Intelligent scheduling and tracking method and tracking system

Abstract

The present invention discloses an intelligent scheduling and tracking method and tracking system, which specifically relates to the field of data management technology. By setting a comparison between ideal production status data and actual production data, the changes in production data are monitored in real time. By calculating the deviation degree and coupling degree anomaly index, the failure risk of the intelligent scheduling system can be effectively identified, and the system is divided into different adaptability levels according to the adaptability evaluation results. For high adaptability levels, the resource allocation strategy is further optimized; for low adaptability levels, failure decisions are responded to and repaired in a timely manner to reduce the impact on production; for medium adaptability levels, the system's adaptability to data distribution drift is improved by predicting the degree of anomaly and continuously optimizing scheduling decisions, thereby ensuring the balanced operation of the production line, maximizing production capacity and avoiding stagnation. It can not only dynamically adapt to changes in the industrial production environment, but also improve the accuracy of the intelligent scheduling system and ensure the stability and reliability of production efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of data management, and in particular to an intelligent scheduling and tracking method and a tracking system. Background Art

[0002] Intelligent scheduling and tracking refers to the use of technologies such as artificial intelligence (AI), big data, and the Internet of Things (IoT) to optimize resource allocation and location tracking. Through algorithms and real-time data analysis, this approach improves system efficiency, reduces resource waste, and enables precise task allocation and dynamic adjustments. Intelligent scheduling is often used to solve resource allocation problems in complex systems, such as prioritizing personnel, vehicles, equipment, or tasks. Tracking methods use sensors and positioning technologies to monitor the location and status of these resources for visualization and real-time management.

[0003] In industrial production, intelligent scheduling systems leverage artificial intelligence (AI) to monitor and analyze the operating status of production line equipment in real time, optimizing resource allocation and worker scheduling strategies to maximize production efficiency. For example, if a piece of equipment experiences excessive load or an anomaly, the system can automatically dispatch other equipment to take over, avoiding production interruptions. Furthermore, sensors and IoT technologies enable precise tracking of critical equipment, providing real-time feedback on operating status and maintenance needs. This helps companies promptly identify potential failures and implement preventative measures, effectively reducing downtime and maintenance costs.

[0004] The existing technology has the following shortcomings:

[0005] The algorithms of intelligent scheduling systems typically rely on historical data and trained models. However, the industrial production environment can change over time, due to factors such as equipment aging, fluctuating production demands, and the introduction of new processes, causing data distribution to drift. If the model fails to dynamically adapt to these changes, the system's scheduling decisions will quickly become ineffective, potentially leading to decreased production efficiency and even system disruption. Furthermore, if the scheduling system's decisions fail, resource allocation can be irrational, resulting in some equipment being overloaded while others remain idle, leading to unbalanced production line operations. Critical tasks may not be completed on time, overall production capacity may be significantly reduced, and production may even stagnate. Summary of the Invention

[0006] The purpose of the present invention is to provide an intelligent scheduling and tracking method and tracking system to solve the shortcomings of the background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: an intelligent scheduling and tracking method, comprising the following steps:

[0008] S1: Set the ideal production status data in the production process as the target production data, and collect real-time production data in several time periods during the actual production process. Compare the real-time production data with the target production data, and divide the real-time production data into those that deviate from the target and those that do not deviate from the target;

[0009] S2: Analyze the distribution changes of production data that deviate from the target in each time period, select the production data distribution in the initial time period of the intelligent scheduling system as the benchmark distribution, and determine the degree of deviation between the production data distribution in the current time period and the benchmark distribution;

[0010] S3: If the deviation is high, collect historical data on the scheduling predictions of the intelligent scheduling system model and actual equipment utilization, calculate the coupling degree after performing regression analysis, and analyze abnormal changes in the coupling degree under different drift states to assess the impact of failure of the intelligent scheduling system model on production targets.

[0011] S4: Evaluate the adaptability of the intelligent scheduling system model to data distribution drift based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on production targets when the intelligent scheduling system model fails;

[0012] S5: Based on the evaluation results, the adaptability of the intelligent scheduling system model to data distribution drift is divided into high, medium, and low adaptability levels. For the high adaptability level, the resource allocation strategy is further refined. For the low adaptability level, failed decisions are quickly responded to and repaired to reduce the impact on production.

[0013] S6: For the medium adaptability level, the abnormal degree of the intelligent scheduling system model's adaptability to data distribution drift within a fixed time period is predicted. If the abnormal degree is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions.

[0014] Preferably, in S1, the real-time data and the target data are compared within the same time window, and the deviation between the actual production data and the target production data is calculated, and the expression is: ; Compare the deviation between the actual production data and the target production data with the deviation reference threshold under normal production conditions set based on historical data. If the deviation between the actual production data and the target production data is greater than the deviation reference threshold, it is classified as deviating from the target production data; if the deviation between the actual production data and the target production data is less than or equal to the deviation reference threshold, it is classified as not deviating from the target production data.

[0015] Preferably, in S2, the deviation degree between the production data distribution and the reference distribution in the current time period is analyzed to generate a reference distribution deviation index, and the reference distribution deviation index is obtained by:

[0016] Set the baseline distribution and current distribution. The baseline distribution P is the ideal production status data distribution, and the current distribution Q is the distribution of production data collected in real time during the current time period. Calculate the KL divergence between the baseline distribution and the current time period distribution. The calculation formula is: ;in: is the KL divergence between the benchmark distribution and the distribution of the current time period; P(i) is the probability mass of the i-th class in the benchmark distribution P; Q(i) is the probability mass of the i-th class in the current distribution Q; the calculated KL divergence between the benchmark distribution and the distribution of the current time period is used as the benchmark distribution deviation index.

