Intelligent production task scheduling method based on time sequence knowledge base

Abstract

The invention discloses an intelligent production task scheduling method based on a time sequence knowledge base, and the method comprises the steps: carrying out the analysis through a time sequence method according to equipment operation data, production task data and environment data which are collected in real time, and obtaining a production rule; querying a knowledge base according to the production rule to obtain a production task scheduling scheme; and adjusting the time sequence knowledge base according to the execution condition of the production task scheduling scheme. According to the method, the dynamic adaptability of production task scheduling is remarkably improved through a closed-loop process of real-time data acquisition, time sequence analysis, knowledge base dynamic query and feedback iteration. Multi-dimensional data fusion and time sequence rule mining are carried out, a precise scheduling scheme is generated, a knowledge base continuously learns new data, the method adapts to long-term changes of a production environment, manual intervention is reduced, and decision-making precision is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent production task scheduling, and particularly relates to an intelligent production task scheduling method based on a time series knowledge base. Background Art

[0002] With the rapid development of Industry 4.0 and intelligent manufacturing technologies, the intelligence and dynamic adaptability of production scheduling have become the core requirements for enhancing the competitiveness of the manufacturing industry. Traditional production task scheduling methods mostly rely on static rules or empirical models set by humans, and execute tasks through predefined priority and resource allocation strategies. However, such methods have significant limitations when facing complex and changeable production environments.

[0003] Traditional scheduling systems usually perform task allocation based on fixed rules and historical experience, and it is difficult to respond in real time to dynamic changes such as equipment failures, fluctuations in environmental parameters, or sudden task requirements. For example, when production equipment malfunctions or environmental conditions (such as temperature, humidity, pressure) exceed the preset range, the system cannot quickly adjust the scheduling strategy, resulting in production interruptions or resource waste.

[0004] In addition, the existing technologies mostly stay at single-point or fragmented analysis of production data, and fail to fully explore the temporal dependence relationships and potential laws in time series data. For example, key information such as the periodic fluctuations of equipment operating states and the bottleneck laws of task execution has not been systematically extracted, resulting in the lack of data-driven accuracy in scheduling decisions.

[0005] Generally speaking, the existing production scheduling technologies have significant deficiencies in aspects such as dynamic adaptability and data-driven decision-making capabilities, and there is an urgent need for an intelligent production task scheduling method to achieve the dynamic adaptation of production task scheduling. Summary of the Invention

[0006] The present invention provides an intelligent production task scheduling method based on a time series knowledge base to solve the problems of insufficient dynamic adaptability of the method caused by the difficulty in real-time response to dynamic changes, and insufficient data-driven decision-making accuracy caused by ignoring temporal dependence relationships and potential laws.

[0007] The technical solution adopted by the present invention is as follows:

[0008] An intelligent production task scheduling method based on a time series knowledge base, comprising:

[0009] Analyze the production rules through time series methods based on the device operation data, production task data, and environmental data collected in real time;

[0010] Query the knowledge base according to the production rules to obtain a production task scheduling plan;

[0011] Adjust the time - series knowledge base according to the execution status of the production task scheduling plan.

[0012] The intelligent production task scheduling method based on the time - series knowledge base disclosed in the present invention further has the following additional technical features:

[0013] Query the knowledge base according to the production rules to obtain a production task scheduling plan, specifically:

[0014] According to the data collected in real - time, through the similarity of the real - time data with the cluster classification of multiple scheduling plans in the knowledge base, obtain multiple preliminary scheduling plans corresponding to the production rules;

[0015] According to multiple said preliminary scheduling plans, combined with the corresponding execution status data in the knowledge base, obtain a production task scheduling plan.

[0016] The knowledge base is specifically:

[0017] The knowledge base contains multiple said production rules and the corresponding execution status of the production task scheduling plan;

[0018] The production rules include equipment operation modes, task execution modes, and production process relationships.

[0019] The construction of the knowledge base is specifically:

[0020] According to historical data, obtain time - series embedding vectors through data conversion;

[0021] According to the time - series embedding vectors, through cluster analysis, according to the Euclidean distance between the data points within the cluster and the cluster center, obtain the cluster classification of the production rules of multiple scheduling plans.

[0022] The equipment operation data, production task data, and environmental data are specifically:

[0023] The equipment operation data includes at least any one of the working status, temperature, load, and fault records of the equipment,

[0024] The production task data includes at least any one of the execution requirements, priorities, and required resources of the production tasks,

[0025] The environmental data includes at least any one of temperature and humidity, pressure, and air flow.

[0026] Adjust the time - series knowledge base according to the execution status of the production task scheduling plan, specifically:

[0027] According to the device operation data, production task data, environmental data collected in real time, and the execution feedback data of the production task scheduling scheme, the execution feedback data includes at least any one of task completion time, resource utilization rate, abnormal event record, and manual intervention record;

[0028] According to the device operation data, production task data, and environmental data, obtain a production task scheduling scheme;

[0029] According to the execution feedback data, obtain the execution situation of the production task scheduling scheme to update and optimize the production task scheduling scheme.

[0030] When there are multiple production tasks, the production task scheduling method is specifically as follows:

[0031] Correspondingly obtain multiple production task scheduling schemes according to the multiple production tasks;

[0032] According to the multiple production task scheduling schemes, combined with the priorities of the multiple production tasks, obtain a production task scheduling strategy, and the production task scheduling strategy includes task execution order, device resource allocation, and time window;

[0033] Among them, the priority of the production task is determined according to the task urgency, deadline constraint, and resource demand weight.

