Intelligent workshop scheduling optimization system, method and equipment based on artificial intelligence and storage medium

Abstract

The invention provides an intelligent workshop scheduling optimization system, method and device based on artificial intelligence and a storage medium. The system comprises a data acquisition module, a data fusion and preprocessing module, an intelligent scheduling model training module, a real-time scheduling decision module and an execution monitoring module. The data acquisition module collects production equipment operation parameters, material circulation information, production order details and other data, and transmits the data to the data fusion and preprocessing module for cleaning, feature extraction and labeling. And the intelligent scheduling model training module constructs an optimization model by adopting a deep reinforcement learning algorithm based on the annotation data. And the real-time scheduling decision module receives real-time data, inputs a model to generate a scheduling instruction, and performs risk assessment and adjustment. And the execution monitoring module controls workshop production according to the instruction, collects feedback data to form closed-loop control, and realizes production dynamic optimization. The method can improve the production efficiency, reduce the cost, and enhance the robustness and adaptability of the scheduling scheme.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing technology, and more specifically, to an artificial intelligence-based intelligent workshop scheduling optimization system, method, device and storage medium. Background Art

[0002] In traditional workshop production, scheduling relies primarily on manual experience and simple rule-based engines. Schedulers must manually formulate production plans, taking into account multiple factors, including the operating status of production equipment, material availability, and the urgency of production orders. This approach is not only inefficient but also prone to errors, making it difficult to respond to various emergencies that arise during the production process, such as equipment failures and material shortages. Furthermore, with the continuous expansion of production scale and the increasing complexity of production tasks, the limitations of manual scheduling have become increasingly apparent, making it difficult to meet the high production efficiency and product quality requirements of modern manufacturing.

[0003] While existing shop floor scheduling systems incorporate some automation technologies, such as rule-based scheduling algorithms and simple optimization models, most of these systems can only handle relatively simple, deterministic scheduling problems. These systems often fail to provide effective solutions for complex, dynamically changing shop floor environments, such as those faced with high-mix, small-batch production. They often lack the ability to deeply mine and analyze real-time shop floor data, making it impossible to accurately predict the various uncertainties in the production process. This results in scheduling solutions with poor robustness and adaptability.

[0004] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the existing technology: first, the scheduling efficiency is low and it is impossible to quickly respond to changes in the production process; second, the scheduling accuracy is insufficient and it is difficult to formulate an optimal production plan; third, insufficient consideration is given to uncertainty factors in the production process, resulting in poor robustness and adaptability of the scheduling scheme; fourth, there is a lack of in-depth mining and analysis of real-time workshop data, and it is impossible to fully utilize data resources to optimize scheduling decisions. Summary of the Invention

[0005] The present invention provides an artificial intelligence-based intelligent workshop scheduling optimization system, method, device and storage medium.

[0006] In a first aspect of the present invention, an artificial intelligence-based intelligent workshop scheduling optimization system is provided, comprising:

[0007] Data collection module: This module collects in real time the operating parameters of production equipment in the workshop, material flow information, and production order details. The operating parameters of production equipment include equipment load rate, failure rate, and energy consumption indicators; material flow information includes material batch, specification, inventory level, and work-in-progress status; and production order details include order number, product model, delivery date, and quality requirements. The collected data is then transmitted to the data fusion and preprocessing module.

[0008] Data fusion and preprocessing module: including data cleaning submodule, feature engineering submodule and data labeling submodule;

[0009] The data cleaning submodule is used to perform in-depth cleaning of the collected workshop data, using data mining technology to identify and eliminate outliers, duplicate records, and missing values in the data, while also unifying and standardizing the data format;

[0010] The feature engineering submodule is used to mine high-dimensional features related to shop floor scheduling optimization from cleaned data. It uses dimensionality reduction techniques to extract key features, such as equipment overall efficiency, material supply stability, and order urgency coefficient, and constructs a feature vector space.

[0011] The data annotation submodule is used to accurately annotate feature vectors based on historical production data and expert experience, classifying scheduling results into excellent, good, fair, and poor grades, providing training samples for the supervised learning model. The annotation process uses a semi-automated annotation tool combined with manual review.

[0012] Intelligent Scheduling Model Training Module: Based on labeled feature vector data, a deep reinforcement learning algorithm is used to build a shop scheduling optimization model. The model architecture integrates a convolutional neural network to extract spatial features, a recurrent neural network to capture time series features, and an attention mechanism to highlight key information. Through interaction with the environment, the model parameters are continuously adjusted to learn the optimal scheduling strategy. This enables the model to make autonomous decisions in complex and changing shop floor environments, generate efficient scheduling solutions, and deploy the trained model to the real-time scheduling decision module.

[0013] Real-time scheduling decision module: Receives real-time data streams from the workshop and inputs them into a trained intelligent scheduling model. The model then outputs scheduling instructions in real time, including the equipment's job scheduling sequence, material delivery routes and time nodes, and personnel task allocation and collaboration arrangements. It also considers uncertainties in the production process and uses methods such as Monte Carlo simulation to conduct risk assessment and optimize adjustments to the scheduling plan, ensuring its robustness and adaptability. The finalized scheduling instructions are then sent to the execution monitoring module.

[0014] Execution monitoring module: Based on the received scheduling instructions, the production equipment, material transportation system and operators in the workshop are precisely controlled and monitored in real time. The Internet of Things technology is used to realize remote control and status monitoring of equipment. The automated warehousing and logistics system is used to ensure the accurate distribution of materials. The personnel positioning and task management system is used to ensure that personnel complete tasks on time. At the same time, feedback data during the execution process is collected, such as the actual operating efficiency of the equipment, the real-time inventory level of materials, and the completion progress of the task. The feedback data is sent back to the data acquisition module to form a closed-loop control and realize dynamic optimization and continuous improvement of workshop production.

[0015] Furthermore, the operating parameters of the production equipment are deeply cleaned, specifically:

[0016] Collect operating parameter data of all production equipment in the workshop during the continuous production cycle. The data dimensions include equipment start and stop time, operating speed, processing accuracy, and energy consumption;

[0017] Use time series analysis to detect anomalies in the equipment's start and stop time data, identify abnormal start and stop records that do not conform to production rules, and use an outlier repair algorithm based on a Gaussian mixture model, combined with the equipment's historical normal start and stop patterns, to reasonably correct abnormal start and stop times.

[0018] The running speed data of the equipment is subjected to noise filtering. Wavelet transform combined with adaptive threshold algorithm is used to decompose the running speed data into sub-band signals of different frequencies, retaining the low-frequency useful signals and removing the high-frequency noise interference. At the same time, according to the processing requirements of the equipment, the filtered running speed data is normalized by interval processing. The formula is:

[0019]

[0020] Among them, v is the original running speed data of the device, v min is the minimum value of the device running speed data, v max is the maximum value of the device operating speed data, v mean is the mean of the equipment running speed data, σ is the standard deviation of the equipment running speed data, v norm is the normalized running speed data;

[0021] For the equipment's processing accuracy data, a missing value filling method based on Bayesian network is adopted to construct a Bayesian network model of the equipment's processing accuracy. The type of equipment, the material of the workpiece, and the processing parameters are used as network nodes. The conditional probability distribution of the network is learned according to the known processing accuracy data. For the missing processing accuracy data, the Bayesian inference algorithm is used, combined with the network's prior knowledge and observation data, to calculate the posterior probability distribution of the missing data, thereby obtaining the filling value of the missing data.

[0022] Furthermore, feature extraction is performed on the material flow information, specifically:

[0023] Integrate the batch information, specification parameters, inventory dynamics and work-in-progress flow status of each material in the workshop to build a multi-dimensional data matrix of material flow information;

[0024] To calculate the dynamic inventory turnover rate of materials, the weighted moving average method is combined with the exponential smoothing technology. Different weight coefficients are assigned according to the importance and demand volatility of different materials. The formula is:

[0025]

[0026] Among them, T turn is the dynamic inventory turnover rate of the material, w i is the weight coefficient of the i-th material, D i is the outbound quantity of the i-th material in the statistical period, I i is the average inventory of the i-th material in the statistical period, α and β are exponential smoothing parameters, and t is the time variable;

[0027] Analyze the flow path and processing cycle of materials in the process, use the shortest path algorithm and critical path method in graph theory to construct a network diagram of the flow of materials in the process, calculate the average processing cycle, longest processing path and bottleneck process of each material in the process from raw material input to finished product output, and thus obtain the key characteristic parameters of the flow of materials in the process, including the average residence time of materials in the process, the waiting time ratio between processes, and the capacity utilization rate of key processes.

