviation value, effective processing cycle time of key processes, comprehensive influence of cycle time of key processes, current workpiece release interval, and corresponding workpiece release interval adjustment value.
[0123] Gradient boosting regression tree was selected as the prediction model, and initial hyperparameters were set. The initial hyperparameters specifically include: number of decision trees 100-150, maximum depth of a single tree 4-6, minimum number of samples for node splitting 8-12, maximum number of features considered during splitting 3, learning rate 0.05-0.1, and regularization coefficient (L2) 0.1-0.2.
[0124] During the model training phase, mean squared error is used as the loss function to measure the degree of deviation between the model's predicted values and the true labels. Model training is carried out based on training data. The construction of each new tree aims to fit the regression residual of the training set loss function. The optimal splitting feature is selected through the mean squared error criterion, such as prioritizing core features that are strongly correlated with the prediction target. The samples are divided into different child nodes until the preset stopping conditions are met, such as reaching the preset maximum depth, the number of child node samples being less than the minimum number of samples, or the splitting gain being lower than the threshold.
[0125] The gradient descent method is used to optimize the weights of the leaf nodes of each new tree. The contribution weight of the new tree to the final prediction result is adjusted by the learning rate to avoid a single tree dominating the prediction process and to ensure the stability of the model prediction.
[0126] In the hyperparameter optimization phase, a Bayesian optimization method is employed to search for the optimal combination of hyperparameters within a pre-defined optimization range. The optimization objective is to minimize the root mean square error of the validation set. Specifically, the hyperparameter optimization range is as follows: number of decision trees 80-180, maximum depth of a single tree 3-7, learning rate 0.03-0.12, and regularization coefficient (L2) 0.05-0.25.
[0127] An early stopping mechanism is introduced during training. The root mean square error (RMSE) of the validation set is calculated every 20 trees. Training is stopped when the RMSE of the validation set decreases by less than 0.001 after three consecutive iterations (60 trees in total) to prevent overfitting of the training data. After training, the model parameters with the lowest RMSE of the validation set are saved, including the splitting rules of all decision trees and the weights of the leaf nodes, ensuring the model still has stable predictive ability for unseen new samples.
[0128] The trained model is validated using a test set. The root mean square error, mean absolute error, and coefficient of determination (R²) are calculated. If the root mean square error of the test set is ≤ 5% of the target predicted value, the mean absolute error is ≤ 3% of the target predicted value, and the coefficient of determination (R²) is ≥ 0.9, then the model performance evaluation meets the standards and can be deployed in practice.
[0129] Step S05: Adjust the workpiece release interval between the critical process and adjacent processes based on the workpiece release interval adjustment value.
[0130] Specifically, since both equipment idling and production blockage are caused by the mismatch between the arrival rhythm of workpieces and the actual processing capacity of the process, the workpiece release interval of adjacent processes is dynamically adjusted based on the effective processing rhythm of key processes to keep the arrival rhythm of workpieces consistent with the processing capacity of key processes. This reduces the conditions for equipment to wait for workpieces or for workpieces to accumulate, thereby reducing the probability of equipment idling and production blockage.
[0131] This invention realizes cycle balance and collaborative control guided by workpiece cycle consistency. It can improve the continuous operation capability and overall efficiency of zoom lens and optical crystal component processing production lines while ensuring processing quality, and significantly reduce the probability of equipment idling and production blockage. It is suitable for automated production sites for high-precision, multi-specification crystal processing.
[0132] Example 2
[0133] Please see Figure 4 This invention provides a technical solution: a cycle time balance and collaborative control system for an automated crystal processing production line, which is used to implement the cycle time balance and collaborative control method for the automated crystal processing production line. The system includes:
[0134] The target cycle generation module pre-builds a processing cycle prediction machine learning model, obtains the structural parameters, accuracy constraints and processing steps of the crystal to be processed, inputs them into the pre-built processing cycle prediction machine learning model, and predicts the processing cycle benchmark of the crystal to be processed, which is denoted as the target processing cycle.
[0135] The effective cycle time correction module collects real-time operating status data of each processing step in the production line, corrects the processing cycle time benchmark based on the acquired real-time operating status data, and obtains the current effective processing cycle time of each step.