[0017] Preferably, in S3, after analyzing the abnormal changes in the coupling degree under different drift states, a coupling degree abnormality index is generated. The coupling degree abnormality index is obtained by:

[0018] Construct historical coupling degree series respectively , and construct coupling degree sequence ; First, construct an n×m distance matrix, where n and m are the lengths of the historical coupling sequence and the current coupling sequence respectively; at each element position (i, j), calculate the current time point and The distance between them, using Euclidean distance , the expression is: ; The cumulative distance matrix is ​​calculated by dynamic programming. Each element D(i,j) of the cumulative distance matrix represents the minimum cumulative distance from the starting point to the position (i,j). The calculation formula is: Where, Represents paths in three directions: vertical, horizontal, and diagonal. By selecting the shortest path to accumulate the minimum distance, the shortest path DTW distance is calculated. The DTW distance is the element D(n,m) in the lower right corner of the matrix, that is, the minimum cumulative distance from the historical data sequence to the current data sequence.

[0019] Normalize the DTW distance to get a standardized DTW distance value , the expression is: ; The range of the DTW distance in normal mode is the mean and standard deviation of the DTW distances of multiple normal periods calculated through historical data. The mean of the DTW distance in the historical normal mode is set to , the standard deviation is ; According to the difference between the current DTW distance and the historical mean, the coupling anomaly index is calculated. The formula is: ; Where Gt is the coupling anomaly index.

[0020] Preferably, in S4, the adaptability of the intelligent scheduling system model to data distribution drift is evaluated based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on the production target when the intelligent scheduling system model fails, specifically:

[0021] The baseline distribution deviation index and the coupling degree anomaly index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the adaptability value label of the intelligent scheduling system model to the data distribution drift as the prediction target, and takes minimizing the sum of the prediction errors of the adaptability value labels of all intelligent scheduling system models to the data distribution drift as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The adaptability value of the intelligent scheduling system model to the data distribution drift is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0022] Preferably, in S5, the adaptability of the intelligent scheduling system model to data distribution drift is divided into a high adaptability level, a medium adaptability level, and a low adaptability level, specifically:

[0023] Comparing the obtained adaptability value of the intelligent scheduling system model to data distribution drift with the gradient standard threshold, where the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. Comparing the adaptability value of the intelligent scheduling system model to data distribution drift with the first standard threshold and the second standard threshold respectively;

[0024] If the adaptability value of the intelligent scheduling system model to data distribution drift is greater than the second standard threshold, it means that the intelligent scheduling system model has good adaptability to data distribution drift. At this time, a high adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a high adaptability level.

[0025] If the adaptability value of the intelligent scheduling system model to data distribution drift is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it means that the adaptability of the intelligent scheduling system model to data distribution drift is average. In this case, a medium adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a medium adaptability level.

[0026] If the adaptability value of the intelligent scheduling system model to data distribution drift is less than the first standard threshold, it means that the intelligent scheduling system model has poor adaptability to data distribution drift. At this time, a low adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a low adaptability level.

[0027] Preferably, in S6, for the medium adaptability level, the abnormal degree of adaptability of the intelligent scheduling system model to data distribution drift within a fixed time period is predicted, specifically:

[0028] For the medium adaptability level, that is, the adaptability value of the intelligent scheduling system model to data distribution drift generated within a fixed time period is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the adaptability values ​​generated within the subsequent fixed time period that are greater than or equal to the first standard threshold and less than or equal to the second standard threshold are collected, and the corresponding data set is established, and the standard deviation of the data set is calculated as the adaptability abnormality degree value;

[0029] The obtained adaptability abnormality value is compared with the abnormality reference threshold value pre-set according to historical data. If the adaptability abnormality value is greater than or equal to the pre-set abnormality reference threshold value, it means that the abnormality of the adaptability of the intelligent scheduling system model to the data distribution drift within the fixed time period is high. At this time, an early warning signal is generated and the intelligent scheduling system is optimized; if the adaptability abnormality value is less than the pre-set abnormality reference threshold value, it means that the abnormality of the adaptability of the intelligent scheduling system model to the data distribution drift within the fixed time period is low. At this time, no early warning signal is generated and no additional processing is required.

[0030] Preferably, if the abnormality level is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions, specifically:

[0031] Set the scheduling decision of the current model through a function To express it, the intelligent scheduling system optimizes the decision by adjusting the model parameters θ: ;in: is a scheduling decision based on the current model prediction, is the actual production output result, is the regularization parameter, AE is the abnormality value, and s is the total number of samples;

[0032] According to the optimized objective function, the model parameters θ are adjusted and the new parameters are used to predict the scheduling decision. The optimization process is updated based on the gradient descent method: ; where η is the learning rate, ∇θ is the gradient of the objective function with respect to the model parameters, are the model parameters used in the tth iteration of the intelligent scheduling system, For the In the iteration, the new model parameters are optimized.

[0033] The present invention also provides an intelligent scheduling and tracking system, including a data comparison module, a distribution detection module, a coupling analysis module, an adaptability evaluation module, an adaptability level division module, and an optimization module:

[0034] Data comparison module: Sets the ideal production status data in the production process as the target production data, and collects real-time production data in several time periods during the actual production process. The real-time production data is compared with the target production data, and the real-time production data is divided into those that deviate from the target and those that do not deviate from the target.

[0035] Distribution Detection Module: Analyzes the distribution changes of production data that deviates from the target in each time period, selects the production data distribution of the initial time period of the intelligent scheduling system operation as the benchmark distribution, and determines the degree of deviation between the production data distribution in the current time period and the benchmark distribution;

[0036] Coupling analysis module: If the degree of deviation is high, historical data on the scheduling prediction value of the intelligent scheduling system model and the actual equipment utilization rate are collected, and the coupling degree is calculated after using regression analysis. The abnormal changes in the coupling degree under different drift states are analyzed to evaluate the impact of the failure of the intelligent scheduling system model on production targets.

[0037] Adaptability evaluation module: This module evaluates the adaptability of the intelligent scheduling system model to data distribution drift based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on production targets when the intelligent scheduling system model fails.

[0038] Adaptability level classification module: Based on the evaluation results, the adaptability of the intelligent scheduling system model to data distribution drift is divided into high, medium, and low adaptability levels. For the high adaptability level, the resource allocation strategy is further refined. For the low adaptability level, failed decisions are quickly responded to and repaired to reduce the impact on production.