[0028] Furthermore, feature extraction is performed on the production order details, specifically:

[0029] Sorting out the full lifecycle information of production orders, including the order source, signing date, delivery requirements, product technical parameters, and customer feedback, to construct a feature vector for the production order;

[0030] To quantify the urgency of production orders, we use the fuzzy comprehensive evaluation method combined with the analytic hierarchy process to establish an evaluation index system for order urgency. The indicators include order delivery lead time, customer priority, and product importance. Through expert scoring and weight calculation, we obtain the fuzzy comprehensive evaluation matrix of order urgency. The formula is:

[0031]

[0032] Among them, E is the fuzzy comprehensive evaluation matrix of order urgency, e ijIt represents the fuzzy membership of the i-th order on the j-th evaluation index, n is the number of orders, m is the number of evaluation indicators, and the comprehensive score of the urgency of each order is calculated according to the fuzzy comprehensive evaluation matrix;

[0033] Analyze the product technical parameters of the production order, use semantic analysis technology and data mining algorithms to extract the association rules between the product's key performance indicators and production process requirements, build a feature mapping table of the product technical parameters, and classify and mark the orders according to the complexity of the product technical parameters and special requirements.

[0034] Furthermore, a workshop scheduling optimization model is constructed, specifically:

[0035] The extracted high-dimensional feature data such as equipment comprehensive efficiency, material supply stability, and order urgency coefficient are used as input variables of the model. Minimizing production costs, maximizing production efficiency, and ensuring on-time delivery are used as multi-objective optimization functions to construct a workshop scheduling optimization model based on deep reinforcement learning. The loss function of the model adopts the weighted sum form of multi-objective optimization, and the formula is:

[0036] L=λ1L cost +λ2L efficiency +λ3L delivery

[0037] Among them, L is the total loss function of the model, L cost is the production cost loss, L efficiency is the production efficiency loss, L delivery is the on-time delivery rate loss, λ1, λ2, and λ3 are the corresponding weight coefficients, and the weight coefficients are dynamically adjusted through the genetic algorithm;

[0038] The model's network architecture adopts an encoder-decoder structure. The encoder part consists of a multi-layer convolutional neural network and a long short-term memory network, which is used to encode the spatiotemporal features of the input high-dimensional feature data; the decoder part uses a gated recurrent unit combined with an attention mechanism to generate the optimal scheduling instruction sequence based on the feature vector output by the encoder. At the same time, a multi-agent reinforcement learning mechanism is introduced to simulate the collaborative scheduling behavior between multiple production units in the workshop and optimize the global scheduling strategy.

[0039] Furthermore, a production scheduling plan is generated, specifically:

[0040] Obtain real-time data on the workshop's production equipment status, material inventory levels, work-in-progress information, and pending production order data, and input them into the trained workshop scheduling optimization model;

[0041] The model uses an internal deep reinforcement learning algorithm to intelligently schedule production equipment. Taking into account equipment load balancing, preventive maintenance, and process adaptability, it generates a Gantt chart for equipment scheduling, determines the production tasks, workpieces, and process parameter settings for each piece of equipment in different time periods, and uses a meta-heuristic algorithm to perform local search optimization on the equipment scheduling results.

[0042] Based on the inventory status of materials, the flow of work-in-process products, and the bill of materials of the production order, the model plans the material delivery route and delivery time window. It uses the ant colony optimization algorithm combined with the vehicle routing problem model to consider factors such as traffic congestion, loading and unloading time, and delivery costs during the material delivery process to generate the optimal material delivery route planning diagram;

[0043] Taking into account the urgency of production orders, the complexity of product processes, and the matching of personnel skills, the model prioritizes production tasks and assigns tasks to personnel. It uses the Hungarian algorithm to solve the optimal matching problem between personnel and tasks and formulates a detailed production task execution plan, including the start time, end time, assigned equipment, operators, and quality inspection nodes of the task.

[0044] Furthermore, the production scheduling plan is implemented, specifically:

[0045] After receiving the generated scheduling plan, the execution monitoring module remotely controls and monitors the production equipment in the workshop in real time through the Industrial Internet of Things platform. Based on the equipment's operation scheduling Gantt chart, it accurately controls the equipment's start and stop, operating speed, processing parameters, etc. At the same time, it collects the equipment's real-time operating data and uses machine learning algorithms to conduct real-time diagnosis of the equipment's operating status and provide fault warnings.

[0046] Utilize automated warehousing and logistics systems to automatically dispatch forklifts, AGVs, or automated conveyor belts based on material distribution route planning to achieve automated material handling and precise distribution. At the same time, use radio frequency identification technology and visual recognition systems to track and monitor the material distribution process in real time.

[0047] With the help of the personnel positioning and task management system, task assignment information is pushed to operators in real time according to the production task execution plan. Through smart bracelets or mobile terminal devices, the location and work status of personnel are monitored and managed in real time. At the same time, the personnel's operation data and feedback information are collected and transmitted back to the data acquisition module.

[0048] In a second aspect of the present invention, a method for optimizing intelligent workshop scheduling based on artificial intelligence is provided, comprising:

[0049] During the pre-preparation phase, we collect real-time data on various aspects of production equipment in the workshop, including operating parameters, material flow information, and production order details. We then use data processing technology to perform deep cleaning, feature extraction, and annotation on the data to construct a training dataset.

[0050] During the model building and training phase, based on the labeled feature vector data, a deep reinforcement learning algorithm is used to integrate convolutional neural networks, recurrent neural networks, and attention mechanisms to build a shop floor scheduling optimization model. Through interactive simulation with the virtual production environment, the model parameters are continuously adjusted to learn the optimal scheduling strategy.

[0051] During the real-time scheduling decision-making phase, real-time collected shop floor data is fed into a trained model. The model then outputs real-time scheduling instructions covering multiple dimensions, including equipment operation scheduling, material distribution planning, and personnel task allocation. Risk assessment and adjustment optimization algorithms are also used to address uncertainties in the production process.

[0052] During the execution monitoring stage, according to scheduling instructions, the industrial Internet of Things, automated logistics systems, and personnel management systems are used to accurately control and monitor workshop production in real time. At the same time, execution feedback data is collected to form a closed-loop control, thereby achieving dynamic optimization and continuous improvement of workshop production.

[0053] In a third aspect of the present invention, an electronic device is provided, comprising: at least one processor, a memory, and an input / output unit; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute any one of the methods described in the first aspect.

[0054] In a fourth aspect of the present invention, a computer-readable storage medium is provided, comprising instructions, which, when executed on a computer, enable the computer to execute the method according to any one of the first aspects.

[0055] The above-described embodiments of the present invention have at least the following beneficial effects: The intelligent workshop scheduling optimization system of the present invention can collect and process large amounts of workshop data in real time, including production equipment operating parameters, material flow information, and production order details. Through deep data mining and analysis, it accurately extracts high-dimensional features related to scheduling optimization, providing rich and accurate data support for model training. The system uses a deep reinforcement learning algorithm to construct a workshop scheduling optimization model, integrating convolutional neural networks, recurrent neural networks, and an attention mechanism. It can effectively extract spatial features, capture time series features, and highlight key information, thereby learning the optimal scheduling strategy, generating an efficient scheduling plan, improving production efficiency, and reducing costs.