[0136] The critical process identification module calculates the process cycle deviation relative to the target processing cycle based on the current effective processing cycle of each process, and combines the position weight of the process in the overall production line to classify and identify the process cycle deviation, thereby determining the critical process with the highest impact on the continuous processing rhythm of the entire line.
[0137] The collaborative cycle control module obtains the adjacent processes involved in collaborative control based on the cycle deviation value of the key processes. By adjusting the workpiece input sequence of the adjacent processes in a coordinated manner, the cycle time of the key processes is gradually brought back to the target processing cycle time.
[0138] The workpiece timing execution module adjusts the workpiece release interval between key processes and adjacent processes based on the workpiece release interval adjustment value.
[0139] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.
Claims
1. A method for cycle time balance and coordinated control of an automated crystal processing production line, characterized in that, include: Step S01: Pre-build a machine learning model for predicting processing cycle time, obtain the structural parameters, accuracy constraints and processing steps of the crystal to be processed, input them into the pre-built machine learning model for predicting processing cycle time, and predict the processing cycle time benchmark of the crystal to be processed, which is recorded as the target processing cycle time. Step S02: Collect real-time operating status data of each processing step in the production line, and correct the processing cycle benchmark based on the acquired real-time operating status data to obtain the current effective processing cycle of each step; Step S03: Based on the current effective processing cycle time of each process, calculate the process cycle time deviation relative to the target processing cycle time, and combine the position weight of the process in the overall production line to classify and identify the process cycle time deviation, thereby determining the key process with the highest impact on the continuous processing rhythm of the entire line. The method for determining the key processes that have the greatest impact on the continuous processing rhythm of the entire production line includes: Step S031: Calculate the process cycle deviation value based on the current effective processing cycle time and the target processing cycle time for each processing step; Step S032: Based on the sequence of each processing step in the production line, normalize the relative position of each processing step to obtain a relative position factor that reflects the structural importance of the process. Step S033: Within a preset time window, obtain the transmission relationship of the cycle time change of each processing step in its downstream process, so as to obtain the coupling influence factor reflecting the amplification effect of the processing step on the upstream and downstream cycle time. Step S034: Based on the relative position factor and the coupling influence factor, process position weights are generated. The process position weights are multiplied with the corresponding process cycle deviations to obtain the comprehensive influence of process cycle. A set of comprehensive influence of process cycle is constructed, and the cycle corresponding to the maximum value of the comprehensive influence of process cycle is obtained and marked as a key process. Step S04: Based on the process cycle deviation value of the key process, obtain the adjacent processes that participate in the coordinated control, and adjust the workpiece input sequence of the adjacent processes in a coordinated manner so that the cycle time of the key process gradually returns to the target processing cycle time. Step S05: Adjust the workpiece release interval between the critical process and adjacent processes based on the workpiece release interval adjustment value.
2. The cycle balance and collaborative control method for an automated crystal processing production line according to claim 1, characterized in that, The operational status data includes the actual processing time for a single piece, the operating status parameters of the cutting tool, and the material flow data before and after the corresponding process.
3. The cycle balance and collaborative control method for an automated crystal processing production line according to claim 2, characterized in that, The method for obtaining the effective processing cycle time includes: The processing time for each workpiece in each process is calibrated, and the actual processing time is calculated. The operating status parameters of the tool during the machining process are collected, including load, rotational speed and wear degree, and the operating status coefficient of the tool is obtained based on the operating status parameters of the tool. Obtain the number of workpieces transferred in the buffer zones before and after each process, and calculate the material flow status coefficient. The process cycle correction factor is calculated based on the actual processing time, operating status coefficient and material flow status coefficient. The correction factor is used to correct the target processing cycle in real time to obtain the current effective processing cycle.
4. The cycle balance and collaborative control method for an automated crystal processing production line according to claim 1, characterized in that, The method for obtaining adjacent processes participating in coordinated control based on the process cycle deviation value of key processes includes: Obtain the cycle time deviation value corresponding to the key process, obtain the cycle time deviation type based on the cycle time deviation value, and determine the adjacent processes that participate in the coordinated control. The cycle time deviation type includes lag deviation and lead deviation. When the cycle time deviation value of the process is greater than zero, it indicates that the key process is a cycle time lag type deviation. The upstream process is selected as the main collaborative control object and marked as the adjacent process. When the cycle time deviation value of a process is less than zero, it indicates that the critical process has a cycle time lead type deviation. The downstream process is selected as the main object of coordinated control and marked as the adjacent process.