[0034] The intelligent production task scheduling method based on the time series knowledge base further includes: [[ID=2a3]]

[0035] Preprocess the collected data, and the preprocessing includes denoising, missing value filling, and normalization processing;

[0036] The denoising processing includes at least the wavelet transform method, the missing value filling includes at least the interpolation method, and the normalization processing includes at least the linear normalization method.

[0037] The present invention also discloses a storage medium,

[0038] A computer program is stored on the storage medium, and when the computer program is executed, the steps of the intelligent production task scheduling method based on the time series knowledge base are implemented.

[0039] The present invention further discloses a processing device, including:

[0040] A memory for storing a computer program;

[0041] A processor for implementing the steps of the intelligent production task scheduling method based on the time series knowledge base when executing the computer program.

[0042] Due to the adoption of the above technical solution, the beneficial effects achieved by the present invention are as follows:

[0043] 1. In the present invention, by collecting device operation data, production task data, and environmental data in real time, and using time series analysis to mine production rules, combined with historical scheduling cases and execution feedback in the knowledge base, a dynamic scheduling strategy is generated. This real-time response mechanism significantly reduces the risk of production interruption caused by static rules in traditional methods. At the same time, the automation decision-making of the production task scheduling scheme reduces the dependence on manual experience and ensures the continuity of the production process.

[0044] According to the execution situation of the production task scheduling scheme, the time series knowledge base is adjusted. After each production cycle ends, new data is added to the knowledge base, and an online learning iteration strategy is adopted. The dynamic update of the historical case base gradually adapts to long-term changes and reduces the rule maintenance cost. The closed-loop feedback mechanism continuously improves the intelligence level of this method, and there is no need for manual regular update of the rule base.

[0045] In addition, this method supports multi-dimensional data fusion (device operation data, production task data, environmental data) to construct the knowledge base. When adding new devices or task types, only the corresponding data needs to be supplemented to the knowledge base to adapt to production lines of different scales. The scalability of the knowledge base meets the differentiated requirements of complex multi-task scenarios.

[0046] Generally speaking, through the closed-loop process of real-time data collection, time series analysis, dynamic query of the knowledge base, and feedback iteration, the present invention significantly improves the dynamic adaptability of production task scheduling. Multi-dimensional data fusion and time series law mining generate accurate scheduling schemes. The knowledge base continuously learns new data, adapts to long-term changes in the production environment, reduces manual intervention, and improves the decision-making accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0048] Figure 1 It is a schematic flow chart of the intelligent production task scheduling method based on the time series knowledge base under an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] In order to more clearly illustrate the overall concept of the present invention, the following will be described in detail by way of examples in conjunction with the drawings of the specification.

[0050] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited by the specific embodiments disclosed below.

[0051] As Figure 1 shown, an intelligent production task scheduling method based on a time series knowledge base includes:

[0052] S100: Analyze the device operation data, production task data, and environmental data collected in real time through time series methods to obtain production rules.

[0053] The main purpose of this step is to identify potential rules and patterns in the production process through real-time monitoring and analysis of the operation status of production equipment, the execution of production tasks, and environmental parameters. These rules help predict future production demands, optimize resource allocation, and provide a scientific basis for subsequent task scheduling. Through this method, the system can better adapt to the dynamically changing production environment and improve production efficiency and resource utilization.

[0054] Among them, the time series method is a statistical analysis method used to study patterns and trends in time series data. Through time series analysis, the system can identify potential rules such as periodic failure patterns of equipment and time rules for task execution, predict potential problems in production in advance, and optimize scheduling strategies.

[0055] In this step, the collected data includes device operation data, production task data, and environmental data. Device operation data can monitor the device status in real time, helping to detect device anomalies in a timely manner, reduce downtime, and improve production efficiency. Production task data provides accurate task data, supporting more reasonable resource allocation and task arrangement, and avoiding resource waste and bottleneck problems. Environmental data provides accurate environmental monitoring data, helping to maintain the best production conditions and ensure product quality and consistency.

[0056] According to the collected data, apply time series analysis methods (such as dynamic time warping DTW, K-means clustering, etc.) to mine production rules from historical data. For example, convert time series data into high-dimensional vector embeddings through ACGE-Text-Embedding, and then perform clustering analysis to identify periodic maintenance requirements of equipment or task execution bottlenecks.

[0057] Based on a time series model (such as an LSTM network), predict the production trend in the next period of time. For example, use historical data to predict possible failures of equipment in the next few days and arrange maintenance plans in advance. Accurate trend prediction helps to respond to potential problems in advance and reduce the risk of unexpected downtime and production interruptions.

[0058] In a specific embodiment, in an automobile manufacturing factory, the operation data of welding robot equipment (such as working status, temperature, load, etc.), production task data (such as the current batch of welding tasks and their priorities), and environmental data (such as temperature and humidity in the workshop) are collected in real time. It is identified that the temperature of the equipment fluctuates greatly during a specific period, and there is a phenomenon of excessive load. Further analyzing these data, it is found that the equipment shows a periodic pattern of temperature increase on Tuesday afternoons every week, which may be due to the concentration of production tasks during this period.

[0059] Based on this pattern, it is convenient to adopt corresponding production task scheduling schemes, remind the management to check the equipment cooling system in advance on Tuesday mornings, and adjust the task allocation to avoid equipment overload during peak hours. In addition, through the production pattern, bottleneck links under certain production task combinations can be found, and optimization suggestions are put forward, such as adjusting the priorities of some tasks or reallocating resources, thereby improving the efficiency and stability of the overall production line.