[0056] In addition, the system can also consider the uncertainty factors in the production process and use methods such as Monte Carlo simulation to conduct risk assessment and adjustment optimization of the scheduling plan to ensure the robustness and adaptability of the scheduling plan. The real-time scheduling decision module can receive real-time data streams from the workshop and quickly output scheduling instructions covering multiple dimensions such as equipment operation scheduling, material distribution planning, and personnel task allocation, thereby achieving precise control and real-time monitoring of workshop production. The execution monitoring module uses Internet of Things technology to achieve remote control and status monitoring of equipment, uses automated warehousing and logistics systems to ensure the accurate distribution of materials, and uses personnel positioning and task management systems to ensure that personnel complete tasks on time. At the same time, it collects feedback data during the execution process to form a closed-loop control, further realizing dynamic optimization and continuous improvement of workshop production. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily apparent by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation, in which:

[0058] Figure 1 A schematic diagram of the structure of an artificial intelligence-based intelligent workshop scheduling optimization system provided by one embodiment of the present invention;

[0059] Figure 2 A flowchart of an artificial intelligence-based intelligent workshop scheduling optimization method provided by one embodiment of the present invention;

[0060] Figure 3 The figure schematically shows the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided solely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. Rather, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0062] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software.

[0063] It should be noted that any number of elements in the drawings is for illustration only and not for limitation, and any naming is only for distinction and does not have any limiting meaning.

[0064] Reference below Figure 1 , Figure 1 This is a schematic diagram of the structure of an intelligent workshop scheduling optimization system based on artificial intelligence provided by an embodiment of the present invention. Figure 1 As shown, an artificial intelligence-based intelligent workshop scheduling optimization system 100 includes:

[0065] Data collection module 101: This module collects in real time the operating parameters of production equipment in the workshop, material flow information, and production order details. The operating parameters of production equipment include equipment load rate, failure rate, and energy consumption indicators; material flow information includes material batch, specification, inventory level, and work-in-progress status; and production order details include order number, product model, delivery date, and quality requirements. The collected data is then transmitted to the data fusion and preprocessing module.

[0066] Data fusion and preprocessing module 102: including data cleaning submodule, feature engineering submodule and data labeling submodule;

[0067] The data cleaning submodule is used to perform in-depth cleaning of the collected workshop data, using data mining technology to identify and eliminate outliers, duplicate records, and missing values in the data, while also unifying and standardizing the data format;

[0068] The feature engineering submodule is used to mine high-dimensional features related to shop floor scheduling optimization from cleaned data. It uses dimensionality reduction techniques to extract key features, such as equipment overall efficiency, material supply stability, and order urgency coefficient, and constructs a feature vector space.

[0069] The data annotation submodule is used to accurately annotate feature vectors based on historical production data and expert experience, classifying scheduling results into excellent, good, fair, and poor grades, providing training samples for the supervised learning model. The annotation process uses a semi-automated annotation tool combined with manual review.

[0070] Intelligent Scheduling Model Training Module 103: Based on the labeled feature vector data, a deep reinforcement learning algorithm is used to build a shop scheduling optimization model. The model architecture integrates a convolutional neural network to extract spatial features, a recurrent neural network to capture time series features, and an attention mechanism to highlight key information. Through interaction with the environment, the model parameters are continuously adjusted to learn the optimal scheduling strategy. This enables the model to make autonomous decisions in complex and changing shop floor environments, generate efficient scheduling solutions, and deploy the trained model to the real-time scheduling decision module.

[0071] Real-time scheduling decision module 104: Receives real-time data streams from the workshop and inputs them into a trained intelligent scheduling model. The model then outputs scheduling instructions in real time, including the equipment's job scheduling sequence, material delivery routes and time nodes, and personnel task allocation and collaboration arrangements. Taking into account uncertainties in the production process, it uses Monte Carlo simulation and other methods to conduct risk assessment and adjust and optimize the scheduling plan to ensure its robustness and adaptability. The finalized scheduling instructions are then sent to the execution monitoring module.

[0072] Execution monitoring module 105: Based on the received scheduling instructions, the production equipment, material transportation system and operators in the workshop are precisely controlled and monitored in real time. The Internet of Things technology is used to realize remote control and status monitoring of the equipment. The automated warehousing and logistics system is used to ensure the accurate distribution of materials. The personnel positioning and task management system is used to ensure that personnel complete tasks on time. At the same time, feedback data during the execution process is collected, such as the actual operating efficiency of the equipment, the real-time inventory level of materials, and the completion progress of the task. The feedback data is sent back to the data acquisition module to form a closed-loop control, thereby realizing dynamic optimization and continuous improvement of workshop production.

[0073] It's important to note that the data acquisition module is the foundation of the entire system. It's responsible for collecting real-time information on the operating parameters of production equipment within the workshop, material flow information, and production order details. Equipment operating parameters include load rate, failure rate, and energy consumption, reflecting the equipment's operational status and efficiency. Material flow information includes information such as batch, specification, inventory level, and work-in-progress status. This information is crucial for ensuring the timely and accurate supply of materials. Production order details include order number, product model, delivery date, and quality requirements. This information helps the system understand the specific requirements and priorities of production tasks.

[0074] Specifically, the operating parameters of production equipment can be collected through sensors and monitoring devices installed on the equipment. For example, load rate can be calculated by monitoring the equipment's power consumption and operating hours; failure rate can be calculated by recording the number of equipment failures and operating hours; and energy consumption indicators can be obtained through metering devices such as electricity and water meters. Material flow information, including operations such as material inbound and outbound storage and inventory counts, can be tracked through warehouse management systems and production execution systems. Production order details can be obtained from the company's order management system, including order creation, modification, and completion status information.

[0075] Preferably, the data collection module can utilize a variety of data collection technologies, such as IoT and industrial internet protocols, to ensure real-time and accurate data. For example, wireless sensor networks can be used to collect equipment operating parameters, RFID technology can be used to track material flow, and ERP systems can be used to obtain production order details. Furthermore, to improve the efficiency and reliability of data collection, the frequency and accuracy of data collection can be set, for example, collecting equipment operating parameters every minute and updating material inventory levels every hour.

[0076] It should be noted that the data fusion and preprocessing module is the core link of data processing, which includes a data cleaning submodule, a feature engineering submodule, and a data labeling submodule. The data cleaning submodule is used to deeply clean the collected workshop data, identify and eliminate outliers, duplicate records, and missing values in the data through data mining technology, and unify and standardize the data format. The feature engineering submodule is used to mine high-dimensional features related to workshop scheduling optimization from the cleaned data, use dimensionality reduction technology to extract key features such as equipment comprehensive efficiency, material supply stability, and order urgency coefficient, and construct a feature vector space. The data labeling submodule is used to accurately label feature vectors based on production history data and expert experience, and classify scheduling results into excellent, good, medium, and poor levels to provide training samples for the supervised learning model. The labeling process uses a semi-automatic labeling tool combined with manual review.

[0077] Specifically, the data cleaning submodule can use a variety of data mining techniques, such as time series analysis and outlier detection algorithms. For example, for the start and stop time data of the equipment, the time series analysis method can be used to identify abnormal start and stop records, and the outlier repair algorithm based on the Gaussian mixture model can be used to correct them. For the running speed data of the equipment, the wavelet transform can be used in combination with the adaptive threshold algorithm to perform noise filtering and interval normalization. The feature engineering submodule can use dimensionality reduction techniques, such as principal component analysis (PCA) and linear discriminant analysis (LDA), to extract key features from high-dimensional data. The data labeling submodule can combine production history data and expert experience, use semi-automatic labeling tools for preliminary labeling, and then ensure the accuracy of the labeling through manual review.

[0078] Preferably, the data cleaning submodule can set specific cleaning rules and parameters, such as outlier thresholds and duplicate record deletion criteria. The feature engineering submodule can further refine feature extraction methods, such as using different combinations of dimensionality reduction algorithms or introducing new feature extraction methods, such as deep feature extraction. The data annotation submodule can utilize more advanced annotation technologies, such as machine learning-based automatic annotation algorithms, combined with manual review, to improve annotation efficiency and accuracy.

[0079] It should be noted that the intelligent scheduling model training module is the core component of the system. Based on labeled feature vector data, it uses a deep reinforcement learning algorithm to construct a shop scheduling optimization model. The model architecture integrates a convolutional neural network to extract spatial features, a recurrent neural network to capture time series features, and an attention mechanism to highlight key information. By continuously adjusting model parameters through interaction with the environment, the model learns the optimal scheduling strategy, enabling it to make autonomous decisions in complex and changing shop floor environments, generate efficient scheduling solutions, and deploy the trained model to the real-time scheduling decision module.

