An Internet intelligent manufacturing service platform
By applying dynamic ecological optimization algorithms, machine learning and deep learning technologies in the Internet intelligent manufacturing service platform, the shortcomings of traditional platforms in data processing, fault prediction, quality control and demand prediction are solved, and more efficient and flexible production management is achieved. , improve equipment health management and product quality control, and reduce maintenance costs and production interruption risks.
Patent Information
- Application Number
- CN202410356260.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-27
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-03-27
AI Technical Summary
Traditional Internet intelligent manufacturing service platforms have shortcomings in real-time data processing, equipment failure prediction, product quality control, demand and supply prediction, etc., lack flexibility and adaptability, and are difficult to cope with the rapid changes in the production environment and market demand, resulting in equipment maintenance costs. The high risk of production interruption increases, product quality control is not high, and demand forecasting is low, which affects the formulation and implementation of production plans.
Using dynamic ecological optimization algorithms, isolated forest algorithms, self-coded neural networks, convolutional neural networks, long and short-term memory networks, genetic algorithms and deep learning models, etc., to build production optimization scheduling, equipment health management, quality control, safety monitoring, and requirements With supply forecasting and fault diagnosis modules, real-time data analysis and dynamic adjustments are implemented to improve production efficiency and resource utilization.
By dynamically adjusting production processes and resource configuration, improve the accuracy of equipment failure prediction, automatically identify product defects, improve product quality, improve the accuracy and efficiency of demand and supply prediction, optimize production scheduling, reduce maintenance costs, and enhance the intelligence of production systems level, ensuring the efficiency and reliability of the production process.
Smart Images

Figure CN118261439B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent manufacturing technology, and in particular to an Internet intelligent manufacturing service platform. Background Art
[0002] The field of intelligent manufacturing technology is a manufacturing model that integrates information technology, artificial intelligence, the Internet of Things and other advanced technologies, aiming to achieve automation, intelligence and networking in the production process. The field of intelligent manufacturing technology not only focuses on the intelligent production of products, but also covers data analysis, decision support, resource optimization and allocation in the production process, and is committed to improving manufacturing efficiency, reducing costs, enhancing product quality and production flexibility, and ultimately promoting the manufacturing industry to a higher level of intelligent development.
[0003] Among them, the Internet intelligent manufacturing service platform is a service system that uses Internet technology to achieve efficient configuration and intelligent management of manufacturing resources. The core purpose of the platform is to integrate various resources of the manufacturing industry, including manpower, machines, materials and information, through technical means to optimize the production process, improve product quality and effectively reduce costs, enhance the innovation ability and market response speed of the manufacturing industry, and enhance the competitiveness and sustainable development capabilities of enterprises.
[0004] Traditional Internet intelligent manufacturing service platforms have many deficiencies in real-time data processing, dynamic adjustment of production processes, equipment failure prediction, product quality control, production demand and supply prediction, production scheduling optimization, fault diagnosis and maintenance, etc. Traditional methods rely on static planning and manual experience, lack flexibility and adaptability, and are difficult to cope with the rapid changes in production environment and market demand. The prediction and diagnosis of equipment failures mostly rely on manual inspection, which is inefficient and inaccurate, resulting in high equipment maintenance costs and increased risks of production interruptions. In terms of product quality control, traditional methods lack efficient automatic detection mechanisms, which can easily cause the outflow of defective products and affect brand reputation. In terms of demand and supply forecasting, traditional methods have low prediction accuracy and are difficult to effectively deal with uncertainties in the supply chain, affecting the formulation and execution of production plans. Summary of the invention
[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an Internet intelligent manufacturing service platform.
[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions: an Internet intelligent manufacturing service platform includes a production optimization scheduling module, an equipment health management module, a quality control management module, a safety monitoring management module, a demand and supply forecasting module, an intelligent scheduling optimization module, a fault diagnosis and maintenance module, and a decision support module;
[0007] The production optimization scheduling module uses a dynamic ecological optimization algorithm based on real-time production line data to simulate the competition and symbiotic relationship of species in the natural ecology. Through the selection, crossover and mutation operations within the algorithm, the production process and configuration are dynamically adjusted to match environmental changes and generate an optimized production configuration plan;
[0008] The equipment health management module uses the isolation forest algorithm to analyze abnormal points in equipment operation data based on the optimized production configuration plan, identify potential equipment failures, and encode and decode equipment status features in combination with the autoencoder neural network to predict equipment failures and generate equipment health assessment results;
[0009] The quality control management module uses a convolutional neural network based on the equipment health assessment results to automatically extract features in the image through stacked filters and classify them, identify and mark product defects on the production line, monitor product quality in real time, and generate quality control results;
[0010] The safety monitoring management module uses the isolation forest algorithm to analyze production data based on the quality control results, identifies abnormal patterns in the data by building an isolation tree, identifies safety risks in the production process, and generates safety monitoring results;
[0011] The demand and supply forecasting module uses a long short-term memory network to process time series data based on safety monitoring results, learns historical production demand and raw material supply patterns, and uses a time convolutional network to optimize the efficiency and accuracy of sequence data processing, provides forecasts for production demand and raw material supply in future time periods, and generates comprehensive forecast analysis results;
[0012] The intelligent scheduling optimization module uses genetic algorithms to simulate natural selection and genetic mechanisms to optimize the solution of production scheduling problems based on comprehensive prediction and analysis results, and uses simulated annealing algorithms to perform random analysis to overcome local optimal problems, dynamically adjust production plans and resource allocation, and generate scheduling optimization solutions;
[0013] The fault diagnosis and maintenance module uses a deep learning model based on the scheduling optimization scheme to learn the equipment fault data pattern through a deep neural network, diagnose and predict the fault type, and generate preventive maintenance measures;
[0014] The decision support module is based on preventive maintenance measures, uses a rule engine to process information based on predefined decision logic, combines a decision tree to analyze decision results through a tree-structured decision path, and generates execution effect feedback results.
[0015] As a further scheme of the present invention, the optimized production configuration scheme includes production rhythm adjustment parameters, robot arm scheduling priorities, and raw material delivery speed settings; the equipment health assessment results include equipment maintenance cycle recommendations, key component wear assessments, and fault warning levels; the quality control results include defect size ranges, defect location information, and product qualification rate statistics; the safety monitoring results include safety risk levels, hazardous area identification, and emergency shutdown trigger points; the comprehensive forecasting and analysis results include production demand increase and decrease trends, raw material gap forecasts, and supply chain stability scores; the scheduling optimization scheme includes optimal production line configuration, key production node priorities, and spare resource allocation strategies; the preventive maintenance measures include maintenance time window arrangements, key component inspection lists, and spare parts inventory plans; the execution effect feedback results include production adjustment decision paths, risk mitigation measures selection, and benefit-maximizing resource allocation.
[0016] As a further solution of the present invention, the production optimization scheduling module includes an ecological model submodule, an optimization execution submodule, and a feedback adjustment submodule;
[0017] The ecological model submodule uses a dynamic ecological optimization algorithm to simulate the competition and symbiosis mechanism of species in the natural ecology based on real-time production line data, and dynamically adjusts the production process and configuration by simulating the selection, crossover and mutation operations in the biological evolution process to generate an ecological optimization configuration plan;
[0018] The optimization execution submodule uses a linear programming method to optimize the production process and equipment configuration based on the ecological optimization configuration scheme, and uses the simplex method and interior point method solvers to capture the optimal solution of resource configuration and generate an optimization execution configuration scheme by establishing an objective function and constraints;
[0019] The feedback adjustment submodule is based on the optimized execution configuration plan and adopts a feedback control strategy to perform real-time monitoring of production execution and collect feedback information. Through continuous analysis and comparison of production data, it adjusts production parameters, matches real-time production needs and conditions, and generates an optimized production configuration plan.