[0060] Generally speaking, this step not only realizes the comprehensive management and intelligent analysis of production data, but also provides a solid foundation for subsequent task scheduling.

[0061] S200: According to the production pattern, query the knowledge base to obtain a production task scheduling scheme.

[0062] The main purpose of this step is to use the extracted production pattern, combined with the constructed time series knowledge base, to quickly generate an optimal production task scheduling scheme. In this way, the system can make more accurate and efficient scheduling decisions by making full use of historical data and real-time data, thereby improving production efficiency, resource utilization rate, and reducing production bottlenecks.

[0063] Based on the potential patterns in the production process obtained from the previous steps. These patterns may include the periodic failure modes of equipment, the time patterns of task execution, production bottlenecks, etc. By mining these patterns, the system can predict and address potential problems in advance, optimize resource allocation, and improve production efficiency.

[0064] It can be understood that the knowledge base is a data set storing historical production data and scheduling strategies, which is used to support intelligent decision-making.

[0065] The knowledge base contains historical records of various production patterns and their corresponding production task scheduling schemes. For example, information such as different equipment operation modes, task execution modes, and production process relationships is stored in the knowledge base.

[0066] The system compares the current production pattern with historical cases in the knowledge base to find the most similar historical scheduling plan. Common methods include the Dynamic Time Warping (DTW) algorithm, which can measure the similarity between two time series.

[0067] Among them, Dynamic Time Warping (DTW) refers to an algorithm used to calculate the optimal alignment path between two time series, especially suitable for sequences of different lengths.

[0068] Through similarity matching in this step, the system can quickly find the scheduling plan closest to the current situation from a large amount of historical data, reducing the time and complexity of redesigning the scheduling strategy.

[0069] Based on the similarity matching results, the system combines the specific requirements of the current production task to generate an optimal task scheduling plan. This includes the execution order of tasks, the allocation of equipment resources, and the necessary time windows. The generated scheduling plan not only considers historical experience but also combines the current actual situation, ensuring the accuracy and adaptability of scheduling decisions.

[0070] In a specific embodiment, in a semiconductor manufacturing factory, it is identified that there is a phenomenon of peak equipment load during a specific period in the task execution process of a certain batch of chip production. This phenomenon may be caused by the simultaneous start of multiple high-priority tasks during this period.

[0071] The system queries the knowledge base to find the historical case most similar to the current production pattern. Suppose there is a record in the knowledge base indicating that in a similar situation, by adjusting the time windows of some non-critical tasks, the equipment load can be effectively dispersed to avoid overload problems during peak periods. The system calculates the similarity between the current production pattern and this historical case through the DTW algorithm and confirms it as the closest reference case.

[0072] Based on this historical case, the system generates a new production task scheduling plan.

[0073] The production task scheduling plan can include: adjusting the time windows of some non-critical tasks to periods with lower load; arranging the equipment resources required for critical tasks in advance to ensure their completion on time; dynamically monitoring the progress of task execution and being ready to adjust the scheduling strategy at any time to cope with emergencies.

[0074] Through this intelligent scheduling method, the factory has successfully avoided the risk of equipment overload, improving the overall efficiency and stability of the production line. In addition, since the system can quickly find the optimal scheduling plan, the need for manual intervention is reduced, further enhancing the continuity and automation level of production.

[0075] S300: Adjust the time series knowledge base according to the execution situation of the production task scheduling plan.

[0076] The main purpose of this step is to dynamically update and optimize the time series knowledge base by analyzing the actual execution of the production task scheduling plan. Through continuous learning and adjustment, the system can better adapt to the changing production environment, continuously improve the scheduling strategy, and improve production efficiency and resource utilization. This process ensures the long-term performance improvement of the system and reduces the need for manual intervention.

[0077] In this step, during the production process, the execution of tasks is monitored in real-time (such as task completion time, resource utilization, abnormal event records, manual intervention records, etc.). These data are the key basis for evaluating the effectiveness of the scheduling plan. Real-time monitoring helps to promptly detect and solve problems in production, reducing downtime and resource waste.

[0078] The system compares the actual execution with the expected goals to evaluate the effectiveness of the scheduling plan. For example, check whether tasks are completed on time, whether resources are utilized efficiently, and whether there are abnormal events. If tasks are completed smoothly as planned and resource utilization is high, the scheduling plan is considered effective; if there are delays or resource waste, the reasons need to be further analyzed. By continuously evaluating the effectiveness of the scheduling plan, the system can identify which strategies are effective and which need improvement, thus gradually optimizing the scheduling strategy.

[0079] Incorporate the execution feedback data into the time series knowledge base. This includes information such as task completion time, resource utilization, and abnormal event records. By continuously updating the knowledge base, the system can better adapt to long-term changes, such as equipment aging and process improvement, and maintain the effectiveness of the scheduling strategy.

[0080] In a specific embodiment, in an automobile manufacturing plant, the system generates a production task scheduling plan to arrange multiple welding robots to complete the welding tasks of a certain batch of products within a specific time period. During the execution process, the system monitors the working status, task progress, and resource usage of each robot in real-time and records all relevant execution feedback data (such as task completion time, resource utilization, abnormal event records, etc.).