[0080] Specifically, deep reinforcement learning algorithms can include Q-learning, DQN (Deep Q-Network), and PPO (Proximal Policy Optimization). Within the model architecture, convolutional neural networks (CNNs) can be used to process image or spatial data, extracting spatial features such as equipment layout and material distribution. Recurrent neural networks (RNNs) or long short-term memory networks (LSTMs) can be used to process time series data, capturing temporal features such as equipment operating status and material flow. Attention mechanisms can be used to highlight key information, such as urgent orders and faulty equipment. The model training process can be simulated interactively with a virtual production environment, continuously adjusting model parameters and optimizing scheduling strategies.

[0081] Preferably, the intelligent scheduling model training module can further refine the model training process, for example, by setting specific training parameters such as learning rate and discount factor. A multi-agent reinforcement learning mechanism can be employed to simulate the coordinated scheduling behavior between multiple production units within a workshop and optimize the global scheduling strategy. Furthermore, transfer learning techniques can be introduced to transfer existing scheduling experience to new production environments, improving the model's adaptability and generalization capabilities.

[0082] In some embodiments, the operating parameters of the production equipment are deeply cleaned, specifically:

[0083] Collect operating parameter data of all production equipment in the workshop during the continuous production cycle. The data dimensions include equipment start and stop time, operating speed, processing accuracy, and energy consumption;

[0084] Use time series analysis to detect anomalies in the equipment's start and stop time data, identify abnormal start and stop records that do not conform to production rules, and use an outlier repair algorithm based on a Gaussian mixture model, combined with the equipment's historical normal start and stop patterns, to reasonably correct abnormal start and stop times.

[0085] The running speed data of the equipment is subjected to noise filtering. Wavelet transform combined with adaptive threshold algorithm is used to decompose the running speed data into sub-band signals of different frequencies, retaining the low-frequency useful signals and removing the high-frequency noise interference. At the same time, according to the processing requirements of the equipment, the filtered running speed data is normalized by interval processing. The formula is:

[0086]

[0087] Among them, v is the original running speed data of the device, v min is the minimum value of the device running speed data, v max is the maximum value of the device operating speed data, v mean is the mean of the equipment running speed data, σ is the standard deviation of the equipment running speed data, v norm is the normalized running speed data;

[0088] For the equipment's processing accuracy data, a missing value filling method based on Bayesian network is adopted to construct a Bayesian network model of the equipment's processing accuracy. The type of equipment, the material of the workpiece, and the processing parameters are used as network nodes. The conditional probability distribution of the network is learned according to the known processing accuracy data. For the missing processing accuracy data, the Bayesian inference algorithm is used, combined with the network's prior knowledge and observation data, to calculate the posterior probability distribution of the missing data, thereby obtaining the filling value of the missing data.

[0089] It's important to note that deep cleaning of production equipment operating parameters is performed to ensure data accuracy and reliability, thereby providing high-quality data support for subsequent scheduling optimization. Equipment operating parameters include start and stop times, operating speed, machining accuracy, and energy consumption, reflecting the equipment's operational status and efficiency. The deep cleaning process involves steps such as anomaly detection, noise filtering, and data normalization. The goal is to remove outliers, duplicate records, and missing values from the data, while also unifying and standardizing the data format.

[0090] Specifically, the start and stop time data of the equipment can be used for anomaly detection through time series analysis methods. For example, the sliding window technology in time series analysis can be used to identify abnormal start and stop records that do not conform to production rules. For these abnormal records, an outlier repair algorithm based on the Gaussian mixture model can be used for correction. This algorithm combines the historical normal start and stop patterns of the equipment to make reasonable corrections to the abnormal start and stop times. The running speed data of the equipment can be processed by noise filtering through wavelet transform combined with adaptive threshold algorithm. Specifically, the running speed data is decomposed into sub-band signals of different frequencies, retaining the low-frequency useful signals and removing the high-frequency noise interference. The filtered running speed data is then interval normalized according to the processing requirements of the equipment. The formula is:

[0091]

[0092] Among them, v is the original running speed data of the device, v min is the minimum value of the device running speed data, v max is the maximum value of the device operating speed data, v mean is the mean of the equipment running speed data, σ is the standard deviation of the equipment running speed data, v norm is the normalized running speed data.

[0093] Preferably, a Bayesian network-based missing value filling method can be used for equipment machining accuracy data. First, a Bayesian network model of equipment machining accuracy is constructed, with the equipment type, workpiece material, and machining process parameters as network nodes. The network's conditional probability distribution is learned based on the known machining accuracy data. For missing machining accuracy data, a Bayesian inference algorithm is used, combining the network's prior knowledge and observed data to calculate the posterior probability distribution of the missing data, thereby obtaining the missing data filling value.

[0094] Furthermore, more data cleaning techniques can be introduced, such as cluster analysis to identify abnormal clusters in the data, or using autoencoders in deep learning for data noise reduction and feature extraction to further improve the effect of data cleaning.

[0095] In some embodiments, feature extraction is performed on material flow information, specifically:

[0096] Integrate the batch information, specification parameters, inventory dynamics and work-in-progress flow status of each material in the workshop to build a multi-dimensional data matrix of material flow information;

[0097] To calculate the dynamic inventory turnover rate of materials, the weighted moving average method is combined with the exponential smoothing technology. Different weight coefficients are assigned according to the importance and demand volatility of different materials. The formula is:

[0098]

[0099] Among them, T turn is the dynamic inventory turnover rate of the material, w i is the weight coefficient of the i-th material, D i is the outbound quantity of the i-th material in the statistical period, I i is the average inventory of the i-th material in the statistical period, α and β are exponential smoothing parameters, and t is the time variable;

[0100] Analyze the flow path and processing cycle of materials in the process, use the shortest path algorithm and critical path method in graph theory to construct a network diagram of the flow of materials in the process, calculate the average processing cycle, longest processing path and bottleneck process of each material in the process from raw material input to finished product output, and thus obtain the key characteristic parameters of the flow of materials in the process, including the average residence time of materials in the process, the waiting time ratio between processes, and the capacity utilization rate of key processes.

[0101] It's important to note that feature extraction of material flow information aims to extract key information relevant to shop floor scheduling optimization from a vast amount of material data. This information includes material batches, specifications, inventory dynamics, and work-in-process flow status. This information is crucial for optimizing material distribution routes and improving the stability of material supply. By constructing a multidimensional data matrix and calculating key feature parameters, we can provide richer input data for intelligent scheduling models, thereby improving the accuracy and efficiency of scheduling solutions.

[0102] Specifically, we first integrate the batch information, specification parameters, inventory dynamics, and work-in-process flow status of each material in the workshop to construct a multi-dimensional data matrix of material flow information. For example, batch information can include the production date and supplier of the material; specification parameters can include the size and weight of the material; inventory dynamics can include the incoming and outgoing quantities, inventory levels, etc.; and the work-in-process flow status can include the current location and processing status of the work-in-process. Next, we calculate the dynamic inventory turnover rate of the material using the weighted moving average method combined with exponential smoothing technology, assigning different weight coefficients according to the importance and demand volatility of different materials. The formula is:

[0103]

[0104] Among them, turnover rate 动态 is the dynamic inventory turnover rate of the material, w i is the weight coefficient of the i-th material, the outbound quantity i is the outbound quantity of the i-th material in the statistical period, the average inventory i is the average inventory of the i-th material during the statistical period, α and Δt are exponential smoothing parameters, and Δt is the time variable.

[0105] Preferably, the flow paths and processing cycles of materials in the work-in-process (WIP) can be analyzed, and the shortest path algorithm and critical path method from graph theory can be applied to construct a network diagram of the material WIP flow. Specifically, the shortest path algorithm can be used to calculate the shortest path from raw material input to finished product output, and the critical path method can be used to identify bottleneck processes in the processing. Using these algorithms, the average processing cycle, the longest processing path, and the capacity utilization rate of bottleneck processes for each material WIP can be calculated, thereby obtaining key characteristic parameters of the material WIP flow, including the average residence time of the material WIP, the waiting time ratio between processes, and the capacity utilization rate of key processes.