[0020] As a further solution of the present invention, the equipment health management module includes a state monitoring submodule, a fault prediction submodule, and a health assessment submodule;
[0021] The state monitoring submodule uses the isolation forest algorithm to analyze abnormal points in the equipment operation data based on the optimized production configuration scheme, isolates the abnormal points on a few branches by constructing multiple isolated trees, identifies the abnormal points in the data, and generates equipment abnormality monitoring results;
[0022] The fault prediction submodule uses an autoencoder neural network to encode and decode the equipment status features based on the equipment abnormality monitoring results, learns the representation of the data by compressing the input data, reconstructs the data, and generates fault prediction results by capturing the key features in the data;
[0023] The health assessment submodule uses a support vector machine to perform a comprehensive assessment of the health status of the equipment based on the fault prediction results, and generates equipment health assessment results by constructing an optimal decision boundary in a multidimensional space to separate differentiated health status categories.
[0024] As a further solution of the present invention, the quality control management module includes an image acquisition submodule, a defect recognition submodule, and a quality feedback submodule;
[0025] The image acquisition submodule acquires product images on the production line based on the equipment health assessment results, uses a camera to capture product surface details, provides raw image data for defect identification, and generates a product image dataset;
[0026] The defect recognition submodule uses a convolutional neural network to automatically extract image features and classify them based on the product image dataset. Through the convolutional layer, pooling layer and fully connected layer in the deep network structure, it captures image features and identifies defects of different types to generate defect recognition results.
[0027] The quality feedback submodule uses a data-driven decision model to monitor and feedback product quality in real time based on defect identification results. It identifies quality trends and anomalies by analyzing historical quality data and real-time feedback, guides production adjustments to optimize product quality, and generates quality control results.
[0028] As a further solution of the present invention, the security monitoring management module includes a feature extraction submodule, an anomaly detection submodule, and an alarm generation submodule;
[0029] The feature extraction submodule extracts security features from production data using an isolation forest algorithm based on quality control results, constructs an isolation tree by randomly selecting features and split values, identifies abnormal patterns in the data, and generates security feature analysis results;
[0030] The anomaly detection submodule uses anomaly point detection technology to analyze the security feature analysis results, calculates the degree of outliers of data points, identifies data anomaly patterns, provides a basis for security risk management, and generates anomaly detection results;
[0031] The alarm generation submodule uses rule-based decision-making to evaluate security warnings based on the abnormal detection results. Through preset security rules and logical judgments, it analyzes the severity and potential impact of abnormal patterns, forms targeted security measures and warning information, and generates security monitoring results.
[0032] As a further solution of the present invention, the demand and supply forecasting module includes a data sorting submodule, a forecasting model building submodule, and a forecasting result analysis submodule;
[0033] The data sorting submodule uses data processing techniques based on the safety monitoring results, including using the Z-score method to standardize the data, identifying and eliminating outliers through the IQR method, and filling missing values through linear interpolation to generate a time series data set;
[0034] The forecast model building submodule is based on a time series data set and uses a long short-term memory network combined with a temporal convolutional network. The long short-term memory network captures the long-term dependencies in the time series through its gating mechanism, and the temporal convolutional network optimizes the efficiency and accuracy of processing sequence data through a causal convolutional layer to generate demand and supply forecast results.
[0035] The forecast result analysis submodule uses trend analysis methods based on demand and supply forecast results to identify the randomness of demand peaks and supply shortages by calculating the difference between the forecast value and the real-time value, providing data support for production decisions in future time periods and generating comprehensive forecast analysis results.
[0036] As a further solution of the present invention, the intelligent scheduling optimization module includes a scheduling strategy submodule, a resource allocation submodule, and an optimization feedback submodule;
[0037] The scheduling strategy submodule uses a genetic algorithm to design an initial production scheduling plan based on the comprehensive prediction and analysis results. After initializing the population, the roulette wheel selection method is used to select excellent individuals, the single point crossover method is used to exchange individual information, and the uniform mutation method is used to introduce new genetic information. The optimal production scheduling strategy is captured through continuous iterative optimization and generates a preliminary scheduling optimization strategy.
[0038] The resource allocation submodule optimizes the resource allocation problem based on the preliminary scheduling optimization strategy using a simulated annealing algorithm. By setting the initial temperature and cooling plan, the probability of accepting a poor solution is gradually reduced as the temperature decreases, thereby capturing the global optimal resource allocation strategy and generating a resource allocation optimization solution.
[0039] The optimization feedback submodule implements a dynamic adjustment and feedback mechanism based on the resource allocation optimization plan, adjusts the production plan and resource allocation according to the feedback information in the real-time production process, uses real-time data monitoring and adjustment strategies, and generates a scheduling optimization plan.
[0040] As a further solution of the present invention, the fault diagnosis and maintenance module includes a fault data analysis submodule, a fault diagnosis submodule, and a maintenance solution submodule;
[0041] The fault data analysis submodule optimizes the fault data based on the scheduling optimization scheme by using data preprocessing technology, removes abnormal data points and missing values through outlier processing, and fills missing data with linear interpolation to generate a processed fault data set;
[0042] The fault diagnosis submodule uses a deep neural network to perform fault diagnosis based on the processed fault data set, captures fault features through a convolutional layer, merges features through a fully connected layer, and uses a deep network structure to learn and identify fault modes to generate fault diagnosis and prediction results;
[0043] The maintenance plan submodule uses decision support technology to analyze the fault type and frequency based on the fault diagnosis and prediction results, and uses fault tree analysis and risk assessment to determine the priority and strategy of maintenance measures to generate preventive maintenance measures.
[0044] As a further solution of the present invention, the decision support module includes a data summary and analysis submodule, a decision generation submodule, and an execution feedback submodule;
[0045] The data summary and analysis submodule summarizes and analyzes production, maintenance and fault data based on preventive maintenance measures, using data warehouse and big data analysis technology, provides data views required for decision-making, and generates a comprehensive decision-making data set;
[0046] The decision generation submodule performs decision analysis based on the comprehensive decision data set in combination with the rule engine and the decision tree. The rule engine processes information based on logical rules, and the decision tree evaluates the randomness of the decision results through path analysis to generate a decision support plan.
[0047] The execution feedback submodule implements decision execution monitoring and feedback based on the decision support solution, evaluates the execution effect of the decision through performance indicator monitoring and deviation analysis, provides data support for decision adjustments in future time periods, and generates execution effect feedback results.
[0048] Compared with the prior art, the advantages and positive effects of the present invention are:
[0049] In the present invention, the production configuration is optimized by a dynamic ecological optimization algorithm, which effectively improves the production efficiency and the ability to match environmental changes. The isolation forest algorithm and the autoencoder neural network are used together to improve the prediction accuracy of equipment failures and reduce unexpected downtime. The application of convolutional neural networks in quality control automatically identifies product defects and improves product quality. The application of long short-term memory networks and time convolutional networks improves the accuracy and efficiency of prediction in demand and supply forecasting. The combination of genetic algorithms and simulated annealing algorithms optimizes production scheduling and improves resource utilization. The application of deep learning models in fault diagnosis and maintenance strengthens early diagnosis and prevention of faults and reduces maintenance costs. The comprehensive application of technologies improves the intelligence level of production systems, optimizes resource allocation, reduces production costs, improves production safety and product quality, and ensures the efficiency and reliability of production processes. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Figure 1 is a flow chart of the platform of the present invention;
[0051] Figure 2 It is a schematic diagram of the platform framework of the present invention;
[0052] Figure 3 It is a flow chart of the production optimization scheduling module of the present invention;
[0053] Figure 4 This is a flow chart of the equipment health management module of the present invention;
[0054] Figure 5 It is a flow chart of the quality control management module of the present invention;
[0055] Figure 6 It is a flow chart of the security monitoring management module of the present invention;
[0056] Figure 7 It is a flow chart of the demand and supply forecasting module of the present invention;
[0057] Figure 8 This is a flow chart of the intelligent scheduling optimization module of the present invention;
[0058] Fig. 9 It is a flow chart of the fault diagnosis and maintenance module of the present invention;
[0059] Fig.10 It is a flow chart of the decision support module of the present invention. DETAILED DESCRIPTION
[0060] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0061] In the description of the present invention, it is to be understood that the terms "length", "width", "upper", "lower", "front",
[0062] The orientation or position relationship is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.