[0081] First, analyze and evaluate this execution feedback data. Suppose it is found that a certain robot overheats during the task execution, resulting in an extended task completion time and affecting the overall production progress. The system records this abnormal event and incorporates it into the time series knowledge base.

[0082] Based on this abnormal data, identify the peak load phenomenon of this robot within a specific time period. To prevent similar problems from occurring again, the system adjusts the scheduling strategy and decides to arrange cooling time in advance for future similar tasks, or allocate some tasks to other idle robots.

[0083] Generally speaking, this step not only realizes the comprehensive evaluation and feedback on the execution of the production task scheduling plan, but also continuously optimizes the scheduling strategy by dynamically updating the time series knowledge base, ensuring the long-term performance improvement of the system. This adaptive learning mechanism enables the system to better cope with the complex and changeable production environment, reduce manual intervention, and improve production efficiency and resource utilization.

[0084] As a preferred embodiment of the present invention, the device operation data, production task data, and environmental data are specifically:

[0085] The device operation data includes at least any one of the working state, temperature, load, and fault record of the device.

[0086] The production task data includes at least any one of the execution requirements, priority, and required resources of the production task.

[0087] The environmental data includes at least any one of temperature and humidity, pressure, and air flow.

[0088] The main purpose of this embodiment is to determine specifically which types of device operation data, production task data, and environmental data to collect. By clarifying the types of these data that have a great impact on production scheduling, it can ensure that various factors in the production process can be comprehensively and accurately captured, thus providing a solid data foundation for subsequent time series analysis and intelligent scheduling.

[0089] The device operation data includes at least any one of the working state, temperature, load, and fault record of the device. Among them, the working state refers to whether the device is currently in the on, off, standby, or other states. The temperature refers to the temperature change of the device during operation, used to monitor whether the device has an overheating risk. The load refers to the current working load of the device, helping to evaluate the usage efficiency and potential bottlenecks of the device. The fault record refers to the faults that occur to the device and their repair records, used to predict future maintenance needs.

[0090] Real-time monitoring of the working state of the device helps to detect abnormal situations in a timely manner and reduce downtime. Temperature and load data can help identify the periodic maintenance needs of the device and arrange maintenance plans in advance. The fault record helps to accumulate historical data and optimize future maintenance strategies.

[0091] The production task data includes at least any one of the execution requirements, priority, and required resources of the production task. Among them, the execution requirements refer to the specific operation steps and standards of each production task to ensure that the task is completed according to the specifications. The priority refers to determining the priority order according to the urgency and importance of the task to ensure that key tasks are completed first. The required resources refer to the raw materials, equipment, and other resources required to complete the task, used for reasonable resource allocation.

[0092] Clear execution requirements help ensure the consistency and stability of product quality. Reasonable task prioritization can prevent critical tasks from being delayed and improve overall production efficiency. Accurate resource requirement information supports more reasonable resource allocation, reducing resource waste and bottleneck problems.

[0093] The environmental data includes at least any one of temperature and humidity, pressure, and air flow. Among them, temperature and humidity refer to the temperature and humidity levels in the production environment, which are particularly important in production processes sensitive to the environment. Pressure refers to the air pressure level in the production workshop, which affects air flow and pollutant control. Air flow refers to the air flow situation in the workshop, ensuring good ventilation and pollution control.

[0094] Precise temperature and humidity control helps maintain optimal production conditions, ensuring product quality and consistency. Appropriate pressure and air flow management can effectively control the air quality and pollutant levels in the workshop, reducing unexpected events during production. Comprehensive environmental monitoring data supports dynamic adjustment of the production environment, improving production efficiency and safety.

[0095] This implementation method can comprehensively understand various factors in the production process through the detailed collection and processing of equipment operation data, production task data, and environmental data, providing a solid foundation for subsequent time series analysis and intelligent scheduling. This multi-dimensional data fusion not only improves the level of intelligence but also significantly enhances production efficiency and resource utilization rate, ensuring the continuity and stability of production.

[0096] As a preferred implementation method under the present invention, the intelligent production task scheduling method based on the time series knowledge base further includes:

[0097] Preprocessing the collected data, where the preprocessing includes denoising, missing value filling, and normalization processing;

[0098] The denoising processing includes at least the wavelet transform method, the missing value filling includes at least the interpolation method, and the normalization processing includes at least the linear normalization method.

[0099] The main purpose of this implementation method is to ensure the accuracy of subsequent analysis and scheduling decisions by preprocessing the collected data. The preprocessing includes denoising, missing value filling, and normalization processing. These operations can improve the quality of the data, reduce noise interference, fill in missing values, and unify the data to the same scale, thereby enhancing the prediction ability of the model and the overall performance of the system.

[0100] Wavelet transform is a signal processing technology used to decompose signals, extract useful information, and remove noise at the same time. The wavelet transform method is used to denoise the data. The wavelet transform method effectively removes noise interference by decomposing and reconstructing the signal.

[0101] Removing noise in this step can improve the purity of the data and reduce misjudgments in subsequent analyses. Wavelet transform can effectively separate the high-frequency noise components in the signal and retain useful information.

[0102] Interpolation is a data filling method that estimates missing values based on known data points. Interpolation fills in missing data according to the trends before and after in the time series.

[0103] Filling in missing values in this step can avoid analysis biases caused by incomplete data. Interpolation can reasonably estimate missing values and maintain the time continuity of the data.

[0104] Linear normalization scales the data to a unified range (usually between 0 and 1) for subsequent analysis and processing. The linear normalization method is used to normalize the data.