[0039] Optimization module: For the medium adaptability level, the abnormal degree of the intelligent scheduling system model's adaptability to data distribution drift within a fixed time period is predicted. If the abnormal degree is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions.

[0040] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0041] 1. The present invention can promptly identify adaptability issues of the intelligent scheduling system in the production environment by real-time monitoring and analysis of production data deviations and data distribution drift. Based on the KL divergence calculation between the baseline distribution and the current data distribution, coupling degree anomaly analysis, and comprehensive feature vector input machine learning model for adaptability prediction, the scheduling system can dynamically adapt to data drift caused by factors such as equipment aging, production demand fluctuations, and process changes. By classifying adaptability into high, medium, and low levels and taking corresponding resource optimization and decision-making repair measures, the present invention can effectively improve the robustness of the scheduling system and ensure the smooth operation of the production line.

[0042] 2. This invention targets medium adaptability levels by predicting and optimizing the degree of adaptability anomalies to continuously improve the accuracy of scheduling decisions. By monitoring and implementing an early warning mechanism for adaptability anomaly values, it can promptly issue optimization signals when system adaptability decreases. By adjusting model parameters for optimization, it further improves production efficiency, avoids equipment overloads, production line imbalances, and other issues, and thus reduces the risk of production stagnation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0044] Figure 1 Flow chart of the method of the present invention.

[0045] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0047] Example 1, please refer to Figure 1 and Figure 2 As shown, the intelligent scheduling and tracking method described in this embodiment includes the following steps:

[0048] S1: Set the ideal production status data in the production process as the target production data, and collect real-time production data in several time periods during the actual production process. Compare the real-time production data with the target production data, and divide the real-time production data into those that deviate from the target and those that do not deviate from the target;

[0049] S2: Analyze the distribution changes of production data that deviate from the target in each time period, select the production data distribution in the initial time period of the intelligent scheduling system as the benchmark distribution, and determine the degree of deviation between the production data distribution in the current time period and the benchmark distribution;

[0050] S3: If the deviation is high, collect historical data on the scheduling predictions of the intelligent scheduling system model and actual equipment utilization, calculate the coupling degree after performing regression analysis, and analyze abnormal changes in the coupling degree under different drift states to assess the impact of failure of the intelligent scheduling system model on production targets.

[0051] S4: Evaluate the adaptability of the intelligent scheduling system model to data distribution drift based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on production targets when the intelligent scheduling system model fails;

[0052] S5: Based on the evaluation results, the adaptability of the intelligent scheduling system model to data distribution drift is divided into high, medium, and low adaptability levels. For the high adaptability level, the resource allocation strategy is further refined. For the low adaptability level, failed decisions are quickly responded to and repaired to reduce the impact on production.

[0053] S6: For the medium adaptability level, the abnormal degree of the intelligent scheduling system model's adaptability to data distribution drift within a fixed time period is predicted. If the abnormal degree is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions.

[0054] In S1, the ideal production state data in the production process is set as the target production data, and real-time production data in several time periods are collected in real time during the actual production process. The real-time production data is compared with the target production data, and the real-time production data is divided into production data that deviates from the target and production data that does not deviate from the target. Specifically:

[0055] Target production data is standard data defined based on ideal or optimal production conditions, including: Production rate (production line capacity): the output per unit time in the ideal production process. Equipment status (utilization rate, failure rate, etc.): the normal operating time and maintenance cycle of the equipment when it is in the ideal working condition. Worker efficiency (worker scheduling and production efficiency): the work arrangement of workers under ideal conditions based on production needs. Resource utilization: the rational allocation and use of resources such as raw materials, equipment, and energy. This part of the data can be obtained through long-term production experience, historical data, or production simulation, reflecting the standardization requirements and the expected efficient operation model in the production process.

[0056] Actual production data is collected in real time through sensors, manufacturing execution systems (MES), and other intelligent devices. This data primarily includes: Equipment operating status: real-time equipment load, operating hours, fault alarm information, etc. Production data: actual production output, production rate, production quality, etc. Worker and operation status: worker operating efficiency, shift schedules, working hours, etc. Resource usage data: consumption of various resources (such as raw materials and energy), inventory levels, etc. This data is collected through Industrial Internet of Things (IoT) systems, real-time monitoring systems, or sensor networks, enabling comprehensive monitoring of the production process.

[0057] Once the actual production data is collected, it needs to be compared with the target production data. The main steps of the comparison include:

[0058] Compare real-time and target data within the same time window to ensure data synchronization. Typically, production processes can be divided into hours, shifts, or other reasonable time periods. Remove noisy or missing data and normalize the data to ensure comparability across different sources.

[0059] The deviation between the actual production data and the target production data is calculated using difference indicators (such as mean square error, percentage error, etc.). The expression is: ; Compare the deviations between different production indicators (such as output, equipment load, etc.) and evaluate the execution effect of various production activities.

[0060] If the deviation between the actual production data and the target production data exceeds the set allowable range (for example, the deviation is greater than a certain threshold), it is classified as deviating from the target production data. Otherwise, it is classified as not deviating from the target production data, indicating that the actual production status is relatively close to the target production status.

[0061] Based on historical data, production experience, and the fluctuation range of the production line, a reasonable tolerance range is set (for example, a deviation rate exceeding 5% or 10% is considered off target). This threshold can be dynamically adjusted based on the characteristics of different factories, production lines, and production links.

[0062] The deviation between the actual production data and the target production data is compared with the deviation reference threshold under normal production conditions set based on historical data. If the deviation between the actual production data and the target production data is greater than the deviation reference threshold, it is classified as deviating from the target production data; if the deviation between the actual production data and the target production data is less than or equal to the deviation reference threshold, it is classified as not deviating from the target production data.

[0063] S2: Analyze the distribution changes of the production data that deviates from the target in each time period, select the production data distribution of the initial time period of the intelligent scheduling system as the benchmark distribution, and determine the degree of deviation between the production data distribution in the current time period and the benchmark distribution.