[0063] Embodiment 1
[0064] See also Figure 1 , 2 ,An Internet intelligent manufacturing service platform includes a production optimization scheduling module, an equipment health management module, a quality control management module, a safety monitoring management module, a demand and supply prediction module, an intelligent scheduling optimization module, a fault diagnosis and maintenance module, and a decision support module;
[0065] The production optimization scheduling module is based on real-time production line data and adopts a dynamic ecological optimization algorithm to simulate the competition and symbiotic relationship of species in the natural ecology. Through the selection, crossover and mutation operations within the algorithm, the production process and configuration are dynamically adjusted to match environmental changes and generate an optimized production configuration plan;
[0066] The equipment health management module is based on the optimized production configuration plan. It uses the isolation forest algorithm to analyze abnormal points in the equipment operation data, identify potential equipment failures, and combines the autoencoder neural network to encode and decode equipment status features, predict equipment failures, and generate equipment health assessment results.
[0067] The quality control management module uses convolutional neural networks based on equipment health assessment results to automatically extract features from images and classify them through stacked filters, identify and mark product defects on the production line, monitor product quality in real time, and generate quality control results;
[0068] The safety monitoring management module uses the isolation forest algorithm to analyze production data based on quality control results. It identifies abnormal patterns in the data by building an isolation tree, identifies safety risks in the production process, and generates safety monitoring results.
[0069] The demand and supply forecasting module uses long short-term memory networks to process time series data based on safety monitoring results, learns historical production demand and raw material supply patterns, and uses time convolutional networks to optimize the efficiency and accuracy of sequence data processing, providing forecasts for production demand and raw material supply in future time periods, and generating comprehensive forecast analysis results;
[0070] The intelligent scheduling optimization module uses genetic algorithms to simulate natural selection and genetic mechanisms to optimize the solution of production scheduling problems based on comprehensive prediction and analysis results. It uses simulated annealing algorithms to perform random analysis to overcome local optimal problems, dynamically adjust production plans and resource allocation, and generate scheduling optimization solutions.
[0071] The fault diagnosis and maintenance module is based on the scheduling optimization scheme and adopts a deep learning model to learn the equipment fault data pattern through a deep neural network, diagnose and predict the fault type, and generate preventive maintenance measures;
[0072] The decision support module is based on preventive maintenance measures. It uses a rule engine to process information based on predefined decision logic, combines a decision tree to analyze decision results through a tree-structured decision path, and generates feedback results on the execution effect.
[0073] The optimized production configuration plan includes production rhythm adjustment parameters, robot arm scheduling priority, and raw material delivery speed setting. The equipment health assessment results include equipment maintenance cycle recommendations, key component wear assessment and fault warning level. The quality control results include defect size range, defect location information, and product qualification rate statistics. The safety monitoring results include safety risk level, dangerous area identification, and emergency shutdown trigger point. The comprehensive forecasting and analysis results include production demand increase and decrease trends, raw material gap prediction, and supply chain stability score. The scheduling optimization plan includes optimal production line configuration, key production node priority, and backup resource allocation strategy. Preventive maintenance measures include maintenance time window arrangement, key component inspection list, and spare parts inventory plan. The execution effect feedback results include production adjustment decision path, risk mitigation measure selection, and benefit maximization resource allocation.
[0074] In the production optimization scheduling module, the data is processed by the dynamic ecological optimization algorithm through real-time production line data, the data format includes production volume, timestamp, equipment status and raw material consumption. The dynamic ecological optimization algorithm simulates the competition and symbiotic relationship of species in the natural ecosystem, and optimizes the production process through three key operations: selection, crossover and mutation. The selection operation determines the survival probability of individuals based on the fitness function to ensure that the configuration scheme with high fitness has a greater chance of being retained. The crossover operation imitates the chromosome exchange process in biological genetics, and randomly exchanges some parameters in the two production configuration schemes to generate a new production configuration. The mutation operation randomly modifies a part of the parameters in the current configuration scheme, introduces new mutations, and explores a better production configuration space. Through these three operations, the algorithm dynamically adjusts the production process and configuration to match the environmental changes, thereby generating an optimized production configuration scheme, and iterating continuously until the optimal solution is found or the predetermined number of iterations is reached. The optimized production configuration scheme can effectively improve production efficiency, reduce resource consumption, and realize dynamic optimization of the production process.
[0075] In the equipment health management module, the equipment operation data is analyzed using the isolation forest algorithm and autoencoder neural network through the optimized production configuration plan. The isolation forest algorithm specifically detects outliers in the data set. By constructing multiple isolated trees, each data point is isolated. Outliers are easier to isolate due to different attributes, so they have a shorter path length and can effectively identify potential fault points in equipment operation data. The autoencoder neural network captures the key features of the equipment status through the encoding and decoding process of the equipment status features to predict equipment failures. During the training process of the autoencoder, the compressed representation of the data is learned by reducing the reconstruction error between the input and output. The equipment health assessment results finally generated by this module provide equipment managers with a method to accurately predict equipment failures, aiming to reduce equipment downtime through early fault detection and preventive maintenance.
[0076] In the quality control management module, a convolutional neural network (CNN) is used to analyze product images on the production line based on the equipment health assessment results. CNN automatically extracts features from images through its multi-layer structure, and classifies, identifies, and marks product defects. The process starts with automatically extracting low-level features (such as edges and corners) using convolutional layers, then combining low-level features through deeper network structures to form high-level features (such as texture and shape), and finally classifying features through fully connected layers to identify product defects. By adjusting the depth of the network layer and the size of the convolution filter, CNN can adapt to different types of defect recognition tasks, which not only improves the level of automation of product quality control, but also greatly reduces the negative impact of product defects on production and customer satisfaction through real-time monitoring.
[0077] In the safety monitoring management module, the isolation forest algorithm is used to analyze production data based on quality control results to identify safety risks in the production process. The isolation forest algorithm identifies abnormal patterns in the data set by constructing isolated trees, and abnormal patterns indicate safety risks in the production process. Similar to the application in the equipment health management module, the anomaly analysis here focuses on identifying abnormal situations that lead to safety accidents from production data. By monitoring and analyzing abnormal patterns in production data, the module can generate safety monitoring results in real time and promptly warn of potential safety risks on the production line, which not only improves the safety monitoring capabilities of the production line, but also provides important support for preventing accidents and ensuring worker safety.
[0078] In the demand and supply forecasting module, time series data is processed by long short-term memory network (LSTM), and the data format used is mainly the time series data of historical production demand and raw material supply. LSTM effectively captures the long-term dependencies in the time series through its gating mechanism to learn the historical production demand and raw material supply patterns. The data is preprocessed to meet the model input requirements. Temporal convolutional network (TCN) is used to optimize the efficiency and accuracy of sequence data processing, including using convolutional layers to capture local features in the sequence and increasing the receptive field through dilated convolution to process long sequence data. The advantage of TCN is that it can process sequence data in parallel, thereby speeding up the learning process. Integrating the results of LSTM and TCN, the model predicts production demand and raw material supply in the future by analyzing the historical patterns of time series. The generated comprehensive forecast analysis results are saved in the form of files, which list the predicted demand and supply at each time point in the future in detail, providing decision support for downstream modules.
[0079] In the intelligent scheduling optimization module, the production scheduling problem is optimized by combining genetic algorithm and simulated annealing algorithm. The genetic algorithm simulates natural selection and genetic mechanism, searches for the optimal solution in the solution space of the production scheduling scheme through selection, crossover and mutation operations, initializes a set of solutions as a population, each solution represents a production scheduling scheme, and evaluates its quality by calculating the fitness of each scheme. The scheme with high fitness is selected for crossover and mutation operations to generate new schemes. The simulated annealing algorithm is used to overcome the local optimal problem by randomly selecting adjacent solutions and accepting poor solutions according to a certain probability to jump out of the local optimal. In the process of dynamically adjusting the production plan and resource allocation, the algorithm comprehensively considers factors such as production efficiency, cost and delivery time to generate an optimized scheduling scheme. The scheme details the start and end time of each task on the production line and the resource allocation, effectively improving production efficiency and resource utilization.