[0105]

[0106] In this step, normalization processing can eliminate the influence of different dimensions and enable the data to be compared on the same scale. Linear normalization can simplify the subsequent calculation and analysis processes and improve the training efficiency of the model.

[0107] As a preferred embodiment of the present invention, according to the production rules, query the knowledge base to obtain a production task scheduling plan, specifically:

[0108] Based on the data obtained by real-time acquisition, through the similarity of the real-time data with the cluster classification of multiple scheduling plans in the knowledge base, obtain multiple preliminary scheduling plans corresponding to the production rules;

[0109] Based on the multiple preliminary scheduling plans, in combination with the corresponding execution situation data in the knowledge base, obtain a production task scheduling plan.

[0110] The main purpose of this embodiment is to match the real-time acquired data with the historical scheduling plans in the time series knowledge base, and identify the preliminary scheduling plans that best conform to the current production rules through similarity analysis. Then, in combination with the historical execution situation data of these preliminary plans, finally generate an optimal production task scheduling plan. This step aims to improve the accuracy and adaptability of scheduling decisions and ensure the efficient operation of the production process.

[0111] Real-time data processing and preliminary screening. Based on the equipment operation data, production task data, and environmental data obtained by real-time acquisition, use time series methods to analyze and obtain production rules. Ensure that the system can promptly capture new trends or changes in the production process and provide the latest basis for subsequent scheduling.

[0112] Compare the real-time data with multiple scheduling schemes stored in the knowledge base. Usually, the dynamic time warping (DTW) algorithm is used to measure the similarity between two time series, and based on this, several preliminary scheduling schemes closest to the current production pattern are found.

[0113]

[0114] Among them, X and U are two time series, and φ(i) represents the index of the i-th element in X corresponding to that in Y, which minimizes the total distance between the two series.

[0115] Through similarity matching, the scheduling scheme most suitable for the current situation can be quickly located from a large amount of historical data, reducing the time cost of re-designing the scheduling strategy.

[0116] Based on the above-mentioned preliminary scheduling schemes, further combined with the corresponding execution situation data in the knowledge base (such as task completion rate, resource utilization rate, abnormal event records, etc.), each scheme is comprehensively evaluated, and the best one is selected as the final production task scheduling scheme.

[0117] Multiple factors need to be considered during the evaluation, including but not limited to task priority, required resources, expected completion time, and potential risks, etc. By comprehensively evaluating the historical performance of each preliminary scheme, its actual effect under the current conditions can be predicted more accurately, thus making a more scientific and reasonable scheduling decision.

[0118] Through the above steps, not only the effective utilization of the production pattern is realized, but also the historical experience in the knowledge base is combined to generate a precise and highly adaptable production task scheduling scheme.

[0119] As a preferred embodiment of this implementation manner, the knowledge base is specifically:

[0120] The knowledge base contains various production patterns and the corresponding execution situations of production task scheduling schemes;

[0121] The production patterns include equipment operation modes, task execution modes, and production process relationships.

[0122] The main purpose of this embodiment is to construct and maintain a time series knowledge base containing various production patterns and the corresponding execution situations of production task scheduling schemes. This knowledge base not only records key information such as equipment operation modes, task execution modes, and production process relationships, but also continuously optimizes the scheduling strategy through historical data and real-time feedback to improve production efficiency and resource utilization rate.

[0123] In this embodiment, a comprehensive knowledge base is constructed, which contains different types of production rules (such as equipment operation modes, task execution modes, production process relationships) and the historical execution situations of the production task scheduling schemes corresponding to these rules. It provides a comprehensive data basis, enabling the system to make accurate matching and scheduling decisions based on real-time data in the current production environment.

[0124] The production rules are classified and stored according to equipment operation modes, task execution modes, and production process relationships. For example, equipment operation modes can include normal operation, high-temperature operation, low-load status, etc.; task execution modes can be continuous operation, intermittent operation, etc.; and production process relationships involve the dependency relationships between different processes.

[0125] Among them, the equipment operation mode is used to describe the working state or performance characteristics of the equipment within a specific time period. The task execution mode is used to describe the operation process or method followed during the execution of production tasks. The production process relationship is used to indicate the dependency relationship in the time sequence between different production tasks, ensuring that the next process can only start after the previous process is completed.

[0126] Generally speaking, this step realizes the effective management and utilization of production rules.

[0127] Specifically, the construction of the knowledge base is as follows:

[0128] According to historical data, time series embedding vectors are obtained through data conversion;

[0129] According to the time series embedding vectors, through clustering analysis, based on the Euclidean distance between the data points within the cluster and the cluster center, the production rule clusters of multiple scheduling schemes are obtained.

[0130] The main purpose of this embodiment is to construct a structured knowledge base through historical data, convert the original data into high-dimensional feature vectors using time series embedding technology, and then classify similar production rules into different clusters through clustering analysis. This process can systematically organize the patterns in historical data, provide an efficient and accurate classification basis for subsequent scheduling scheme matching, and thus improve the intelligent level of production task scheduling.

[0131] Converting the preprocessed data into time series embedding vectors, common methods include:

[0132] Based on the position encoding of Transformer, position-sensitive embedding vectors are generated through sine / cosine functions.

[0133]

[0134] Among them, pos is the time step position and d is the dimension of the embedding vector.

[0135] The delay embedding of the Havok-Method reconstructs the phase space of the time series by selecting appropriate embedding dimensions and delay times.

[0136] Deep learning models (such as LSTM / GRU) extract time-dependent features through sequence modeling.