[0106] Furthermore, more feature extraction methods can be introduced, such as cluster analysis to identify the similarity of materials, or convolutional neural networks in deep learning can be used for feature extraction to further improve the effect of feature extraction.

[0107] In some embodiments, feature extraction is performed on the production order details, specifically:

[0108] Sorting out the full lifecycle information of production orders, including the order source, signing date, delivery requirements, product technical parameters, and customer feedback, to construct a feature vector for the production order;

[0109] To quantify the urgency of production orders, we use the fuzzy comprehensive evaluation method combined with the analytic hierarchy process to establish an evaluation index system for order urgency. The indicators include order delivery lead time, customer priority, and product importance. Through expert scoring and weight calculation, we obtain the fuzzy comprehensive evaluation matrix of order urgency. The formula is:

[0110]

[0111] Among them, E is the fuzzy comprehensive evaluation matrix of order urgency, e ij It represents the fuzzy membership of the i-th order on the j-th evaluation index, n is the number of orders, m is the number of evaluation indicators, and the comprehensive score of the urgency of each order is calculated according to the fuzzy comprehensive evaluation matrix;

[0112] Analyze the product technical parameters of the production order, use semantic analysis technology and data mining algorithms to extract the association rules between the product's key performance indicators and production process requirements, build a feature mapping table of the product technical parameters, and classify and mark the orders according to the complexity of the product technical parameters and special requirements.

[0113] It's crucial to compile information about the entire lifecycle of production orders. Orders can originate from a variety of sources, including direct customer orders and distributor orders. The signing date formalizes the order. Delivery requirements include product quantity, quality standards, and delivery location. Product technical parameters cover dimensions, performance indicators, and raw material specifications. Customer feedback includes customer satisfaction with previous orders and suggestions for improvement. By integrating this information, a vector comprehensively reflecting the characteristics of production orders can be constructed, providing foundational data support for subsequent analysis and scheduling.

[0114] Specifically, when it comes to quantifying the urgency of production orders, the fuzzy comprehensive evaluation method, based on fuzzy mathematics, is used. It is a comprehensive evaluation method capable of handling multi-factor, ambiguity, and uncertainty. The analytic hierarchy process (AHP) is a decision-making method that decomposes decision-making elements into hierarchical levels, such as goals, criteria, and plans, and then conducts qualitative and quantitative analysis based on these levels. The delivery lead time for an order refers to the time interval between order signing and required delivery. Customer priority can be determined based on factors such as customer cooperation history, order amount, and credit rating. The importance of a product can be determined based on its market scarcity and contribution to corporate profits.

[0115] Furthermore, in actual operations, experts score each indicator based on their own experience and the company's actual situation. For example, orders with short delivery lead times, high customer priority, and high product importance are likely to receive high scores on the corresponding indicators. By rationally determining the weights of each evaluation indicator, a fuzzy comprehensive evaluation method is used to calculate the overall urgency score for each order, providing a basis for order prioritization.

[0116] More specifically, when parsing product technical parameters in production orders, semantic analysis technology can leverage natural language processing tools to analyze the descriptive text of product technical parameters and extract key information. For example, if a product requires high-strength aluminum alloy, machining accuracy must reach ±0.01mm, and surface roughness must be less than Ra0.8, key performance indicators such as ±0.01mm machining accuracy and Ra0.8 surface roughness can be accurately extracted. Data mining algorithms can analyze product technical parameters and production process data from a large number of historical orders to uncover association rules between key performance indicators of different products and production process requirements. For example, they can identify the relationship between specific performance indicator requirements of a product and specific processing techniques and equipment parameters. Order classification and tagging can be categorized into simple, general, and complex based on the complexity of product technical parameters. Special requirements can be categorized into categories such as those with special environmental requirements and those with special safety requirements. This allows for tailored strategies to be adopted for different types of orders during production scheduling, improving production efficiency and product quality.

[0117] In some embodiments, a workshop scheduling optimization model is constructed, specifically:

[0118] The extracted high-dimensional feature data such as equipment comprehensive efficiency, material supply stability, and order urgency coefficient are used as input variables of the model. Minimizing production costs, maximizing production efficiency, and ensuring on-time delivery are used as multi-objective optimization functions to construct a workshop scheduling optimization model based on deep reinforcement learning. The loss function of the model adopts the weighted sum form of multi-objective optimization, and the formula is:

[0119] L=λ1L cost +λ2L efficiency +λ3L delivery

[0120] Among them, L is the total loss function of the model, L cost is the production cost loss, L efficiency is the production efficiency loss, L delivery is the on-time delivery rate loss, λ1, λ2, and λ3 are the corresponding weight coefficients, and the weight coefficients are dynamically adjusted through the genetic algorithm;

[0121] The model's network architecture adopts an encoder-decoder structure. The encoder part consists of a multi-layer convolutional neural network and a long short-term memory network, which is used to encode the spatiotemporal features of the input high-dimensional feature data; the decoder part uses a gated recurrent unit combined with an attention mechanism to generate the optimal scheduling instruction sequence based on the feature vector output by the encoder. At the same time, a multi-agent reinforcement learning mechanism is introduced to simulate the collaborative scheduling behavior between multiple production units in the workshop and optimize the global scheduling strategy.

[0122] It is important to note the importance of using extracted high-dimensional feature data such as equipment overall efficiency, material supply stability, and order urgency coefficient as model input variables. Equipment overall efficiency is an indicator derived by comprehensively considering factors such as equipment availability, performance efficiency, and product quality qualification rate, reflecting the overall effectiveness of the equipment in the production process. Material supply stability involves factors such as the timeliness of material supply and inventory fluctuations. A stable material supply ensures production continuity. The order urgency coefficient reflects the urgency of orders and plays a key role in prioritizing production scheduling. Using these feature data as input enables the model to more accurately grasp the actual production situation on the shop floor, providing a strong basis for formulating optimization strategies.

[0123] Specifically, in a multi-objective optimization function, production costs include raw material procurement costs, equipment operating costs, and labor costs. For example, raw material procurement costs vary depending on the type of raw material, price fluctuations, and procurement volume; equipment operating costs are related to energy consumption, maintenance costs, and other factors; and labor costs depend on the number of workers, working hours, and wage levels. Production efficiency can be measured through indicators such as output per unit time and equipment utilization. On-time delivery rate refers to the proportion of orders delivered on time. The genetic algorithm dynamically adjusts the weight coefficients λ1, λ2, and λ3 based on different production scenarios and business needs by simulating biological evolution. For example, when focusing on cost control, λ1 can be appropriately increased; when pursuing production efficiency improvements, the weight of λ2 can be increased.

[0124] More specifically, in terms of the model's network architecture, the convolutional neural network in the encoder uses convolution kernels of varying sizes to extract spatial features of data such as equipment, materials, and orders. For example, the impact of equipment layout and material storage relationships on scheduling is captured by the long-short-term memory network. For example, trends in equipment failure rates over time and dynamic patterns in material inventory are captured by the long-short-term memory network. The gated recurrent unit in the decoder incorporates an attention mechanism. When generating scheduling instruction sequences, the attention mechanism focuses on key information such as equipment status and urgent orders, ensuring that the generated instructions are more aligned with actual production needs.

[0125] Furthermore, when simulating the coordinated scheduling behavior between multiple production units within a workshop, the multi-agent reinforcement learning mechanism can treat each production unit as an agent, with agents interacting and collaborating to jointly optimize the global scheduling strategy. For example, different equipment agents can negotiate the allocation of processing tasks, and the material delivery agent can collaborate with the equipment agents to determine material delivery times to maximize overall production efficiency.