[0080] In the fault diagnosis and maintenance module, deep learning models, including deep neural networks (DNNs), are used to diagnose and predict equipment faults. The data format used is the equipment's operating data and historical fault records. The data is preprocessed, including denoising and feature engineering, to extract information useful for fault diagnosis. The DNN model learns complex patterns and relationships in the data through its multi-layer structure to achieve accurate diagnosis of fault types. During the training process, the model continuously optimizes weights through a back-propagation algorithm to reduce the difference between predicted and actual values. The generated preventive maintenance measures are provided in the form of a document, which lists in detail the recommended maintenance activities for the predicted fault type, including repair time, required materials, and expected costs, aiming to reduce equipment failure rates and maintenance costs.
[0081] In the decision support module, preventive maintenance measures are analyzed through the rule engine and decision tree to provide decision support for management. The rule engine processes information based on predefined decision logic, such as evaluating maintenance priorities based on equipment importance and failure impact. The decision tree analyzes different decision paths through a tree structure, providing a logically clear decision process. The system comprehensively considers maintenance costs, impact scope, and production losses to generate specific decision plans. The plan describes in detail the measures taken, expected results, and implementation schedule, enabling management to make more effective maintenance decisions based on comprehensive analysis results. The generated execution effect feedback results further verify the accuracy and effectiveness of the decision, providing a basis for continuous improvement of the decision-making process.
[0082] See also Figure 2 , 3 ,The production optimization scheduling module includes the ecological model submodule, the ,optimization execution submodule, and the feedback adjustment submodule;
[0083] The ecological model submodule uses a dynamic ecological optimization algorithm to simulate the competition and symbiosis mechanism of species in the natural ecology based on real-time production line data. It dynamically adjusts the production process and configuration by simulating the selection, crossover and mutation operations in the biological evolution process to generate an ecological optimization configuration plan.
[0084] The optimization execution submodule uses linear programming methods to optimize production processes and equipment configurations based on the ecological optimization configuration plan. By establishing objective functions and constraints, the simplex method and interior point method solvers are used to capture the optimal solution for resource configuration and generate an optimization execution configuration plan.
[0085] The feedback adjustment submodule is based on the optimized execution configuration plan and adopts feedback control strategy to conduct real-time monitoring of production execution and collect feedback information. Through continuous analysis and comparison of production data, it adjusts production parameters, matches real-time production needs and conditions, and generates an optimized production configuration plan.
[0086] In the ecological model submodule, the competition and symbiosis mechanism of species in the natural ecology is simulated by using the dynamic ecological optimization algorithm through real-time production line data. The data format is a structured data set, including the operating parameters, output rate, quality control indicators and energy consumption information of the production line. The dynamic ecological optimization algorithm is based on data, performs selection, crossover and mutation operations in the simulated biological evolution process, and dynamically adjusts the production process and configuration. The selection operation evaluates the adaptability of the production configuration scheme based on output efficiency and quality indicators, and the preferred scheme enters the crossover and mutation stage. The crossover operation simulates the recombination of biological genetic information and promotes the complementary advantages of different production configuration schemes; the mutation operation introduces randomness and explores new production configuration possibilities. The operation is based on the principles of ecological competition and symbiosis, optimizes the production configuration, and generates an ecological optimization configuration scheme, which directly affects the production efficiency and resource utilization, and realizes the optimization of the production process.
[0087] In the optimization execution submodule, based on the ecological optimization configuration scheme, the linear programming method is used to finely optimize the production process and equipment configuration. The input data format is the optimized production configuration scheme, including production tasks, resource allocation and time schedule. By constructing a mathematical model containing objective functions and constraints, the simplex method and interior point method solvers are used to accurately capture the optimal solution of resource configuration. The objective function aims to maximize output efficiency or minimize costs, and the constraints reflect production capacity, resource limitations and time windows. The simplex method gradually approaches the optimal solution through linear boundaries, and the interior point rule finds the optimal path from within the feasible domain. The combination of the two methods ensures the efficiency and accuracy of the solution in a complex production environment. The generated optimized execution configuration scheme specifies the optimal configuration and time schedule of each resource in detail, improving the overall operating efficiency and resource utilization of the production line.
[0088] In the feedback adjustment submodule, based on the optimized execution configuration plan, a feedback control strategy is used to monitor production execution in real time and collect feedback information. The data format includes real-time production data and the optimized execution configuration plan. The real-time production data covers various monitoring indicators in the production process, such as output, quality level and resource consumption. By continuously analyzing the deviation between production data and the optimized execution configuration plan, anomalies and improvement opportunities in the production process are identified. The feedback control strategy adjusts production parameters according to the deviation analysis results, such as adjusting equipment operating speed, raw material supply ratio and quality control standards to match real-time production needs and conditions. This operation ensures the flexibility and adaptability of production execution and realizes the continuous optimization of the production process. The generated optimized production configuration plan improves production efficiency and product quality, ensuring that production activities can respond quickly to market changes.
[0089] See also Figure 2 , 4,The equipment health management module includes a condition monitoring submodule, a fault prediction submodule, and a health assessment submodule;
[0090] The status monitoring submodule uses the isolation forest algorithm to analyze abnormal points in the equipment operation data based on the optimized production configuration plan. By constructing multiple isolated trees, the abnormal points are isolated on a few branches, the abnormal points in the data are identified, and the equipment abnormality monitoring results are generated;
[0091] The fault prediction submodule uses an autoencoder neural network to encode and decode the equipment status features based on the equipment abnormality monitoring results, learns the data representation by compressing the input data, reconstructs the data, and generates fault prediction results by capturing the key features in the data;
[0092] The health assessment submodule uses support vector machines to conduct a comprehensive assessment of the equipment health status based on the fault prediction results. It constructs the optimal decision boundary in the multidimensional space, separates the differentiated health status categories, and generates the equipment health assessment results.
[0093] In the condition monitoring submodule, the isolation forest algorithm is used to analyze the abnormal points of the production equipment operation data. The submodule first receives the equipment operation data. The data format includes time series data, including multi-dimensional indicators such as the equipment's operating rate, temperature, and vibration. The isolation forest algorithm realizes the identification of abnormal points by constructing multiple isolated trees, including randomly selecting data features and their segmentation values as tree nodes, dividing the data set into two subsets, and repeating this process until the data point is isolated or the set tree depth limit is reached. In the process of building the isolated tree, the abnormal data points are isolated at a lower depth of the tree due to the abnormal characteristics of their values, while normal data points require more segmentation to be isolated. By evaluating the average depth of the isolated data points, the algorithm can effectively identify the abnormal points. The execution of this method does not rely on the distribution assumption of the data, and is particularly suitable for processing real-time production data with complex distribution. The condition monitoring submodule finally generates the equipment abnormal monitoring results, which clearly identify the abnormal points in the operation data, provide basic data for subsequent fault prediction, and effectively improve the accuracy and efficiency of equipment monitoring.
[0094] In the fault prediction submodule, the autoencoder neural network is used to deeply learn the equipment status features in the equipment abnormality monitoring results. This submodule uses the encoding and decoding mechanism of the autoencoder to compress and reconstruct the input equipment status data. The autoencoder neural network first encodes the high-dimensional equipment status data into a low-dimensional feature representation, and then attempts to reconstruct the original data through the decoding process. In the process, the network is trained by minimizing the reconstruction error. The network can learn the intrinsic characteristics and structure of the data, so that the model can capture the key features representing the abnormal state of the equipment. The design of the network structure includes the input layer, hidden layer (encoding layer and decoding layer) and output layer. By adjusting the dimension and activation function of the hidden layer, the autoencoder neural network can effectively learn the nonlinear relationship in the data. The fault prediction results generated by the fault prediction submodule can not only accurately predict the potential faults of the equipment, but also further understand the intrinsic mechanism of the fault occurrence by analyzing the differences in feature reconstruction.
[0095] In the health assessment submodule, the health status of the equipment is comprehensively evaluated by the support vector machine (SVM). Based on the equipment status characteristics in the fault prediction results, this submodule uses the classification mechanism of SVM to find the optimal decision boundary in the multidimensional space to distinguish different equipment health states. The support vector machine realizes data classification by constructing one or more hyperplanes, and the optimization goal is to maximize the interval between different categories of data. It includes selecting a suitable kernel function (such as a linear kernel, a polynomial kernel, or a radial basis kernel) to transform the data into a high-dimensional space, where linear separation is easier to achieve. By adjusting the penalty parameter C and the kernel function parameters, the SVM model can achieve highly accurate classification on complex data sets. In this way, the health assessment submodule generates the health assessment results of the equipment, which not only includes the current health status classification of the equipment, but also provides early warning information about changes in the health status of the equipment, providing a scientific basis for the maintenance and repair of the equipment, and optimizing the operation and maintenance strategy of the equipment.