[0137] The SEANet model combines dilated convolution and SOS constraints to generate compact embedding vectors.

[0138] This step captures long-term dependencies in the time series (such as device periodic failure patterns). Compresses the high-dimensional time series into low-dimensional vectors, reducing computational complexity.

[0139] Use clustering algorithms (such as K-means, hierarchical clustering, or DBSCAN) to group the embedding vectors and divide the cluster boundaries according to the Euclidean distance between the data points within the cluster and the cluster center.

[0140] Among them, the K-means algorithm is

[0141]

[0142] where v i is the i-th embedding vector, and c k is the k-th cluster center.

[0143] Through iterative optimization, minimize the total distance within the cluster.

[0144]

[0145] Hierarchical clustering is to merge / decompose clusters from bottom to top or from top to bottom through a dendrogram, which is suitable for scenarios with unknown number of clusters.

[0146] It can be understood that during the clustering process, the optimal number of clusters is selected through the Silhouette Score.

[0147]

[0148] The closer S is to 1, the better the clustering effect.

[0149] This step classifies similar production rules (such as device high-temperature operation mode, task sudden peak) into the same cluster, facilitating the quick matching of scheduling schemes. Through the representativeness of the cluster center, reduce the search range during real-time calculation.

[0150] Associate each cluster with the execution effects of historical scheduling schemes (such as task completion rate and resource utilization rate) to form a mapping relationship of "production law cluster - scheduling scheme". After embedding real-time data, quickly locate the optimal scheduling scheme by calculating the minimum distance to the cluster center. Combine historical execution data to improve the reliability of the scheduling scheme.

[0151] In a specific embodiment, for the equipment scheduling optimization of a semiconductor manufacturing factory, the SEANet model is adopted, and dilated convolution is combined to extract the characteristics of the equipment operation mode. The embedding dimension d = 64, and the stride of the dilated convolution increases exponentially (such as powers of 2). The loss function is

[0152]

[0153] where λ1 and λ2 are weight coefficients to balance the accuracy of compression and reconstruction.

[0154] Compression Error is the compression error, where V is the embedding vector and X is the original data.

[0155] Reconstruction Error is the reconstruction error, where X′ is the data after reconstruction and X is the original data.

[0156] The model output generates a 64-dimensional embedding vector for each equipment operation mode (such as "high temperature and high load" and "low load standby").

[0157] For the embedding vectors, through K-means clustering, the number of clusters K = 3 is determined by the silhouette coefficient. Cluster 1 is the high temperature and high load mode (temperature > 80 °C, load > 90%); Cluster 2 is the intermittent task mode (task interval > 2 hours); Cluster 3 is the continuous high-priority task mode (task priorities are all "urgent").

[0158] Cluster 1 corresponds to scheduling scheme A (increase the cooling interval and reduce the failure risk). Cluster 3 corresponds to scheduling scheme C (prioritize resource allocation to ensure the completion of urgent tasks).

[0159] The current temperature of equipment A is 85 °C and the load is 95%. The distance between the embedding vector and the center of Cluster 1 is the smallest (d = 0.12). Automatically trigger scheduling scheme A to adjust the cooling system to avoid overheating shutdown of the equipment.

[0160] This embodiment transforms the complex production laws in historical data into a structured knowledge base through time series embedding and clustering analysis, significantly improving the matching efficiency and accuracy of the scheduling scheme.

[0161] As a preferred embodiment of the present invention, according to the execution situation of the production task scheduling plan, the time series knowledge base is adjusted, specifically as follows:

[0162] According to the device operation data, production task data, environmental data obtained by real-time acquisition, and the execution feedback data of the production task scheduling plan, the execution feedback data at least includes any one of task completion time, resource utilization rate, abnormal event record, and manual intervention record;

[0163] According to the device operation data, production task data, and environmental data, a production task scheduling plan is obtained;

[0164] According to the execution feedback data, the execution situation of the production task scheduling plan is obtained to update and optimize the production task scheduling plan.

[0165] The main purpose of this embodiment is to dynamically update the production rule cluster classification and scheduling plan template in the time series knowledge base through the real-time acquired production data and execution feedback data, so as to realize the continuous optimization of production task scheduling. This process can improve the adaptability of the scheduling plan, reduce abnormal events, and improve resource utilization rate.

[0166] Real-time acquisition of device operation data, including device status (such as temperature, load, fault code), operation mode (such as high speed / low speed), energy consumption, etc.; production task data, including task type, priority, required resources (such as devices, raw materials), task completion time, etc.; environmental data, including workshop temperature, humidity, power supply stability, etc.

[0167] Execution feedback data, including task completion time, resource utilization rate (such as device idle rate), abnormal event record (such as device failure, material shortage), and manual intervention record (such as number of manual adjustments).

[0168] According to the current device, task and environmental data, convert them into time series embedding vectors for knowledge base retrieval. By calculating the Euclidean distance between the embedding vectors and the cluster centers in the knowledge base, the most matching production rule cluster is determined. Select the scheduling plan template with the best historical performance from the matching cluster. The cluster classification based on real-time data ensures a high degree of adaptation of the scheduling plan to the current production rules, and adjusts the scheduling plan according to the current execution feedback data.

[0169] According to the execution feedback data, compare the difference between the planned time and the actual completion time, such as

[0170] ΔT = T 实际 -T 计划

[0171] Calculate the resource utilization rate

[0172]

[0173] And the abnormal event frequency, which counts the number of abnormal events occurring within a unit time (such as the number of equipment failures). The degree of manual intervention is used to record the number of times and reasons for manual adjustment (such as the system automatically switching to the manual mode).