[0064] The baseline distribution refers to the distribution of production data during the initial stages of the intelligent scheduling system's operation (typically, when the system is operating stably). This initial data is chosen as the baseline because production conditions at this stage are typically relatively stable and ideal, and in the absence of external interference, the production process can maintain optimal performance.

[0065] The baseline distribution should include: Production volume within a time period: Output per time period (e.g., per hour, per shift), equipment utilization rate, etc. Equipment utilization: Data such as equipment load, idle time, and fault conditions. Production process data: This includes production-related process parameters such as temperature, humidity, and machine speed. Worker operating status: Worker efficiency, scheduling, and task execution.

[0066] During each time period, the intelligent scheduling system collects production data in real time and compares it with the baseline distribution. The following analysis is required for the real-time production data:

[0067] Histogram comparison: Generate a histogram of actual production data and benchmark production data to check whether the shape of the data distribution has changed. This mainly checks whether there are significant deviations in the central tendency, distribution width, kurtosis, etc. of the data.

[0068] Distribution fitting (e.g., Gaussian, Poisson, etc.): Statistically fit the data to determine whether the actual data distribution still conforms to the initial distribution model. Model fitting is performed using maximum likelihood estimation (MLE) or least squares methods.

[0069] Distance metrics (such as KL divergence or JS divergence): By calculating the difference between the actual data distribution and the benchmark distribution, KL divergence (Kullback-Leibler Divergence) or JS divergence (Jensen-Shannon Divergence) is used to measure the similarity or difference of the distribution.

[0070] Sliding Window Method: A time window of a certain length (e.g., one hour, one shift) is selected as the analysis range for real-time data. As time passes, the window slides, gradually updating the distribution of the real-time data for comparison with the baseline distribution.

[0071] Fixed time period comparison: For certain time periods in the production process (such as peak production periods or equipment maintenance periods), compare the deviation of the data with the baseline distribution.

[0072] The deviation degree between the production data distribution and the benchmark distribution in the current time period is analyzed to generate the benchmark distribution deviation index, which is used to judge the deviation degree between the production data distribution and the benchmark distribution in the current time period. The benchmark distribution deviation index is obtained as follows:

[0073] Setting the baseline distribution and current distribution: The baseline distribution P is the ideal production status data distribution, which is usually constructed based on historical data, previous production standards, target production status, etc. The current distribution Q is the distribution of production data collected in real time during the current time period.

[0074] If the production data is continuous, it needs to be discretized. This can usually be done through binning or selecting an appropriate probability density function (PDF). Common discretization methods include: discretizing into a probability distribution with a fixed interval (such as a histogram) or using a probability density estimation method (such as kernel density estimation (KDE)).

[0075] Calculate the KL divergence between the baseline distribution and the current time period distribution. The calculation formula is: ;in: is the KL divergence between the baseline distribution and the current time period distribution; P(i) is the probability mass (or probability density) of the i-th class in the baseline distribution P. Q(i) is the probability mass (or probability density) of the i-th class in the current distribution Q. The calculated KL divergence between the baseline distribution and the current time period distribution is used as the baseline distribution deviation index.

[0076] KL divergence is very sensitive to zero probability values. If there are zero values ​​in P(i) or Q(i), the result may tend to infinity. To avoid this, a smoothing constant v is added to the distribution, that is: ; is the probability mass of the i-th class in the smoothed benchmark distribution P, is the probability mass of the i-th class in the current distribution Q after smoothing.

[0077] A larger deviation index indicates a greater degree of deviation between the production data distribution and the baseline distribution during the current time period. This indicates that the actual production data differs significantly from the ideal production state, possibly due to factors such as changes in the production environment, equipment status, or fluctuations in production demand. When the deviation is large, the intelligent scheduling system may fail to adapt to these changes, leading to irrational resource allocation, decreased production efficiency, or increased risk of potential equipment failure. Therefore, the system needs to adjust and re-optimize the scheduling strategy to address the current changes in the production environment.

[0078] Conversely, a smaller baseline distribution deviation index indicates a smaller deviation between the production data distribution and the baseline distribution during the current time period. This indicates that actual production data is closer to the ideal target production state, and that the intelligent scheduling system is able to effectively track and adapt to the current production environment, ensuring optimal resource allocation and smooth production operations. Smaller deviations mean that the scheduling system maintains good adaptability to changes in the production environment, and that system decisions effectively meet production needs, thereby improving production efficiency and reducing the risk of resource waste or equipment overload.

[0079] S3: If the degree of deviation is high, collect historical data on the scheduling prediction value of the intelligent scheduling system model and the actual equipment utilization rate, calculate the coupling degree after using regression analysis, and analyze the abnormal changes in the coupling degree under different drift states to evaluate the impact of the failure of the intelligent scheduling system model on the production target.

[0080] When the degree of deviation is high, there may be a significant discrepancy between the intelligent scheduling system's scheduling predictions and actual equipment utilization, indicating poor model adaptability and leading to problems in resource allocation and production scheduling. Therefore, to further analyze and evaluate the impact of this deviation, we can collect historical data on the intelligent scheduling system's scheduling predictions and actual equipment utilization, use regression analysis to quantify the relationship between them, and further assess the impact of system failures on production targets.

[0081] Scheduling predictions: This is the intelligent scheduling system's prediction of equipment utilization and production schedule based on historical data and current production conditions. Actual equipment utilization: This is the actual equipment operating status data collected from the production line, including equipment operating hours and load levels. By accumulating historical data, we collect predicted and actual values ​​for different time periods.

[0082] Regression analysis models (such as linear regression and ridge regression) are used to establish the relationship between the scheduling prediction value and the actual equipment utilization. The goal is to estimate the deviation between the actual utilization and the predicted utilization through the regression model.

[0083] The basic form of a linear regression model is: ; Where: Y is the actual equipment utilization rate, X is the scheduling prediction value of the intelligent scheduling system, β0 is the intercept term, β1 is the regression coefficient, which indicates the strength of the relationship between the predicted value and the actual value, and ϵ is the error term, which indicates the deviation of the model.

[0084] Through regression analysis, the regression coefficient β1 and the model's fit can be calculated to quantify the degree of match between the predicted value and the actual equipment utilization.