[0096] See also Figure 2 , 5 ,The quality control management module includes an image acquisition submodule, a defect recognition submodule, and a quality feedback submodule;
[0097] The image acquisition submodule collects product images on the production line based on the equipment health assessment results, uses the camera to capture product surface details, provides raw image data for defect identification, and generates a product image dataset;
[0098] The defect recognition submodule uses a convolutional neural network to automatically extract image features and classify them based on the product image dataset. Through the convolutional layer, pooling layer, and fully connected layer in the deep network structure, it captures image features and identifies defects of different types to generate defect recognition results.
[0099] The quality feedback submodule uses a data-driven decision-making model to monitor and provide feedback on product quality in real time based on defect identification results. By analyzing historical quality data and real-time feedback, it identifies quality trends and anomalies, guides production adjustments to optimize product quality, and generates quality control results.
[0100] In the image acquisition submodule, images of products on the production line are acquired based on the equipment health assessment results. During the acquisition process, the camera device automatically adjusts parameters such as focal length, exposure time, and light intensity according to the assessment results to ensure the capture of high-quality product surface detail images. The image data uses high-resolution JPEG or PNG format, contains the full picture of the product and its details, and provides the necessary raw data for defect identification. This module automatically triggers the image acquisition process, responds to the current status of the equipment and the speed of the production line, optimizes the acquisition time, reduces interference, and ensures the consistency and comparability of the image data. The generated product image dataset is pre-processed, such as cropping, rotation, and scale adjustment, to unify the image format and size, and provide standardized input for the subsequent defect recognition module, ensuring the image quality and the integrity of the dataset, providing a basis for recognition accuracy, and directly affecting the efficiency and accuracy of defect recognition.
[0101] In the defect recognition submodule, the features in the product image are automatically extracted and classified through the convolutional neural network (CNN), and the defects of different types are accurately identified. The CNN architecture includes multiple convolutional layers, pooling layers, and at least one fully connected layer. Each layer performs specific mathematical operations on the input image to extract and compress image features. The convolutional layer uses filters to scan the local area of the image to capture basic features such as edges and textures; the pooling layer reduces the number of parameters and computational complexity by downsampling while retaining important feature information. The fully connected layer integrates the output of the previous layer and outputs the final classification result. This submodule uses the training data set to train the CNN model, optimize the weight and bias parameters, and improve the accuracy of defect recognition. The generated defect recognition results annotate the defect type for each product image, support subsequent quality control decisions, and improve the automation and intelligence level of product quality monitoring.
[0102] In the quality feedback submodule, real-time monitoring and feedback of product quality are carried out through a data-driven decision model. This module comprehensively utilizes historical quality data and real-time defect identification results, and adopts statistical analysis and machine learning algorithms, such as decision trees or random forests, to identify product quality trends and abnormal patterns. The model analyzes the distribution, fluctuations and anomalies of product quality data, and evaluates systemic problems or incidental events in the production process. Based on the analysis results, the module automatically generates quality control feedback, including adjustment plans and early warnings, to guide the production line to adjust parameters or processes and optimize product quality. The generated quality control result file records the quality trend analysis, identified problems and solution measures, provides decision support for production management, realizes dynamic monitoring and continuous improvement of product quality, improves production efficiency and product qualification rate, reduces rework and scrap rate, and enhances the market competitiveness of enterprises.
[0103] See also Figure 2 , 6 ,The security monitoring management module includes a feature extraction submodule, anomaly detection submodule, and alarm generation submodule;
[0104] The feature extraction submodule uses the isolation forest algorithm to extract security features from production data based on quality control results. It constructs an isolation tree by randomly selecting features and split values, identifies abnormal patterns in the data, and generates security feature analysis results.
[0105] The anomaly detection submodule uses anomaly point detection technology to analyze the security feature analysis results. It calculates the degree of outliers of data points, identifies data anomaly patterns, provides a basis for security risk management, and generates anomaly detection results.
[0106] The alarm generation submodule uses rule-based decision-making to evaluate security warnings based on the anomaly detection results. Through preset security rules and logical judgments, it analyzes the severity and potential impact of abnormal patterns, forms targeted security measures and warning information, and generates security monitoring results.
[0107] In the feature extraction submodule, the isolation forest algorithm is used to extract security features from the production data in the quality control results. The data format used in this submodule is mainly various monitoring indicators and operating parameters in the production process, including temperature, pressure, speed, etc., to form a high-dimensional data set. The isolation forest algorithm constructs an isolation tree by randomly selecting features and split values. Its uniqueness lies in its randomness and simplicity. It can effectively process high-dimensional data and does not require any assumptions for the detection of outliers. The algorithm first randomly selects a feature, then randomly selects a split value within the value range of this feature, and divides the data set into two parts with this split value as the boundary. The process is repeated until the data is completely isolated or the preset tree depth is reached. In a forest composed of multiple isolated trees, normal data points require more splits to be isolated, while abnormal data are more likely to be isolated early due to their small number or unique characteristics. By calculating the average path length of the data points isolated in the forest, the security feature analysis results are generated. This result lists the abnormal score of each data point in detail, thereby identifying abnormal patterns in the production process.
[0108] In the anomaly detection submodule, the security features extracted based on the isolation forest algorithm are analyzed by the outlier detection technology. This submodule inherits the security feature analysis results generated by the previous module and takes the anomaly score of the data point as input. The outlier detection technology evaluates the degree of outliers of the data point by calculating the distance between each data point and its neighboring points. The core algorithm is to define a metric that identifies abnormal patterns in the data based on the position of the outlier degree of the data point relative to the overall data set. The algorithm takes into account the distance and density between data points, so that abnormal data points can be accurately marked. Through this technology, points that are significantly different from the majority of the data can be identified, providing a reliable basis for security risk management. The generated anomaly detection results are saved in the form of a file, which records in detail each data point identified as an anomaly and its corresponding outlier degree score, providing accurate data support for the next step of risk assessment and preventive measures.
[0109] In the alarm generation submodule, rule-based decision evaluation is used to issue safety warnings for anomaly detection results. Based on the anomaly detection results generated by the anomaly detection submodule, this submodule uses preset safety rules and logical judgments to analyze the identified abnormal patterns. Rule-based decision evaluation evaluates the severity and potential impact of abnormal data points by defining a series of rules and thresholds. For example, if the degree of outlier of a data point exceeds the set threshold, the system determines that the anomaly causes production safety risks, and then triggers corresponding warnings, which include not only simple numerical comparisons, but also comprehensive judgments on the causes and impacts behind the abnormal data, so as to form more targeted safety measures. The generated safety monitoring results are provided in the form of a report, which not only lists each abnormal data point that triggers the alarm, but also includes targeted warning information and recommended safety measures, aiming to guide risk management and emergency response in the production process, thereby effectively reducing safety risks in the production process.
[0110] See also Figure 2 , 7 ,The demand and supply forecasting module includes a data collating submodule, a forecasting model building submodule, and a forecasting result analysis submodule;
[0111] The data collation submodule uses data processing techniques based on the safety monitoring results, including standardizing the data using the Z-score method, processing outliers by identifying and removing outliers through the IQR method, and filling missing values through linear interpolation to generate a time series data set;
[0112] The forecast model construction submodule is based on the time series data set and uses a long short-term memory network combined with a temporal convolutional network. The long short-term memory network captures the long-term dependencies in the time series through its gating mechanism, and the temporal convolutional network optimizes the efficiency and accuracy of processing sequence data through the causal convolution layer to generate demand and supply forecast results.
[0113] The forecast result analysis submodule is based on the demand and supply forecast results. It uses trend analysis methods to calculate the difference between the forecast value and the real-time value, identify the randomness of demand peaks and supply shortages, provide data support for production decisions in future time periods, and generate comprehensive forecast analysis results.