[0126] In some embodiments, a production scheduling plan is generated, specifically:

[0127] Obtain real-time data on the workshop's production equipment status, material inventory levels, work-in-progress information, and pending production order data, and input them into the trained workshop scheduling optimization model;

[0128] The model uses an internal deep reinforcement learning algorithm to intelligently schedule production equipment. Taking into account equipment load balancing, preventive maintenance, and process adaptability, it generates a Gantt chart for equipment scheduling, determines the production tasks, workpieces, and process parameter settings for each piece of equipment in different time periods, and uses a meta-heuristic algorithm to perform local search optimization on the equipment scheduling results.

[0129] Based on the inventory status of materials, the flow of work-in-process products, and the bill of materials of the production order, the model plans the material delivery route and delivery time window. It uses the ant colony optimization algorithm combined with the vehicle routing problem model to consider factors such as traffic congestion, loading and unloading time, and delivery costs during the material delivery process to generate the optimal material delivery route planning diagram;

[0130] Taking into account the urgency of production orders, the complexity of product processes, and the matching of personnel skills, the model prioritizes production tasks and assigns tasks to personnel. It uses the Hungarian algorithm to solve the optimal matching problem between personnel and tasks and formulates a detailed production task execution plan, including the start time, end time, assigned equipment, operators, and quality inspection nodes of the task.

[0131] It should be noted that real-time access to various types of workshop data is fundamental to generating effective scheduling plans. Production equipment status encompasses various states, including operation, shutdown, standby, and failure. This status information directly impacts the equipment's ability to accept new tasks and the order in which these tasks are scheduled. Material inventory levels, including the quantities of raw materials, work-in-progress, and finished products, determine the feasibility of material distribution and production tasks. Work-in-progress information, including the quantity, location, and processing progress of semi-finished products in each production process, is crucial for the continuity of the production process. Pending production order data, containing key information such as product requirements and delivery times, is an important basis for determining production priorities. Inputting this data into a trained workshop scheduling optimization model, the model can generate a reasonable scheduling plan by comprehensively considering various factors based on its internal algorithmic logic and learned experience.

[0132] Specifically, in intelligent equipment scheduling, load balancing involves rationally allocating tasks to various devices to ensure a relatively balanced workload, avoiding overworking some devices while leaving others idle. For example, this can be achieved by calculating the average load rate of each device and setting a reasonable load variance threshold. Tasks can then be adjusted when the variance exceeds the threshold. Preventive maintenance involves predicting the time and components of likely equipment failures based on equipment operating data and historical failure records, allowing for proactive maintenance scheduling. For example, predictive models can be developed using equipment operating hours and wear data on key components. Process adaptability requires that task allocations be consistent with the equipment's processing capabilities, such as accuracy and dimensional range. When generating a Gantt chart for equipment job scheduling, parameters such as setup time, processing time, and changeover time must be considered to accurately determine the production tasks, workpieces, and process parameter settings for each device within different time periods. For metaheuristic algorithms that perform local search optimization on equipment scheduling results, such as simulated annealing, parameters such as the initial temperature and cooling rate can be set. By randomly changing the equipment task allocation scheme within a certain range and evaluating the performance of each new scheme based on the objective function, a more optimal equipment scheduling result can be gradually found.

[0133] More specifically, in material distribution planning, the ant colony optimization algorithm simulates the pheromone transmission of ants during their foraging process to find the optimal path. In this application, the various material storage points and production equipment locations in a workshop can be considered nodes, and the material distribution paths can be considered edges. The concentration of pheromones left by ants along the paths guides the choices of subsequent ants, with paths with higher pheromone concentrations being more likely to be chosen. Vehicle routing problem models consider factors such as vehicle load capacity, driving speed, and delivery range. For example, the required vehicle type is determined based on the weight and volume of the materials, and delivery routes are planned in conjunction with the workshop layout and traffic regulations. Factors such as traffic congestion, loading and unloading times, and delivery costs must also be considered during the distribution process. When generating an optimal material distribution path plan, real-time traffic data can be used to determine traffic congestion conditions, historical loading and unloading time data can be used to estimate the current loading and unloading time, and delivery costs can be comprehensively calculated. The path plan can then be continuously adjusted to achieve efficient and cost-effective material distribution. In the area of personnel and task allocation, the Hungarian algorithm is a matrix-based combinatorial optimization algorithm used to solve the problem of optimally matching personnel with tasks. In actual applications, a matching assessment is conducted based on the skill requirements of the production task and the skill level of the personnel. For example, welding tasks are assigned to workers with welding skills and high proficiency. At the same time, production tasks are prioritized based on factors such as the urgency of the production order and the complexity of the product process. For example, urgent orders with complex processes are given priority to experienced and well-skilled workers. A detailed production task execution plan is then developed to clearly define the task's start and end time, assigned equipment, operators, and quality inspection nodes to ensure the smooth progress of the production process. In addition, an AI-based intelligent recommendation system can be used as an alternative. By analyzing large amounts of historical production data and personnel performance data, it can intelligently recommend suitable operators for each production task and dynamically adjust the recommendation results based on real-time production conditions.

[0134] In some embodiments, a production scheduling scheme is executed, specifically:

[0135] After receiving the generated scheduling plan, the execution monitoring module remotely controls and monitors the production equipment in the workshop in real time through the Industrial Internet of Things platform. Based on the equipment's operation scheduling Gantt chart, it accurately controls the equipment's start and stop, operating speed, processing parameters, etc. At the same time, it collects the equipment's real-time operating data and uses machine learning algorithms to conduct real-time diagnosis of the equipment's operating status and provide fault warnings.

[0136] Utilize automated warehousing and logistics systems to automatically dispatch forklifts, AGVs, or automated conveyor belts based on material distribution route planning to achieve automated material handling and precise distribution. At the same time, use radio frequency identification technology and visual recognition systems to track and monitor the material distribution process in real time.

[0137] With the help of the personnel positioning and task management system, task assignment information is pushed to operators in real time according to the production task execution plan. Through smart bracelets or mobile terminal devices, the location and work status of personnel are monitored and managed in real time. At the same time, the personnel's operation data and feedback information are collected and transmitted back to the data acquisition module.

[0138] It should be noted that the execution monitoring module plays a key control role in the entire production process. After receiving the generated scheduling plan, the Industrial Internet of Things platform becomes a bridge connecting the execution monitoring module and the production equipment. Through this platform, remote control and real-time monitoring of production equipment can be achieved. The operating parameters of production equipment, such as start and stop status, operating speed, and processing parameters (including processing accuracy, cutting depth, etc.) can all be precisely controlled and monitored. For example, for CNC machine tools, their spindle speed, feed speed and other processing parameters can be set remotely to ensure that the equipment is produced according to the scheduling plan. At the same time, the equipment's operating data is collected in real time, which will provide a basis for subsequent analysis and optimization.

[0139] Specifically, when utilizing automated warehousing and logistics systems, material distribution routing maps are a key guide for material handling. The dispatch of equipment such as forklifts, automated guided vehicles (AGVs), and automated conveyor belts is based on this map. Radio frequency identification (RFID) technology, by attaching electronic tags to materials, enables rapid identification and location during the material distribution process, ensuring that materials are accurately and securely delivered to designated locations. Visual recognition systems, using cameras and other equipment, identify and monitor the appearance, shape, and position of materials, further improving the accuracy and reliability of material distribution. For example, in a large warehouse, AGVs automatically travel to the material storage area according to a routing map, identify materials using RFID technology, and deliver them to designated locations on the production line. The visual recognition system monitors the correct placement of materials in real time throughout the process.

[0140] More specifically, with the help of personnel positioning and task management systems, task assignment information is pushed to operators based on the production task execution plan. Smart bracelets or mobile devices not only allow operators to keep up to date with task details but also monitor their location and work status in real time. For example, smart bracelets can determine an operator's location through built-in GPS modules or indoor positioning technology, and use sensors to monitor information such as their movements and heart rate to determine their work status. When collecting operator operation data and feedback, detailed records of the operator's steps, operation time, and encountered problems can be recorded. This data is then transmitted back to the data acquisition module via wireless transmission technology. Furthermore, more advanced wearable devices can integrate multiple sensors, such as pressure sensors and accelerometers, to more comprehensively monitor the operator's operating force, movement frequency, and other information. This provides more detailed data support for optimizing operations, thereby further improving production efficiency and quality.