[0085] The coupling degree reflects the strength of the relationship between the predicted value and the actual equipment utilization rate. In regression analysis, this coupling degree can be measured by calculating the correlation coefficient r, which is: Where, and are the predicted value and actual value at the i-th moment, and is the average of the predicted value and the actual value. The coupling degree r ranges from [−1,1]: It indicates that the predicted value is highly correlated with the actual value, and the system's scheduling model has good adaptability and small error. It means that there is no obvious relationship between the predicted value and the actual value, and the model is invalid or has poor adaptability. Indicates a high negative correlation between the predicted and actual values.

[0086] In actual applications, data distribution and the production environment may change, resulting in changes in the relationship between the predicted value of the intelligent scheduling system and the actual equipment utilization rate. If the coupling degree fluctuates abnormally, it means that the adaptability and prediction accuracy of the system are affected, which may cause production scheduling to fail. For example: Normal state: When the production environment does not change much, the coupling degree between the predicted value and the actual value is usually stable, reflecting the effectiveness of the scheduling system over the past period of time. Drift state: When the production environment changes (such as equipment aging, fluctuations in production demand, etc.), the coupling degree may decrease. If the coupling degree is significantly lower than the normal state, it means that the risk of model failure has increased and the system needs to be adjusted.

[0087] After analyzing the abnormal changes in coupling degree under different drift states, a coupling degree abnormality index is generated to evaluate the impact of failure of the intelligent scheduling system model on the production target. The coupling degree abnormality index is obtained as follows:

[0088] Historical coupling data (normal mode): represents the coupling data of the system when it is operating stably and without anomalies. This is typically based on historical production data (such as equipment utilization and production efficiency). Current coupling data: This is the coupling data collected in real time from the intelligent scheduling system during the current time period.

[0089] Construct historical coupling degree series respectively , and construct coupling degree sequence ; First, construct an n×m distance matrix, where n and m are the lengths of the historical coupling sequence and the current coupling sequence respectively; at each element position (i, j), calculate the current time point and The distance between them is usually the Euclidean distance , the expression is: ; The cumulative distance matrix is ​​calculated by dynamic programming. Each element D(i,j) of the cumulative distance matrix represents the minimum cumulative distance from the starting point to the position (i,j). The calculation formula is: Where, Represents paths in three directions: vertical, horizontal, and diagonal. Minimum distances are accumulated by selecting the shortest path. The shortest path DTW distance is calculated. The DTW distance is the element D(n,m) in the lower right corner of the matrix, which is the minimum cumulative distance from the historical data sequence to the current data sequence.

[0090] In order to make the DTW distance comparable between time series of different lengths, it is usually necessary to normalize the DTW distance. This can be done by dividing the minimum distance by the total length of the two sequences. To normalize, get a standardized DTW distance value , the expression is: ; The degree of anomaly is assessed by comparing the current DTW distance with the historical normal pattern.

[0091] The range of the DTW distance in normal mode is to calculate the mean and standard deviation of the DTW distance in multiple normal periods through historical data, and set the mean of the DTW distance in the historical normal mode to , the standard deviation is ; According to the difference between the current DTW distance and the historical mean, the coupling anomaly index is calculated. The formula is: ; Where Gt is the coupling anomaly index.

[0092] The larger the coupling anomaly index, the more significant the difference between the current intelligent scheduling system's coupling and historical normal patterns. This indicates a significant anomaly in the system's current operating state, likely due to scheduling failures, equipment overload, or irrational resource allocation, resulting in a significant impact on production targets. In this scenario, production efficiency declines significantly, critical tasks are delayed, and production lines may even stall, impacting overall capacity and progress.

[0093] The smaller the coupling anomaly index, or if it approaches zero, the closer the coupling degree is to the historical normal pattern for the current intelligent scheduling system, indicating relatively stable system operation. In this case, the scheduling system can effectively adapt to changes in the production environment, making timely adjustments to ensure smooth production. As a result, production targets are minimally impacted, production lines operate stably, resources are allocated rationally, production capacity is not significantly impacted, and overall production efficiency remains high.

[0094] S4: Evaluate the adaptability of the intelligent scheduling system model to data distribution drift based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on the production target when the intelligent scheduling system model fails.

[0095] The baseline distribution deviation index and the coupling degree anomaly index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the adaptability value label of the intelligent scheduling system model to the data distribution drift as the prediction target, and takes minimizing the sum of the prediction errors of the adaptability value labels of all intelligent scheduling system models to the data distribution drift as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The adaptability value of the intelligent scheduling system model to the data distribution drift is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0096] The method for obtaining the adaptability value of the intelligent scheduling system model to data distribution drift is to obtain the corresponding function expression from the comprehensive feature vector training data of the trained machine learning model: Where, is the output function of the model, Hr is the benchmark distribution deviation index, Gt is the coupling anomaly index, is the adaptability value of the intelligent scheduling system model to data distribution drift.

[0097] S5: Based on the evaluation results, the adaptability of the intelligent scheduling system model to data distribution drift is divided into high, medium, and low adaptability levels. For the high adaptability level, the resource allocation strategy is further refined. For the low adaptability level, failed decisions are quickly responded to and repaired to reduce the impact on production.

[0098] Comparing the obtained adaptability value of the intelligent scheduling system model to data distribution drift with the gradient standard threshold, where the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. Comparing the adaptability value of the intelligent scheduling system model to data distribution drift with the first standard threshold and the second standard threshold respectively;

[0099] If the adaptability value of the intelligent scheduling system model to data distribution drift is greater than the second standard threshold, it means that the intelligent scheduling system model has good adaptability to data distribution drift. At this time, a high adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a high adaptability level.

[0100] If the adaptability value of the intelligent scheduling system model to data distribution drift is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it means that the adaptability of the intelligent scheduling system model to data distribution drift is average. In this case, a medium adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a medium adaptability level.

[0101] If the adaptability value of the intelligent scheduling system model to data distribution drift is less than the first standard threshold, it means that the intelligent scheduling system model has poor adaptability to data distribution drift. At this time, a low adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a low adaptability level.