[0114] In the data sorting submodule, the data is processed through the safety monitoring results, and a variety of data processing techniques are used to ensure data quality and consistency. First, the Z-score method is used to standardize the data, which involves calculating the deviation of each data point from the mean of the data set and dividing it by the standard deviation, thereby converting the data to a distribution with zero mean and unit variance, making data of different scales and dimensions comparable, and improving the accuracy of subsequent analysis. Next, outliers are identified and eliminated by the IQR (interquartile range) method. This method calculates the IQR value based on the first quartile (Q1) and the third quartile (Q3), and defines the range of outliers based on this, thereby effectively excluding outliers in the data set and reducing their negative impact on model training and analysis results. For missing values, linear interpolation is used to fill them. This method assumes that the data has a linear trend and calculates missing values through adjacent data points, maintaining the continuity of the data. After step processing, it provides high-quality input data for the subsequent prediction model construction.
[0115] In the prediction model construction submodule, the long short-term memory network (LSTM) combined with the temporal convolutional network (TCN) is used to build a prediction model through the time series data set. LSTM uses its gating mechanism to effectively capture the long-term dependency in the time series, controls the flow of information through the input gate, forget gate and output gate, retains useful information and forgets irrelevant information, thereby solving the gradient vanishing or gradient exploding problem of traditional recurrent neural networks when processing long sequence data. At the same time, TCN uses the causal convolution layer to ensure that only past information is used when predicting the value at the current moment, and future information will not be leaked. Its dilated convolution increases the receptive field and improves the ability to process long time series. The model combining LSTM and TCN can not only capture the long-term dependency characteristics of time series, but also improve the efficiency and accuracy of processing sequence data. By adjusting parameters such as the number of hidden layer units of LSTM and the depth of the convolution layer of TCN, the model is trained to minimize the difference between the predicted value and the actual value, and finally generates demand and supply forecast results, providing support for accurate demand forecasting and supply chain management.
[0116] In the forecast result analysis submodule, the forecast data is deeply analyzed by trend analysis method through the demand and supply forecast results. Trend analysis evaluates the accuracy and reliability of the forecast model by calculating the difference between the forecast value and the real-time value. By identifying patterns in the differences, such as the frequency of demand peaks and supply shortage conditions, the analysis submodule can reveal the changing trends and randomness of demand and supply. The analysis not only helps to evaluate the performance of the model, but also provides data support for production decisions and guides the adjustment of production plans to cope with market demand fluctuations. The statistical methods and pattern recognition techniques used in the analysis process further refine the forecast results. The generated comprehensive forecast analysis report describes in detail the accuracy of the forecast results, potential risks and their impact on future production decisions, thus providing management with a data-based decision-making tool that enables more effective production planning based on more accurate demand forecasts.
[0117] See also Figure 2 , 8 ,The intelligent scheduling optimization module includes a scheduling strategy submodule, a resource allocation submodule, and an optimization feedback submodule;
[0118] The scheduling strategy submodule uses genetic algorithms to design the initial production scheduling plan based on the comprehensive prediction and analysis results. After initializing the population, the roulette wheel selection method is used to select excellent individuals, the single-point crossover method is used to exchange individual information, and the uniform mutation method is used to introduce new genetic information. The optimal production scheduling strategy is captured through continuous iterative optimization and generates a preliminary scheduling optimization strategy.
[0119] The resource allocation submodule uses the simulated annealing algorithm to optimize the resource allocation problem based on the preliminary scheduling optimization strategy. By setting the initial temperature and cooling plan, the probability of accepting a poor solution is gradually reduced as the temperature decreases, the global optimal resource allocation strategy is captured, and the resource allocation optimization plan is generated;
[0120] The optimization feedback submodule implements a dynamic adjustment and feedback mechanism based on the resource allocation optimization plan, adjusts the production plan and resource allocation according to the feedback information in the real-time production process, uses real-time data monitoring and adjustment strategies, and generates scheduling optimization plans.
[0121] In the scheduling strategy submodule, the preliminary optimization of the production scheduling strategy is achieved through genetic algorithms, which involves the analysis of production tasks and resources, including task duration, resource requirements, and dependencies between tasks. The genetic algorithm starts with the initialization of a set of solutions (populations), representing different production scheduling solutions. Each solution (individual) is represented by a string of codes, which is a binary code or a task sequence. After the initial population is generated, the algorithm selects the better individual through the roulette selection method, and based on the fitness score of the individual, the individual with high fitness is preferentially selected for subsequent operations. The single-point crossover method is used to exchange information between two individuals, that is, select an intersection point and exchange the code part before or after this point to generate a new individual. The uniform mutation rule randomly changes the code in the individual, introduces new genetic information to explore the solution space, including selection, crossover, and mutation, and the genetic algorithm continuously optimizes to find the optimal production scheduling strategy. The preliminary optimization strategy finally generated by this submodule describes in detail the execution order of production tasks and resource allocation plan, effectively improving production efficiency and resource utilization.
[0122] In the resource allocation submodule, the resource allocation problem based on the scheduling preliminary optimization strategy is further optimized by the simulated annealing algorithm. The algorithm simulates the annealing process in the physical process, starting from a high temperature state and gradually cooling until the system is stable. The initial temperature is set high to allow the algorithm to accept poor solutions in the early stage. As the temperature decreases, the algorithm tends to accept better solutions and reduce the probability of accepting poor solutions. The data format of the resource allocation problem covers information such as the type and quantity of resources on the production line, the demand for resources by tasks, and time constraints. The simulated annealing algorithm defines an initial solution, based on the scheduling preliminary optimization strategy, and then performs a series of random perturbations at each temperature to explore new resource allocation schemes. After each perturbation, the quality of the new scheme determines whether to accept or reject the scheme. As the temperature gradually decreases, the algorithm gradually converges to a global optimal or approximately global optimal resource allocation scheme. The generated resource allocation optimization scheme clearly points out the resource allocation details of each production task, including the allocation time and quantity of resources, which helps to balance the load of the production line and improve the efficiency and responsiveness of the overall production system.
[0123] In the optimization feedback submodule, the dynamic adjustment and feedback mechanism is implemented based on the resource allocation optimization scheme. This submodule adopts real-time data monitoring and strategy adjustment mechanism, and dynamically adjusts the production plan and resource allocation based on the feedback information in the real-time production process. The data format includes real-time production data, resource utilization status, production progress and other information. The implementation strategy includes collecting real-time production data, evaluating the deviation between the current production status and the predetermined plan, and adjusting the execution order of production tasks or resource allocation plan according to the deviation results. By continuously monitoring the status on the production line, the optimization feedback mechanism can promptly discover problems in the production process, respond quickly based on real-time data, and adjust the plan to match the actual production situation. The generated scheduling optimization plan comprehensively considers the variables and uncertainties in actual production, improves the flexibility and adaptability of production scheduling, and ensures efficient and stable production process.
[0124] See also Figure 2 , 9 ,The fault diagnosis and maintenance module includes a fault data analysis submodule, a fault diagnosis submodule, and a maintenance plan submodule;
[0125] The fault data analysis submodule optimizes the fault data based on the scheduling optimization scheme, uses data preprocessing technology to optimize the fault data, processes outliers, uses Z-score to remove abnormal data points and missing values, and uses linear interpolation to fill in missing data to generate a processed fault data set;
[0126] The fault diagnosis submodule uses a deep neural network to perform fault diagnosis based on the processed fault data set. It captures fault features through the convolution layer, merges features through the fully connected layer, and uses the deep network structure to learn and identify fault modes to generate fault diagnosis and prediction results.
[0127] The maintenance plan submodule uses decision support technology to analyze the fault type and frequency based on the fault diagnosis and prediction results, and uses fault tree analysis and risk assessment to determine the priority and strategy of maintenance measures and generate preventive maintenance measures.