[0174] This step quickly locates the defects of the scheduling scheme through the correlation analysis between abnormal events and equipment / environment data. The quantification of key indicators provides a quantifiable basis for subsequent optimization.

[0175] According to the execution feedback data, the knowledge base is dynamically updated. Add the current production data and execution feedback data to the knowledge base, and update the time series embedding vector library. Through online learning or periodic offline training, recalculate the cluster centers and boundaries. Adjust the weights of the scheduling schemes within the clusters according to the execution feedback data (such as reducing the priority of the scheme with a high failure rate), and optimize the scheduling scheme. Based on the abnormal event records, update the abnormal thresholds of the equipment operation modes.

[0176] Through continuous updates, the knowledge base can adapt to long-term factors such as equipment aging and environmental changes. Through long-term optimization, avoid the problem of the scheduling scheme becoming invalid due to outdated historical data.

[0177] This implementation method realizes the closed-loop optimization of production task scheduling through real-time data acquisition, execution feedback analysis, and dynamic update of the knowledge base.

[0178] As a preferred implementation method under the present invention, when there are multiple production tasks, the production task scheduling method is specifically as follows:

[0179] Obtain multiple production task scheduling schemes corresponding to the multiple production tasks;

[0180] According to the multiple production task scheduling schemes, combined with the priorities of the multiple production tasks, obtain a production task scheduling strategy, where the production task scheduling strategy includes the task execution order, equipment resource allocation, and time window;

[0181] Among them, the priority of the production task is determined according to the task urgency, deadline constraint, and resource requirement weight.

[0182] The main purpose of this implementation method is to generate a globally optimal production task scheduling strategy when there are multiple production tasks by comprehensively considering task priorities, resource constraints, and the feasibility of the scheduling scheme. This strategy needs to clarify the task execution order, equipment resource allocation, and time window to maximize resource utilization, reduce task delays, and meet the priority requirements of urgent tasks.

[0183] In this embodiment, based on the detailed information of multiple production tasks (task type, required resources, processing time, priority), equipment resource status (available equipment, production capacity, maintenance plan), and environmental constraints (such as workshop capacity, energy limit), multiple candidate scheduling plans are generated, and each plan includes task sequence, resource allocation, and time window.

[0184] The priority is determined according to the task urgency, deadline constraint, and resource requirement weight. Weights are assigned according to the urgency level of the task (e.g., urgent task weight = 0.5). According to the deadline, the remaining time between the task deadline and the current time is calculated (remaining time = T 截止 -T 当前 ), and the shorter the remaining time, the higher the priority.

[0185] According to the resource requirement weight

[0186]

[0187] Tasks with higher resource requirements have lower priorities (to avoid resource competition).

[0188] Determine the comprehensive score

[0189]

[0190] Among them, α, β, γ are weight coefficients (e.g., α = 0.4, β = 0.3, γ = 0.3).

[0191] According to P i Tasks are sorted in descending order to form a priority queue.

[0192] This step avoids subjective judgment bias through quantitative decision-making and ensures the objectivity of priority allocation. Through weight coefficient adjustment, it can flexibly handle the insertion of urgent tasks or resource changes.

[0193] Filter out candidate plans that meet the priority requirements from the generated scheduling plans (e.g., high-priority tasks must be completed within the first 3 time windows). If a high-priority task competes with a low-priority task for the same equipment, the low-priority task is forcibly interrupted. Reserve the minimum resource quota for high-priority tasks.

[0194] Use Gantt charts to visualize task time arrangements and adjust the time window to eliminate resource bottlenecks.

[0195]

[0196] The higher the priority, the greater the adjustment range.

[0197] Finally, the task execution order is achieved (e.g., A→B→C). Device resource allocation (e.g., task A is allocated to device X, task B is allocated to device Y). Determine the time window (e.g., task A: 08:00 - 10:00, task B: 10:30 - 12:00).

[0198] In a specific embodiment, an automotive parts factory needs to process the following three tasks simultaneously:

[0199] Task A, an urgent order (produce 100 brake pads, deadline within 24 hours, requires device X). Task B, a regular order (produce 500 filters, deadline 48 hours later, requires device Y). Task C, a maintenance task (maintain device Z, requires 1 hour, no deadline).

[0200] Generate Plan 1: Prioritize completing tasks A and C, and delay task B. Task order A→C→B, device X / Y / Z allocation, time window 08:00 - 18:00.

[0201] Generate Plan 2: Balance the resource requirements of task B, and task A uses the idle period of device X. Task order A→B→C, device X / Y / Z allocation, time window 08:00 - 19:00.

[0202] For the quantitative evaluation of task priorities, for task A, the urgency level is (Level A, 0.5), the remaining time is 24 hours (weight 0.3 → 0.3 / 24 = 0.0125), and the resource requirement weight (device X occupancy rate 100% → 0.3 / 100 = 0.003).

[0203] P A = 0.4·0.5 + 0.3·0.0125 + 0.3·0.003 = 0.203

[0204] For task B, the urgency level is (Level B, 0.3), the remaining time is 48 hours (0.3 / 48 = 0.00625), and the resource requirement weight (device Y occupancy rate 80% → 0.3 / 80 = 0.00375).