[0141] The above-mentioned embodiments of the present invention have the following beneficial effects: The intelligent workshop scheduling optimization system of the present invention can realize the comprehensive collection and in-depth processing of workshop production data. The data acquisition module can collect multivariate data such as the operating parameters of production equipment, material flow information, and production order details in real time, providing a detailed data basis for subsequent scheduling decisions. The data fusion and preprocessing module can effectively improve data quality through data cleaning, feature engineering, and data labeling, mine high-dimensional features closely related to workshop scheduling optimization, and construct accurate training samples, thereby providing high-quality data support for the training of intelligent scheduling models and improving the training effect and generalization ability of the model.

[0142] The intelligent scheduling model training module, based on a deep reinforcement learning algorithm, integrates convolutional neural networks, recurrent neural networks, and an attention mechanism to construct a shop-floor scheduling optimization model with powerful spatiotemporal feature extraction capabilities. This model continuously adjusts parameters through interaction with the environment, learns the optimal scheduling strategy, and generates efficient scheduling solutions. The real-time scheduling decision module receives real-time shop-floor data streams and rapidly outputs scheduling instructions covering multiple dimensions, including equipment operation scheduling, material distribution planning, and personnel task allocation. It also utilizes risk assessment and adjustment optimization algorithms to address uncertainties in the production process, ensuring the robustness and adaptability of the scheduling solution. Based on scheduling instructions, the execution monitoring module uses the Industrial Internet of Things, automated logistics systems, and personnel management systems to precisely control and monitor shop-floor production in real time. It also collects execution feedback data to form a closed-loop control system, enabling dynamic optimization and continuous improvement of shop-floor production, driving the development of intelligent and efficient shop-floor production.

[0143] like Figure 2 As shown, in some embodiments, an artificial intelligence-based intelligent workshop scheduling optimization method 200 includes:

[0144] Step 201, the pre-preparation phase, collects multiple data such as the operating parameters of production equipment in the workshop, material flow information, and production order details in real time, and uses data processing technology to perform deep cleaning, feature extraction, and annotation on the data to construct a training data set;

[0145] Step 202, the model construction and training phase, uses a deep reinforcement learning algorithm that integrates convolutional neural networks, recurrent neural networks, and attention mechanisms based on the labeled feature vector data to build a shop floor scheduling optimization model. Through interactive simulation with the virtual production environment, the model parameters are continuously adjusted to learn the optimal scheduling strategy.

[0146] Step 203, the real-time scheduling decision-making phase, inputs the real-time collected shop floor data into the trained model. The model then outputs real-time scheduling instructions covering multiple dimensions, including equipment operation scheduling, material distribution planning, and personnel task allocation. Risk assessment and adjustment optimization algorithms are also used to address uncertainties in the production process.

[0147] Step 204, the execution monitoring phase, according to the scheduling instructions, through the industrial Internet of Things, automated logistics system and personnel management system, the workshop production is precisely controlled and monitored in real time, and at the same time, execution feedback data is collected to form a closed-loop control to achieve dynamic optimization and continuous improvement of workshop production.

[0148] It is understandable that the steps described in the artificial intelligence-based intelligent workshop scheduling optimization method 200 are similar to those in the reference Figure 1 The modules, features, and beneficial effects described above for the intelligent workshop scheduling optimization system based on artificial intelligence are also applicable to the intelligent workshop scheduling optimization method 200 based on artificial intelligence and the operations contained therein, and are not further described here.

[0149] Reference below Figure 3 , which shows a schematic structural diagram of an electronic device 300 suitable for implementing some embodiments of the present invention. The electronic devices in some embodiments of the present invention may include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 3 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0150] like Figure 3As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. Various programs and data required for the operation of the electronic device 300 are also stored in the RAM 303. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0151] Typically, the following devices may be connected to the I / O interface 305: an input device 306 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 308 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 309. The communication device 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as needed.

[0152] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0153] The above descriptions are merely some preferred embodiments of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also encompass other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned inventive concept. For example, a technical solution formed by mutually replacing the above-mentioned features with (but not limited to) technical features having similar functions disclosed in the embodiments of the present invention.

Claims

1. An intelligent workshop scheduling optimization system based on artificial intelligence, characterized in that: include: Data collection module: This module collects in real time the operating parameters of production equipment in the workshop, material flow information, and production order details. The operating parameters of production equipment include equipment load rate, failure rate, and energy consumption indicators; material flow information includes material batch, specification, inventory level, and work-in-progress status; and production order details include order number, product model, delivery date, and quality requirements. The collected data is then transmitted to the data fusion and preprocessing module. Data fusion and preprocessing module: including data cleaning submodule, feature engineering submodule and data labeling submodule; The data cleaning submodule is used to perform in-depth cleaning of the collected workshop data, using data mining technology to identify and eliminate outliers, duplicate records, and missing values in the data, while also unifying and standardizing the data format; The feature engineering submodule is used to mine high-dimensional features related to shop floor scheduling optimization from the cleaned data and extract key features using dimensionality reduction techniques; The data annotation submodule is used to accurately annotate feature vectors based on historical production data and expert experience, classifying scheduling results into excellent, good, fair, and poor grades, providing training samples for the supervised learning model. The annotation process uses a semi-automated annotation tool combined with manual review. Intelligent Scheduling Model Training Module: Based on labeled feature vector data, a deep reinforcement learning algorithm is used to build a shop scheduling optimization model. The model architecture integrates a convolutional neural network to extract spatial features, a recurrent neural network to capture time series features, and an attention mechanism to highlight key information. Through interaction with the environment, the model parameters are continuously adjusted to learn the optimal scheduling strategy. This enables the model to make autonomous decisions in complex and changing shop floor environments, generate efficient scheduling solutions, and deploy the trained model to the real-time scheduling decision module. Real-time scheduling decision module: Receives real-time data streams from the workshop and inputs them into a trained intelligent scheduling model. The model then outputs scheduling instructions in real time, including the equipment's job scheduling sequence, material delivery routes and time nodes, and personnel task allocation and collaboration arrangements. It also considers uncertainties in the production process and uses methods such as Monte Carlo simulation to conduct risk assessment and optimize adjustments to the scheduling plan, ensuring its robustness and adaptability. The finalized scheduling instructions are then sent to the execution monitoring module. Execution monitoring module: Based on the received scheduling instructions, the production equipment, material transportation system and operators in the workshop are precisely controlled and monitored in real time. The Internet of Things technology is used to realize remote control and status monitoring of equipment. The automated warehousing and logistics system is used to ensure the accurate distribution of materials. The personnel positioning and task management system is used to ensure that personnel complete tasks on time. At the same time, feedback data during the execution process is collected and transmitted back to the data acquisition module to form a closed-loop control, realizing dynamic optimization and continuous improvement of workshop production.

2. The intelligent workshop scheduling optimization system based on artificial intelligence according to claim 1 is characterized in that: Deeply clean the operating parameters of production equipment, specifically: Collect operating parameter data of all production equipment in the workshop during the continuous production cycle. The data dimensions include equipment start and stop time, operating speed, processing accuracy, and energy consumption; Use time series analysis to detect anomalies in the equipment's start and stop time data, identify abnormal start and stop records that do not conform to production rules, and use an outlier repair algorithm based on a Gaussian mixture model, combined with the equipment's historical normal start and stop patterns, to reasonably correct abnormal start and stop times. The running speed data of the equipment is subjected to noise filtering. Wavelet transform combined with adaptive threshold algorithm is used to decompose the running speed data into sub-band signals of different frequencies, retaining the low-frequency useful signals and removing the high-frequency noise interference. At the same time, according to the processing requirements of the equipment, the filtered running speed data is normalized by interval processing. The formula is: Among them, v is the original running speed data of the device, v min is the minimum value of the device operating speed data, v max is the maximum value of the device operating speed data, v mean is the mean of the equipment running speed data, σ is the standard deviation of the equipment running speed data, v norm is the normalized running speed data; For the equipment's processing accuracy data, a missing value filling method based on Bayesian network is adopted to construct a Bayesian network model of the equipment's processing accuracy. The type of equipment, the material of the workpiece, and the processing parameters are used as network nodes. The conditional probability distribution of the network is learned according to the known processing accuracy data. For the missing processing accuracy data, the Bayesian inference algorithm is used, combined with the network's prior knowledge and observation data, to calculate the posterior probability distribution of the missing data, thereby obtaining the filling value of the missing data.