[0128] In the fault data analysis submodule, the fault data is finely optimized by using the data preprocessing technology through the scheduling optimization scheme to ensure the quality and availability of the data. The fault data is collected from the equipment sensors and operation logs and formatted into a structured data set, which contains information such as timestamp, fault code, equipment identification and operation status. The first step of data preprocessing is outlier processing. The Z score method is used to identify and remove abnormal data points in the data set. The Z score calculates the degree of deviation of each data point from the mean value in units of standard deviation. After removing the outliers, the missing values are processed and the missing data are filled by linear interpolation. This method estimates the missing values based on adjacent data points to maintain the continuity and consistency of the data. The operation improves the accuracy and completeness of the data set and provides reliable input for fault diagnosis. The generated processed fault data set is used for deep learning model training and fault mode identification, ensuring the effectiveness of data analysis and the accuracy of fault diagnosis.
[0129] In the fault diagnosis submodule, the processed fault data set is diagnosed through a deep neural network. The structural design of the deep neural network contains multiple convolutional layers and fully connected layers, which are specifically used to capture fault features and learn fault modes. The convolutional layer automatically extracts relevant fault features, such as frequency distribution, vibration mode, etc., by applying multiple filters to the input data, while the fully connected layer integrates the features for the final fault classification. The network continuously adjusts parameters through the back-propagation algorithm to minimize the difference between the predicted results and the actual labels. After training, the network can accurately diagnose new fault data and generate fault diagnosis and prediction results, including fault type and occurrence time, providing a scientific basis for subsequent maintenance decisions and significantly improving the efficiency and accuracy of fault handling.
[0130] In the maintenance plan submodule, the fault type and frequency are analyzed by decision support technology through fault diagnosis and prediction results. The determination of maintenance strategy relies on fault tree analysis and risk assessment to systematically identify and evaluate the causes of faults and their impact on production. Fault tree analysis provides a logical structure that reversely traces the various conditions and events that lead to faults, while risk assessment focuses on assessing the probability of faults and the severity of consequences, thereby determining the priority of maintenance measures, taking into account the complexity and diversity of faults, and using a combination of quantitative and qualitative analysis methods to provide comprehensive information for the formulation of maintenance plans. The generated preventive maintenance measures list in detail the priorities, recommended maintenance activities and schedules for various types of faults, optimize the allocation of maintenance resources, reduce equipment downtime, and improve production efficiency and equipment reliability.
[0131] See also Figure 2 , 10 ,The decision support module includes data summary and analysis submodule, decision generation submodule, and execution feedback submodule;
[0132] The data summary and analysis submodule summarizes and analyzes production, maintenance and fault data based on preventive maintenance measures, using data warehouse and big data analysis technology, provides the data views required for decision-making, and generates a comprehensive decision-making data set;
[0133] The decision generation submodule performs decision analysis based on the comprehensive decision data set, combining the rule engine and decision tree. The rule engine processes information based on logical rules, and the decision tree evaluates the randomness of decision results through path analysis to generate decision support solutions.
[0134] The execution feedback submodule implements decision execution monitoring and feedback based on the decision support solution, evaluates the execution effect of the decision through performance indicator monitoring and deviation analysis, provides data support for decision adjustments in future time periods, and generates execution effect feedback results.
[0135] In the data aggregation and analysis submodule, production, maintenance and failure data are aggregated and analyzed through data warehouse and big data analysis technology. The core of this submodule is to build a comprehensive data warehouse, which contains a large amount of data generated during the production process, such as production line speed, product quality test results, machine failure records and maintenance activity records. The data warehouse design adopts a star schema for fast query and analysis, and the data format is standardized to ensure the consistency and comparability of various data sources. Big data analysis technology is applied to this data set, and the MapReduce programming model is used to process data in a distributed manner to achieve efficient data aggregation, filtering and transformation operations. In particular, frameworks such as Apache Hadoop and Apache Spark support the processing of PB-level data sets, ensuring the scalability and efficiency of processing. By applying data mining algorithms, such as cluster analysis to identify potential problems in the production process, association rule mining to discover the correlation between maintenance activities and machine failures. The generated comprehensive decision data set is provided in the form of a report, which lists the statistical analysis results of key performance indicators (KPIs) in detail, providing data views and in-depth insights for decision-making, supporting management to make more informed strategic adjustments and resource allocation decisions.
[0136] In the decision generation submodule, decision analysis is performed through the rule engine and decision tree. Based on the comprehensive decision data set generated by the data summary analysis submodule, this submodule uses the rule engine to process information based on logical rules, and the decision tree analysis evaluates the randomness of the decision results. The rule engine automates the decision logic by encoding business logic and policy decision rules. For example, if the failure frequency of a machine exceeds the threshold, it recommends replacement or deep maintenance. As a predictive model, the decision tree analyzes the results of different decision paths by constructing a tree structure to evaluate the risks and benefits of each decision. The construction process of the decision tree includes the selection of node splitting criteria, such as information gain or Gini coefficient, to maximize the purity of the classification. Through the combined use of the rule engine and decision tree, this submodule can generate specific and feasible decision support solutions for complex production and maintenance scenarios. The generated solution details the strategies adopted, the expected effects and risks, helping management to make more scientific and reasonable decisions in a complex and changing production environment.
[0137] In the execution feedback submodule, the execution effect of the decision is evaluated through performance indicator monitoring and deviation analysis. Based on the monitoring and feedback mechanism implemented by the decision support program, this submodule uses a series of pre-defined performance indicators, such as production efficiency, failure rate and maintenance cost, to monitor the actual effect of decision execution through real-time data collection and analysis. Deviation analysis is the core operation of this submodule. By comparing the difference between the expected effect of the decision and the actual execution result, problems and deviations in the execution process are identified. Statistical analysis methods such as analysis of variance (ANOVA) and control chart technology are used to quantify the severity of deviations and analyze the causes of deviations. Through continuous monitoring and deviation analysis, this submodule can adjust and optimize the decision plan in a timely manner to improve the accuracy and execution efficiency of the decision. The generated execution effect feedback results are provided in the form of reports, which record in detail the trend of performance indicators, problems identified in the execution process and their cause analysis, providing data support and empirical basis for future decision adjustment and optimization.
[0138] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.
Claims
1. An Internet intelligent manufacturing service platform, characterized by: The platform includes a production optimization scheduling module, an equipment health management module, a quality control management module, a safety monitoring management module, a demand and supply forecasting module, an intelligent scheduling optimization module, a fault diagnosis and maintenance module, and a decision support module; The production optimization scheduling module uses a dynamic ecological optimization algorithm based on real-time production line data to simulate the competition and symbiotic relationship of species in the natural ecology. Through the selection, crossover and mutation operations within the algorithm, the production process and configuration are dynamically adjusted to match environmental changes and generate an optimized production configuration plan; The production optimization scheduling module includes an ecological model submodule, an optimization execution submodule, and a feedback adjustment submodule; The ecological model submodule uses a dynamic ecological optimization algorithm to simulate the competition and symbiosis mechanism of species in the natural ecology based on real-time production line data, and dynamically adjusts the production process and configuration by simulating the selection, crossover and mutation operations in the biological evolution process to generate an ecological optimization configuration plan; The optimization execution submodule uses a linear programming method to optimize the production process and equipment configuration based on the ecological optimization configuration scheme, and uses the simplex method and interior point method solvers to capture the optimal solution of resource configuration and generate an optimization execution configuration scheme by establishing an objective function and constraints; The feedback adjustment submodule uses a feedback control strategy to perform real-time monitoring of production execution and collection of feedback information based on the optimized execution configuration plan, adjusts production parameters through continuous analysis and comparison of production data, matches real-time production requirements and conditions, and generates an optimized production configuration plan; The equipment health management module uses the isolation forest algorithm to analyze abnormal points in equipment operation data based on the optimized production configuration plan, identify potential equipment failures, and encode and decode equipment status features in combination with the autoencoder neural network to predict equipment failures and generate equipment health assessment results; The quality control management module uses a convolutional neural network based on the equipment health assessment results to automatically extract features in the image through stacked filters and classify them, identify and mark product defects on the production line, monitor product quality in real time, and generate quality control results; The safety monitoring management module uses the isolation forest algorithm to analyze production data based on the quality control results, identifies abnormal patterns in the data by building an isolation tree, identifies safety risks in the production process, and generates safety monitoring results; The demand and supply forecasting module uses a long short-term memory network to process time series data based on safety monitoring results, learns historical production demand and raw material supply patterns, and uses a time convolutional network to optimize the efficiency and accuracy of sequence data processing, provides forecasts for production demand and raw material supply in future time periods, and generates comprehensive forecast analysis results; The intelligent scheduling optimization module uses genetic algorithms to simulate natural selection and genetic mechanisms to optimize the solution of production scheduling problems based on comprehensive prediction and analysis results, and uses simulated annealing algorithms to perform random analysis to overcome local optimal problems, dynamically adjust production plans and resource allocation, and generate scheduling optimization solutions; The intelligent scheduling optimization module includes a scheduling strategy submodule, a resource allocation submodule, and an optimization feedback submodule; The scheduling strategy submodule uses a genetic algorithm to design an initial production scheduling plan based on the comprehensive prediction and analysis results. After initializing the population, the roulette wheel selection method is used to select excellent individuals, the single point crossover method is used to exchange individual information, and the uniform mutation method is used to introduce new genetic information. The optimal production scheduling strategy is captured through continuous iterative optimization and generates a preliminary scheduling optimization strategy. The resource allocation submodule optimizes the resource allocation problem based on the preliminary scheduling optimization strategy using a simulated annealing algorithm. By setting the initial temperature and cooling plan, the probability of accepting a poor solution is gradually reduced as the temperature decreases, thereby capturing the global optimal resource allocation strategy and generating a resource allocation optimization solution. The optimization feedback submodule implements a dynamic adjustment and feedback mechanism based on the resource allocation optimization plan, adjusts the production plan and resource allocation according to the feedback information in the real-time production process, uses real-time data monitoring and adjustment strategies, and generates a scheduling optimization plan; The fault diagnosis and maintenance module uses a deep learning model based on the scheduling optimization scheme to learn the equipment fault data pattern through a deep neural network, diagnose and predict the fault type, and generate preventive maintenance measures; The decision support module is based on preventive maintenance measures, uses a rule engine to process information based on predefined decision logic, combines a decision tree to analyze decision results through a tree-structured decision path, and generates execution effect feedback results.