5. The cycle balance and collaborative control method for an automated crystal processing production line according to claim 1, characterized in that, Methods for gradually returning the cycle time of critical processes to the target processing cycle time by adjusting the workpiece input sequence of adjacent processes in a coordinated manner include: Based on the determined adjacent processes, the timing of workpiece input between the key process and adjacent processes is adjusted in a coordinated manner. Based on the process cycle deviation value between the current effective processing cycle and the target processing cycle of the key process, the workpiece release time of adjacent processes is dynamically corrected so that the arrival rhythm of workpieces entering the key process matches the processing capacity of the key process, thereby eliminating cycle fluctuations caused by concentrated or intermittent input of workpieces.
6. The cycle balance and collaborative control method for an automated crystal processing production line according to claim 5, characterized in that, The method for dynamically correcting the workpiece release time of adjacent processes based on the process cycle deviation value between the current effective processing cycle and the target processing cycle of the key process includes: inputting the obtained process cycle deviation value, the effective processing cycle of the key process, the comprehensive influence degree of the process cycle of the key process, and the current workpiece release interval into a pre-constructed workpiece release interval adjustment model, and outputting the predicted workpiece release interval adjustment value.
7. The cycle balance and collaborative control method for an automated crystal processing production line according to claim 6, characterized in that, The training method for the workpiece release interval adjustment model includes: Q sets of training data are collected in advance, where Q is a positive integer greater than 0. The training data includes process cycle deviation value, effective processing cycle of key processes, comprehensive influence of process cycle of key processes, current workpiece release interval, and corresponding workpiece release interval adjustment value. Gradient boosting regression tree was selected as the prediction model, and initial hyperparameters were set. Mean squared error is used as the loss function to measure the deviation between the model's predicted values and the true labels. Model training is carried out based on training data. The construction of each new tree is aimed at fitting the regression residual of the training set loss function. The optimal splitting feature is selected through the mean squared error criterion to divide the samples into different child nodes until the preset stopping condition is met. Gradient descent is used to optimize the weights of the leaf nodes of each new tree. Bayesian optimization is used to search for the optimal combination of hyperparameters within a preset optimization range. The optimization objective is to minimize the root mean square error of the validation set. An early stopping mechanism is introduced during training: the root mean square error of the validation set is calculated every 20 trees. When the root mean square error of the validation set decreases by less than 0.001 after 3 consecutive iterations (60 trees in total), model training is stopped to avoid overfitting the training data. After training, the model parameters with the lowest root mean square error of the validation set are saved.
8. The cycle balance and collaborative control method for an automated crystal processing production line according to claim 1, characterized in that, The structural parameters of the crystal to be processed include length, width, height, radius of curvature, volume, and weight; the precision constraints include surface roughness, angular tolerance, and dimensional tolerance.
9. A cycle balance and collaborative control system for an automated crystal processing production line, used to implement the cycle balance and collaborative control method for an automated crystal processing production line according to any one of claims 1-8, characterized in that, The system includes: The target cycle generation module pre-builds a processing cycle prediction machine learning model, obtains the structural parameters, accuracy constraints and processing steps of the crystal to be processed, inputs them into the pre-built processing cycle prediction machine learning model, and predicts the processing cycle benchmark of the crystal to be processed, which is denoted as the target processing cycle. The effective cycle time correction module collects real-time operating status data of each processing step in the production line, corrects the processing cycle time benchmark based on the acquired real-time operating status data, and obtains the current effective processing cycle time of each step. The critical process identification module calculates the process cycle deviation relative to the target processing cycle based on the current effective processing cycle of each process, and combines the position weight of the process in the overall production line to classify and identify the process cycle deviation, thereby determining the critical process with the highest impact on the continuous processing rhythm of the entire line. The collaborative cycle control module obtains the adjacent processes involved in collaborative control based on the cycle deviation value of the key process. By adjusting the workpiece input sequence of the adjacent processes in a coordinated manner, the cycle time of the key process gradually returns to the target processing cycle time. The workpiece timing execution module adjusts the workpiece release interval between key processes and adjacent processes based on the workpiece release interval adjustment value.