[0205] P B = 0.4·0.3 + 0.3·0.00625 + 0.3·0.00375 = 0.123

[0206] For task C, the urgency level is (Level D, 0.1), and the resource requirement weight (device Z occupancy rate 10% → 0.3 / 10 = 0.03).

[0207] P C = 0.4·0.1 + 0.3·0 + 0.3·0.03 = 0.049

[0208] Determine the priority order as A > B > C.

[0209] Therefore, the task order is determined as A → B → C. For equipment allocation, task A → equipment X, task B → equipment Y, and task C → equipment Z. For time windows, task A: 08:00 - 12:00 (urgent completion); task B: 12:30 - 17:00 (utilizing the idle period of equipment Y); task C: 17:30 - 18:30 (low-priority task).

[0210] Tasks A and B are executed in parallel on different equipment to avoid contention between equipment X / Y. Task A is completed within 24 hours, and task B is completed within 48 hours, meeting the deadlines.

[0211] Through the multi-task scheduling scheme generation, priority quantification evaluation, and dynamic resource optimization in this embodiment, efficient scheduling in a complex production environment is achieved.

[0212] The present invention also provides a storage medium,

[0213] on which a computer program is stored, and when the computer program is executed, the steps of the intelligent production task scheduling method based on the time series knowledge base are implemented.

[0214] Therefore, any effects of the intelligent production task scheduling method based on the time series knowledge base can be achieved, which will not be elaborated here.

[0215] The present invention further provides a processing device, including:

[0216] a memory for storing a computer program;

[0217] a processor for implementing the steps of the intelligent production task scheduling method based on the time series knowledge base when executing the computer program.

[0218] Therefore, any effects of the intelligent production task scheduling method based on the time series knowledge base can be achieved, which will not be elaborated here.

[0219] What is not described in the present invention can be implemented by adopting or referring to the existing technologies.

[0220] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments.

[0221] The above are only the embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. An intelligent production task scheduling method based on a time series knowledge base, characterized in that, Including: Based on the device operation data, production task data, and environmental data collected in real time, analyze through time series methods to obtain production rules; According to the production rules, query the knowledge base to obtain a production task scheduling plan; Adjust the time series knowledge base according to the execution situation of the production task scheduling plan.

2. The intelligent production task scheduling method based on a time series knowledge base according to claim 1, wherein According to the production rules, query the knowledge base to obtain a production task scheduling plan, specifically: Based on the data collected in real time, obtain multiple preliminary scheduling plans corresponding to the production rules through the similarity of the real-time data and the cluster classification of multiple scheduling plans in the knowledge base; According to the multiple preliminary scheduling plans, combine the corresponding execution situation data in the knowledge base to obtain a production task scheduling plan.

3. The intelligent production task scheduling method based on time series knowledge base according to claim 2 is characterized in that: The knowledge base, specifically: The knowledge base contains various production rules and the corresponding execution situations of production task scheduling plans; The production rules include device operation modes, task execution modes, and production process relationships.

4. The intelligent production task scheduling method based on a time series knowledge base according to claim 3, wherein The construction of the knowledge base, specifically: Based on historical data, obtain time series embedding vectors through data conversion; According to the time series embedding vectors, through clustering analysis, obtain the cluster classification of production rules of multiple scheduling plans according to the Euclidean distance between the data points within the cluster and the cluster center.

5. The intelligent production task scheduling method based on a time series knowledge base according to claim 1, characterized in that The device operation data, production task data, and environmental data, specifically: The device operation data includes at least any one of the working state, temperature, load, and fault records of the device, The production task data includes at least any one of the execution requirements, priorities, and required resources of the production tasks, The environmental data includes at least any one of temperature and humidity, pressure, and air flow.

6. The intelligent production task scheduling method based on time series knowledge base according to claim 1 is characterized in that: Adjust the time series knowledge base according to the execution situation of the production task scheduling plan, specifically: Based on the device operation data, production task data, and environmental data collected in real time, and the execution feedback data of the production task scheduling plan, the execution feedback data includes at least any one of the task completion time, resource utilization rate, abnormal event records, and manual intervention records; Based on the device operation data, production task data, and environmental data, obtain a production task scheduling plan; Obtain the execution situation of the production task scheduling plan according to the execution feedback data to update and optimize the production task scheduling plan.

7. The intelligent production task scheduling method based on a time series knowledge base according to claim 1, wherein When there are multiple production tasks, the production task scheduling method, specifically: Obtain multiple production task scheduling plans corresponding to the multiple production tasks; According to the multiple production task scheduling plans, combine the priorities of the multiple production tasks to obtain a production task scheduling strategy, and the production task scheduling strategy includes the task execution order, device resource allocation, and time window; Among them, the priority of the production task is determined according to the task urgency, deadline constraint, and resource demand weight.

8. The intelligent production task scheduling method based on time series knowledge base according to claim 1 is characterized in that: Also including: Preprocess the collected data, and the preprocessing includes denoising, missing value filling, and normalization processing; The denoising processing includes at least the wavelet transform method, the missing value filling includes at least the interpolation method, and the normalization processing includes at least the linear normalization method.

9. A storage medium, characterized in that, A computer program is stored on the storage medium, and when the computer program is executed, the steps of the intelligent production task scheduling method based on the time series knowledge base according to any one of claims 1 to 8 are implemented.

10. A processing device, characterized in that: It includes: A memory for storing a computer program; A processor for implementing the steps of the intelligent production task scheduling method based on the time series knowledge base according to any one of claims 1 to 8 when executing the computer program.

Images