2. The Internet intelligent manufacturing service platform according to claim 1, characterized in that: The optimized production configuration plan includes production rhythm adjustment parameters, robot arm scheduling priorities, and raw material delivery speed settings. The equipment health assessment results include equipment maintenance cycle recommendations, key component wear assessments, and fault warning levels. The quality control results include defect size ranges, defect location information, and product qualification rate statistics. The safety monitoring results include safety risk levels, hazardous area identification, and emergency shutdown trigger points. The comprehensive forecasting and analysis results include production demand increase and decrease trends, raw material gap forecasts, and supply chain stability scores. The scheduling optimization plan includes optimal production line configuration, key production node priorities, and backup resource allocation strategies. The preventive maintenance measures include maintenance time window arrangements, key component inspection lists, and spare parts inventory plans. The execution effect feedback results include production adjustment decision paths, risk mitigation measures selection, and benefit-maximizing resource allocation.
3. The Internet intelligent manufacturing service platform according to claim 1, characterized in that: The equipment health management module includes a status monitoring submodule, a fault prediction submodule, and a health assessment submodule; The state monitoring submodule uses the isolation forest algorithm to analyze abnormal points in the equipment operation data based on the optimized production configuration scheme, isolates the abnormal points on a few branches by constructing multiple isolated trees, identifies the abnormal points in the data, and generates equipment abnormality monitoring results; The fault prediction submodule uses an autoencoder neural network to encode and decode the equipment status features based on the equipment abnormality monitoring results, learns the representation of the data by compressing the input data, reconstructs the data, and generates fault prediction results by capturing the key features in the data; The health assessment submodule uses a support vector machine to perform a comprehensive assessment of the health status of the equipment based on the fault prediction results, and generates equipment health assessment results by constructing an optimal decision boundary in a multidimensional space to separate differentiated health status categories.
4. The Internet intelligent manufacturing service platform according to claim 1, characterized in that: The quality control management module includes an image acquisition submodule, a defect recognition submodule, and a quality feedback submodule; The image acquisition submodule acquires product images on the production line based on the equipment health assessment results, uses a camera to capture product surface details, provides raw image data for defect identification, and generates a product image dataset; The defect recognition submodule uses a convolutional neural network to automatically extract image features and classify them based on the product image dataset. Through the convolutional layer, pooling layer and fully connected layer in the deep network structure, it captures image features and identifies defects of different types to generate defect recognition results. The quality feedback submodule uses a data-driven decision model to monitor and feedback product quality in real time based on defect identification results. It identifies quality trends and anomalies by analyzing historical quality data and real-time feedback, guides production adjustments to optimize product quality, and generates quality control results.
5. The Internet intelligent manufacturing service platform according to claim 1, characterized in that: The security monitoring management module includes a feature extraction submodule, an anomaly detection submodule, and an alarm generation submodule; The feature extraction submodule extracts security features from production data using an isolation forest algorithm based on quality control results, constructs an isolation tree by randomly selecting features and split values, identifies abnormal patterns in the data, and generates security feature analysis results; The anomaly detection submodule uses anomaly point detection technology to analyze the security feature analysis results, calculates the degree of outliers of data points, identifies data anomaly patterns, provides a basis for security risk management, and generates anomaly detection results; The alarm generation submodule uses rule-based decision-making to evaluate security warnings based on the abnormal detection results. Through preset security rules and logical judgments, it analyzes the severity and potential impact of abnormal patterns, forms targeted security measures and warning information, and generates security monitoring results.
6. The Internet intelligent manufacturing service platform according to claim 1, characterized in that: The demand and supply forecasting module includes a data sorting submodule, a forecasting model building submodule, and a forecasting result analysis submodule; The data sorting submodule uses data processing techniques based on the safety monitoring results, including using the Z-score method to standardize the data, identifying and eliminating outliers through the IQR method, and filling missing values through linear interpolation to generate a time series data set; The forecast model building submodule is based on a time series data set and uses a long short-term memory network combined with a temporal convolutional network. The long short-term memory network captures the long-term dependencies in the time series through its gating mechanism, and the temporal convolutional network optimizes the efficiency and accuracy of processing sequence data through a causal convolutional layer to generate demand and supply forecast results. The forecast result analysis submodule uses trend analysis methods based on demand and supply forecast results to identify the randomness of demand peaks and supply shortages by calculating the difference between the forecast value and the real-time value, providing data support for production decisions in future time periods and generating comprehensive forecast analysis results.
7. The Internet intelligent manufacturing service platform according to claim 1, characterized in that: The fault diagnosis and maintenance module includes a fault data analysis submodule, a fault diagnosis submodule, and a maintenance solution submodule; The fault data analysis submodule optimizes the fault data based on the scheduling optimization scheme by using data preprocessing technology, removes abnormal data points and missing values through outlier processing, and fills missing data with linear interpolation to generate a processed fault data set; The fault diagnosis submodule uses a deep neural network to perform fault diagnosis based on the processed fault data set, captures fault features through a convolutional layer, merges features through a fully connected layer, and uses a deep network structure to learn and identify fault modes to generate fault diagnosis and prediction results; The maintenance plan submodule uses decision support technology to analyze the fault type and frequency based on the fault diagnosis and prediction results, and uses fault tree analysis and risk assessment to determine the priority and strategy of maintenance measures to generate preventive maintenance measures.
8. The Internet intelligent manufacturing service platform according to claim 1, characterized in that: The decision support module includes a data summary and analysis submodule, a decision generation submodule, and an execution feedback submodule; The data summary and analysis submodule summarizes and analyzes production, maintenance and fault data based on preventive maintenance measures, using data warehouse and big data analysis technology, provides data views required for decision-making, and generates a comprehensive decision-making data set; The decision generation submodule performs decision analysis based on the comprehensive decision data set in combination with the rule engine and the decision tree. The rule engine processes information based on logical rules, and the decision tree evaluates the randomness of the decision results through path analysis to generate a decision support plan. The execution feedback submodule implements decision execution monitoring and feedback based on the decision support solution, evaluates the execution effect of the decision through performance indicator monitoring and deviation analysis, provides data support for decision adjustments in future time periods, and generates execution effect feedback results.