[0102] An intelligent scheduling system model rated as highly adaptable demonstrates its ability to effectively adapt to data distribution drift, high prediction accuracy, and strong adaptability to production environments and scheduling decisions. Therefore, the primary goal of this level of system performance is to further optimize and refine resource allocation to improve overall production efficiency. Specific strategies include:

[0103] Based on real-time data, production task scheduling can be further optimized. For example, machine learning algorithms can be used to predict the actual completion time of each task and adjust task scheduling to reduce equipment idle time and waiting time. Based on the production line's load, equipment workloads can be dynamically adjusted to ensure that each device operates at optimal conditions and avoid wasted resources.

[0104] Optimize worker scheduling and assign work based on worker skills and production requirements, ensuring workers are assigned to the most suitable positions. Through data analysis, predict changes in staffing needs, avoid staff redundancies or shortages, and ensure smooth production.

[0105] Leveraging data analysis and predictive maintenance technology, we adjust maintenance cycles based on equipment usage and health status, avoiding over- or under-maintenance and ensuring optimal equipment operation. We dynamically adjust equipment maintenance schedules and production scheduling to reduce downtime and improve equipment utilization.

[0106] Accurately predict energy demand during the production process, optimize energy scheduling through smart grids and energy management systems, reduce unnecessary energy consumption, and improve energy utilization efficiency.

[0107] For intelligent scheduling system models rated as low adaptability, this indicates that the system fails to adapt promptly to changes in the production environment or data distribution drift, resulting in prediction failures and a significant impact on production targets. Therefore, the primary goal of the system at this level is to quickly respond to and repair failed decisions to minimize the negative impact on production. Specific strategies include:

[0108] The system must quickly identify the cause of current scheduling failures and adjust the scheduling plan based on real-time feedback. For example, if the system predicts that uneven equipment loads are causing some equipment to be overloaded while others are idle, the equipment's operating schedule must be adjusted immediately to distribute the workload appropriately. For the scheduling of critical tasks, emergency scheduling strategies are implemented to ensure that critical tasks are completed first and avoid delays or stagnation.

[0109] Because low adaptability levels may be associated with aging equipment or frequent failures, enhanced equipment health monitoring is essential. Real-time monitoring of equipment operating data, such as vibration, temperature, and pressure, allows for timely detection and intervention. If equipment is found to be operating abnormally or about to fail, immediate maintenance is arranged to prevent further production line downtime caused by the failure.

[0110] In situations with low adaptability, automated fault detection and repair mechanisms can be added to the system. For example, automated systems can remotely monitor equipment and, if a fault or abnormal operation is detected, automatically switch to backup equipment or reschedule tasks. This enhances emergency response capabilities. When the system identifies a risk, it can immediately adjust the production process to prevent a single failure from bringing the entire production line to a standstill.

[0111] When the system model's adaptability is low, a simplified production strategy can be temporarily adopted. For example, this reduces complex production scheduling decisions, reduces the need for flexibility in the production process, and mitigates production line risks. Manual intervention can be used to manually adjust the scheduling strategy when the model fails, ensuring that production continues to stagnate and that critical production tasks are completed on time.

[0112] S6: For the medium adaptability level, the abnormal degree of the intelligent scheduling system model's adaptability to data distribution drift within a fixed time period is predicted. If the abnormal degree is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions.

[0113] For the medium adaptability level, that is, the adaptability value of the intelligent scheduling system model generated within a fixed time period to the data distribution drift is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the adaptability values ​​generated in the subsequent fixed time period that are greater than or equal to the first standard threshold and less than or equal to the second standard threshold are collected, and the corresponding data set is established, and the standard deviation of the data set is calculated as the adaptability abnormality degree value.

[0114] The obtained adaptability anomaly degree value is compared with the anomaly degree value reference threshold value pre-set based on historical data. If the adaptability anomaly degree value is greater than or equal to the pre-set anomaly degree value reference threshold value, it means that the anomaly degree of adaptability of the intelligent scheduling system model to data distribution drift within a fixed time period is high. At this time, an early warning signal is generated, indicating that the adaptability of the scheduling system is poor and the decision-making accuracy is low, and the intelligent scheduling system is optimized; if the adaptability anomaly degree value is less than the pre-set anomaly degree value reference threshold value, it means that the anomaly degree of adaptability of the intelligent scheduling system model to data distribution drift within a fixed time period is low. At this time, no early warning signal is generated and no additional processing is required.

[0115] The optimization goal is to reduce the abnormality level (AE) to improve system adaptability. Historical production data, equipment utilization data, and scheduling decision results are collected. Machine learning algorithms such as regression models, support vector machines (SVMs), and neural networks are used to train models to predict and adjust scheduling decisions. The goal of the optimization model is to minimize prediction error and the abnormality level.

[0116] Set the scheduling decision of the current model through a function To express it, the intelligent scheduling system optimizes the decision by adjusting the model parameters θ: ;in: is a scheduling decision based on the current model prediction, is the actual production output result, is the regularization parameter, balancing the weight between prediction error and anomaly degree, AE is the anomaly degree value, and s is the total number of samples, which refers to the number of samples in all training data.

[0117] According to the optimized objective function, the model parameters θ are adjusted and the new parameters are used to predict the scheduling decision. The optimization process is updated based on the gradient descent method: ; where η is the learning rate, ∇θ is the gradient of the objective function with respect to the model parameters, The model parameters used by the intelligent scheduling system in iteration t. It represents the set of all parameters used by the model to generate scheduling decisions in the current iteration. For the In the iteration, the new model parameters are optimized.

[0118] The optimization process is a continuous iterative process. After each update, the system recalculates the abnormality level (AE) and determines whether further optimization is needed. If the abnormality level remains high, the system further adjusts the scheduling strategy and repeatedly optimizes until the AE is reduced to an acceptable range.

[0119] Once the optimization process is complete, the system updates the scheduling decision model and applies it to actual production. The optimized intelligent scheduling system can better adapt to data distribution drift, improve decision accuracy and production efficiency.