3. The intelligent workshop scheduling optimization system based on artificial intelligence according to claim 2 is characterized in that: Feature extraction of material flow information, specifically: Integrate the batch information, specification parameters, inventory dynamics and work-in-progress flow status of each material in the workshop to build a multi-dimensional data matrix of material flow information; To calculate the dynamic inventory turnover rate of materials, the weighted moving average method is combined with the exponential smoothing technology. Different weight coefficients are assigned according to the importance and demand volatility of different materials. The formula is: Among them, T turn is the dynamic inventory turnover rate of the material, w i is the weight coefficient of the i-th material, D i is the outbound quantity of the i-th material in the statistical period, I i is the average inventory of the i-th material in the statistical period, α and β are exponential smoothing parameters, and t is the time variable; Analyze the flow path and processing cycle of materials in the process, use the shortest path algorithm and critical path method in graph theory to construct a network diagram of the flow of materials in the process, calculate the average processing cycle, longest processing path and bottleneck process of each material in the process from raw material input to finished product output, and thus obtain the key characteristic parameters of the flow of materials in the process, including the average residence time of materials in the process, the waiting time ratio between processes, and the capacity utilization rate of key processes.

4. The intelligent workshop scheduling optimization system based on artificial intelligence according to claim 3 is characterized in that: Perform feature extraction on production order details, specifically: Sorting out the full lifecycle information of production orders, including the order source, signing date, delivery requirements, product technical parameters, and customer feedback, to construct a feature vector for the production order; To quantify the urgency of production orders, we use the fuzzy comprehensive evaluation method combined with the analytic hierarchy process to establish an evaluation index system for order urgency. The indicators include order delivery lead time, customer priority, and product importance. Through expert scoring and weight calculation, we obtain the fuzzy comprehensive evaluation matrix of order urgency. The formula is: Among them, E is the fuzzy comprehensive evaluation matrix of order urgency, e ij It represents the fuzzy membership of the i-th order on the j-th evaluation index, n is the number of orders, m is the number of evaluation indicators, and the comprehensive score of the urgency of each order is calculated according to the fuzzy comprehensive evaluation matrix; Analyze the product technical parameters of the production order, use semantic analysis technology and data mining algorithms to extract the association rules between the product's key performance indicators and production process requirements, build a feature mapping table of the product technical parameters, and classify and mark the orders according to the complexity of the product technical parameters and special requirements.

5. The intelligent workshop scheduling optimization system based on artificial intelligence according to claim 4 is characterized in that: Construct a shop floor scheduling optimization model, specifically: The extracted high-dimensional feature data such as equipment comprehensive efficiency, material supply stability, and order urgency coefficient are used as input variables of the model. Minimizing production costs, maximizing production efficiency, and ensuring on-time delivery are used as multi-objective optimization functions to construct a workshop scheduling optimization model based on deep reinforcement learning. The loss function of the model adopts the weighted sum form of multi-objective optimization, and the formula is: L=λ1L cost +λ2L efficiency +λ3L delivery Among them, L is the total loss function of the model, L cost is the production cost loss, L efficiency is the production efficiency loss, L delivery is the on-time delivery rate loss, λ1, λ2, and λ3 are the corresponding weight coefficients, and the weight coefficients are dynamically adjusted through the genetic algorithm; The model's network architecture adopts an encoder-decoder structure. The encoder part consists of a multi-layer convolutional neural network and a long short-term memory network, which is used to encode the spatiotemporal features of the input high-dimensional feature data; the decoder part uses a gated recurrent unit combined with an attention mechanism to generate the optimal scheduling instruction sequence based on the feature vector output by the encoder. At the same time, a multi-agent reinforcement learning mechanism is introduced to simulate the collaborative scheduling behavior between multiple production units in the workshop and optimize the global scheduling strategy.

6. The intelligent workshop scheduling optimization system based on artificial intelligence according to claim 5 is characterized in that: Generate a production scheduling plan, specifically: Obtain real-time data on the workshop's production equipment status, material inventory levels, work-in-progress information, and pending production order data, and input them into the trained workshop scheduling optimization model; The model uses an internal deep reinforcement learning algorithm to intelligently schedule production equipment. Taking into account equipment load balancing, preventive maintenance, and process adaptability, it generates a Gantt chart for equipment scheduling, determines the production tasks, workpieces, and process parameter settings for each piece of equipment in different time periods, and uses a meta-heuristic algorithm to perform local search optimization on the equipment scheduling results. Based on the inventory status of materials, the flow of work-in-process products, and the bill of materials of the production order, the model plans the material delivery route and delivery time window. It uses the ant colony optimization algorithm combined with the vehicle routing problem model to consider factors such as traffic congestion, loading and unloading time, and delivery costs during the material delivery process to generate the optimal material delivery route planning diagram; Taking into account the urgency of production orders, the complexity of product processes, and the matching of personnel skills, the model prioritizes production tasks and assigns tasks to personnel. It uses the Hungarian algorithm to solve the optimal matching problem between personnel and tasks and formulates a detailed production task execution plan, including the start time, end time, assigned equipment, operators, and quality inspection nodes of the task.

7. The intelligent workshop scheduling optimization system based on artificial intelligence according to claim 6 is characterized in that: Execute the production scheduling plan, specifically: After receiving the generated scheduling plan, the execution monitoring module remotely controls and monitors the production equipment in the workshop in real time through the Industrial Internet of Things platform. Based on the equipment's operation scheduling Gantt chart, it accurately controls the equipment's start and stop, operating speed, processing parameters, etc. At the same time, it collects the equipment's real-time operating data and uses machine learning algorithms to conduct real-time diagnosis of the equipment's operating status and provide fault warnings. Utilize automated warehousing and logistics systems to automatically dispatch forklifts, AGVs, or automated conveyor belts based on material distribution route planning to achieve automated material handling and precise distribution. At the same time, use radio frequency identification technology and visual recognition systems to track and monitor the material distribution process in real time. With the help of the personnel positioning and task management system, task assignment information is pushed to operators in real time according to the production task execution plan. Through smart bracelets or mobile terminal devices, the location and work status of personnel are monitored and managed in real time. At the same time, the personnel's operation data and feedback information are collected and transmitted back to the data acquisition module.

8. An artificial intelligence-based intelligent workshop scheduling optimization method, applied to an artificial intelligence-based intelligent workshop scheduling optimization system according to any one of claims 1 to 7, characterized in that: include: During the pre-preparation phase, we collect real-time data on various aspects of production equipment in the workshop, including operating parameters, material flow information, and production order details. We then use data processing technology to perform deep cleaning, feature extraction, and annotation on the data to construct a training dataset. During the model building and training phase, based on the labeled feature vector data, a deep reinforcement learning algorithm is used to integrate convolutional neural networks, recurrent neural networks, and attention mechanisms to build a shop floor scheduling optimization model. Through interactive simulation with the virtual production environment, the model parameters are continuously adjusted to learn the optimal scheduling strategy. During the real-time scheduling decision-making phase, real-time collected shop floor data is fed into a trained model. The model then outputs real-time scheduling instructions covering multiple dimensions, including equipment operation scheduling, material distribution planning, and personnel task allocation. Risk assessment and adjustment optimization algorithms are also used to address uncertainties in the production process. During the execution monitoring stage, according to scheduling instructions, the industrial Internet of Things, automated logistics systems, and personnel management systems are used to accurately control and monitor workshop production in real time. At the same time, execution feedback data is collected to form a closed-loop control, thereby achieving dynamic optimization and continuous improvement of workshop production.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the intelligent workshop scheduling optimization method based on artificial intelligence as described in claim 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the artificial intelligence-based intelligent workshop scheduling optimization method as described in claim 8.

Images