[0120] In this embodiment, the system first sets the ideal production state data as a target and collects production data in real time for comparison, identifying data that deviates from the target. Then, by analyzing the distribution changes of the deviating data and calculating the degree of deviation from the baseline distribution, the system assesses the model failure risk of the intelligent scheduling system under different drift states. Based on these assessments, the system divides adaptability into three levels: high, medium, and low, and takes corresponding countermeasures based on the adaptability level: for high adaptability, the resource allocation strategy is further refined; for low adaptability, failed decisions are quickly responded to and repaired to reduce production impact; for medium adaptability, the system is optimized to improve the accuracy of scheduling decisions and ensure the efficient and smooth operation of the production line.

[0121] Example 2: An intelligent scheduling and tracking system described in this example includes a data comparison module, a distribution detection module, a coupling analysis module, an adaptability evaluation module, an adaptability level division module, and an optimization module.

[0122] Data comparison module: Sets the ideal production status data in the production process as the target production data, and collects real-time production data in several time periods during the actual production process. The real-time production data is compared with the target production data, and the real-time production data is divided into those that deviate from the target and those that do not deviate from the target.

[0123] Distribution Detection Module: Analyzes the distribution changes of production data that deviates from the target in each time period, selects the production data distribution of the initial time period of the intelligent scheduling system operation as the benchmark distribution, and determines the degree of deviation between the production data distribution in the current time period and the benchmark distribution;

[0124] Coupling analysis module: If the degree of deviation is high, historical data on the scheduling prediction value of the intelligent scheduling system model and the actual equipment utilization rate are collected, and the coupling degree is calculated after using regression analysis. The abnormal changes in the coupling degree under different drift states are analyzed to evaluate the impact of the failure of the intelligent scheduling system model on production targets.

[0125] Adaptability evaluation module: This module evaluates the adaptability of the intelligent scheduling system model to data distribution drift based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on production targets when the intelligent scheduling system model fails.

[0126] Adaptability level classification module: Based on the evaluation results, the adaptability of the intelligent scheduling system model to data distribution drift is divided into high, medium, and low adaptability levels. For the high adaptability level, the resource allocation strategy is further refined. For the low adaptability level, failed decisions are quickly responded to and repaired to reduce the impact on production.

[0127] Optimization module: For the medium adaptability level, the abnormal degree of the intelligent scheduling system model's adaptability to data distribution drift within a fixed time period is predicted. If the abnormal degree is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions.

[0128] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0129] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0130] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0131] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. An intelligent scheduling and tracking method, characterized by: The following steps are involved: S1: Set the ideal production status data in the production process as the target production data, and collect real-time production data in several time periods during the actual production process. Compare the real-time production data with the target production data, and divide the real-time production data into those that deviate from the target and those that do not deviate from the target; S2: Analyze the distribution changes of production data that deviate from the target in each time period, select the production data distribution in the initial time period of the intelligent scheduling system as the benchmark distribution, and determine the degree of deviation between the production data distribution in the current time period and the benchmark distribution; S3: If the deviation is high, collect historical data on the scheduling predictions of the intelligent scheduling system model and actual equipment utilization, calculate the coupling degree after performing regression analysis, and analyze abnormal changes in the coupling degree under different drift states to assess the impact of failure of the intelligent scheduling system model on production targets. S4: Evaluate the adaptability of the intelligent scheduling system model to data distribution drift based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on production targets when the intelligent scheduling system model fails; S5: Based on the evaluation results, the adaptability of the intelligent scheduling system model to data distribution drift is divided into high, medium, and low adaptability levels. For the high adaptability level, the resource allocation strategy is further refined. For the low adaptability level, failed decisions are quickly responded to and repaired to reduce the impact on production. S6: For the medium adaptability level, the abnormal degree of the intelligent scheduling system model's adaptability to data distribution drift within a fixed time period is predicted. If the abnormal degree is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions.

2. The intelligent scheduling and tracking method according to claim 1, characterized in that: In S1, the real-time data and the target data are compared within the same time window, and the deviation between the actual production data and the target production data is calculated. The expression is: The deviation between the actual production data and the target production data is compared with the deviation reference threshold under normal production conditions set based on historical data. If the deviation between the actual production data and the target production data is greater than the deviation reference threshold, it is classified as deviating from the target production data. If the deviation between the actual production data and the target production data is less than or equal to the deviation reference threshold, it is classified as not deviating from the target production data.

3. The intelligent scheduling and tracking method according to claim 1, characterized in that: In S2, the deviation degree between the production data distribution and the benchmark distribution in the current time period is analyzed to generate a benchmark distribution deviation index. The benchmark distribution deviation index is obtained as follows: Set the baseline distribution and current distribution. The baseline distribution P is the ideal production status data distribution, and the current distribution Q is the distribution of production data collected in real time during the current time period. Calculate the KL divergence between the baseline distribution and the current time period distribution. The calculation formula is: ;in: is the KL divergence between the benchmark distribution and the distribution of the current time period; P(i) is the probability mass of the i-th class in the benchmark distribution P; Q(i) is the probability mass of the i-th class in the current distribution Q; the calculated KL divergence between the benchmark distribution and the distribution of the current time period is used as the benchmark distribution deviation index.

4. The intelligent scheduling and tracking method according to claim 3, characterized in that: In S3, the abnormal changes of the coupling degree under different drift states are analyzed to generate a coupling degree abnormality index. The coupling degree abnormality index is obtained as follows: Construct historical coupling degree series respectively , and construct coupling degree sequence ; First, construct an n×m distance matrix, where n and m are the lengths of the historical coupling sequence and the current coupling sequence respectively; at each element position (i, j), calculate the current time point and The distance between them, using Euclidean distance , the expression is: ; The cumulative distance matrix is ​​calculated by dynamic programming. Each element D(i, j) of the cumulative distance matrix represents the minimum cumulative distance from the starting point to the position (i, j). The calculation formula is: Where, Represents paths in three directions: vertical, horizontal, and diagonal. By selecting the shortest path to accumulate the minimum distance, the shortest path DTW distance is calculated. The DTW distance is the element D(n,m) in the lower right corner of the matrix, that is, the minimum cumulative distance from the historical data sequence to the current data sequence. Normalize the DTW distance to get a standardized DTW distance value , the expression is: ; The range of the DTW distance in normal mode is the mean and standard deviation of the DTW distances of multiple normal periods calculated from historical data. The mean of the DTW distance in the historical normal mode is set to , the standard deviation is ; According to the difference between the current DTW distance and the historical mean, the coupling anomaly index is calculated. The formula is: ; Where Gt is the coupling anomaly index.

5. The intelligent scheduling and tracking method according to claim 4, characterized in that: In S4, the adaptability of the intelligent scheduling system model to data distribution drift is evaluated based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on the production target when the intelligent scheduling system model fails. Specifically: The baseline distribution deviation index and the coupling degree anomaly index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the adaptability value label of the intelligent scheduling system model to the data distribution drift as the prediction target, and takes minimizing the sum of the prediction errors of the adaptability value labels of all intelligent scheduling system models to the data distribution drift as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The adaptability value of the intelligent scheduling system model to the data distribution drift is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

6. The intelligent scheduling and tracking method according to claim 5, characterized in that: In S5, the adaptability of the intelligent scheduling system model to data distribution drift is divided into high adaptability level, medium adaptability level and low adaptability level, specifically: Comparing the obtained adaptability value of the intelligent scheduling system model to data distribution drift with the gradient standard threshold, where the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. Comparing the adaptability value of the intelligent scheduling system model to data distribution drift with the first standard threshold and the second standard threshold respectively; If the adaptability value of the intelligent scheduling system model to data distribution drift is greater than the second standard threshold, it means that the intelligent scheduling system model has good adaptability to data distribution drift. At this time, a high adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a high adaptability level. If the adaptability value of the intelligent scheduling system model to data distribution drift is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it means that the adaptability of the intelligent scheduling system model to data distribution drift is average. In this case, a medium adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a medium adaptability level. If the adaptability value of the intelligent scheduling system model to data distribution drift is less than the first standard threshold, it means that the intelligent scheduling system model has poor adaptability to data distribution drift. At this time, a low adaptability signal is generated, and the adaptability of the intelligent scheduling system model to data distribution drift is classified as a low adaptability level.

7. The intelligent scheduling and tracking method according to claim 1, characterized in that: In S6, for the medium adaptability level, the abnormal degree of adaptability of the intelligent scheduling system model to data distribution drift within a fixed time period is predicted, specifically: For the medium adaptability level, that is, the adaptability value of the intelligent scheduling system model to data distribution drift generated within a fixed time period is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the adaptability values ​​generated within the subsequent fixed time period that are greater than or equal to the first standard threshold and less than or equal to the second standard threshold are collected, and the corresponding data set is established, and the standard deviation of the data set is calculated as the adaptability abnormality degree value; The obtained adaptability abnormality value is compared with the abnormality reference threshold value pre-set according to historical data. If the adaptability abnormality value is greater than or equal to the pre-set abnormality reference threshold value, it means that the abnormality of the adaptability of the intelligent scheduling system model to the data distribution drift within the fixed time period is high. At this time, an early warning signal is generated and the intelligent scheduling system is optimized; if the adaptability abnormality value is less than the pre-set abnormality reference threshold value, it means that the abnormality of the adaptability of the intelligent scheduling system model to the data distribution drift within the fixed time period is low. At this time, no early warning signal is generated and no additional processing is required.

8. The intelligent scheduling and tracking method according to claim 7, characterized in that: If the abnormality level is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions. Specifically: Set the scheduling decision of the current model through a function To express it, the intelligent scheduling system optimizes the decision by adjusting the model parameters θ: ;in: is a scheduling decision based on the current model prediction, is the actual production output result, is the regularization parameter, AE is the abnormality value, and s is the total number of samples; According to the optimized objective function, the model parameters θ are adjusted and the new parameters are used to predict the scheduling decision. The optimization process is updated based on the gradient descent method: ; where η is the learning rate, ∇θ is the gradient of the objective function with respect to the model parameters, are the model parameters used in the tth iteration of the intelligent scheduling system, For the In the iteration, the new model parameters are optimized.

9. An intelligent scheduling and tracking system, for implementing the intelligent scheduling and tracking method according to any one of claims 1 to 8, characterized in that: It includes data comparison module, distribution detection module, coupling analysis module, adaptability evaluation module, adaptability level division module and optimization module: Data comparison module: Sets the ideal production status data in the production process as the target production data, and collects real-time production data in several time periods during the actual production process. The real-time production data is compared with the target production data, and the real-time production data is divided into those that deviate from the target and those that do not deviate from the target. Distribution Detection Module: Analyzes the distribution changes of production data that deviates from the target in each time period, selects the production data distribution of the initial time period of the intelligent scheduling system operation as the benchmark distribution, and determines the degree of deviation between the production data distribution in the current time period and the benchmark distribution; Coupling analysis module: If the degree of deviation is high, historical data on the scheduling prediction value of the intelligent scheduling system model and the actual equipment utilization rate are collected, and the coupling degree is calculated after using regression analysis. The abnormal changes in the coupling degree under different drift states are analyzed to evaluate the impact of the failure of the intelligent scheduling system model on production targets. Adaptability evaluation module: This module evaluates the adaptability of the intelligent scheduling system model to data distribution drift based on the degree of deviation between the production data distribution and the benchmark distribution in the current time period and the impact on production targets when the intelligent scheduling system model fails. Adaptability level classification module: Based on the evaluation results, the adaptability of the intelligent scheduling system model to data distribution drift is divided into high, medium, and low adaptability levels. For the high adaptability level, the resource allocation strategy is further refined. For the low adaptability level, failed decisions are quickly responded to and repaired to reduce the impact on production. Optimization module: For the medium adaptability level, the abnormal degree of the intelligent scheduling system model's adaptability to data distribution drift within a fixed time period is predicted. If the abnormal degree is high, the intelligent scheduling system model is continuously optimized to improve the accuracy of scheduling decisions.

Images