Intelligent device operation and maintenance management method and system based on industrial big data

Abstract

This invention provides a method and system for intelligent equipment operation and maintenance management based on industrial big data, belonging to the field of equipment operation and maintenance management technology. The method includes: collecting equipment operation and environmental parameters to construct a multi-source heterogeneous dataset; preprocessing, storing, and extracting time-series topological features to output standardized data; combining a rule base and multi-objective optimization to generate preliminary operation and maintenance instructions; optimizing through external system collaboration and constraints; issuing and executing the instructions; and providing feedback on the results. This invention achieves early warning through multi-source data fusion and deep learning feature extraction; avoids system cascading failures caused by localized faults by leveraging topological correlation analysis and external collaboration; balances cost, availability, and risk through multi-objective optimization and risk screening to achieve data-driven decision-making; and dynamically updates maintenance plans and spare parts strategies based on remaining life prediction to improve the predictability and economy of operation and maintenance.

Description

Technical Field

[0001] This invention belongs to the field of equipment operation and maintenance management technology, specifically a method and system for intelligent equipment operation and maintenance management based on industrial big data. Background Technology

[0002] In the current industrial production environment, equipment operation and maintenance management is facing increasingly severe challenges. The traditional manual inspection and regular maintenance model is no longer able to meet the operational needs of today's large-scale and highly complex production lines.

[0003] Existing management methods generally suffer from significant lag, often relying on the experience and judgment of on-site personnel. This makes it difficult to detect minor anomalies in equipment in their early stages, leading to unplanned downtime and severely impacting production continuity. Simultaneously, while the sheer volume of monitoring data generated by various sensors on the production floor is enormous, it is largely fragmented and stored in isolation, lacking effective integration and in-depth analysis mechanisms. This results in a severe underestimation of data value and an inability to form a comprehensive understanding of the overall equipment status. Furthermore, equipment does not operate completely independently but is subject to complex physical connections and process couplings. Current maintenance decisions often focus on individual devices, ignoring the propagation effects of faults between upstream and downstream equipment, easily triggering chain reactions. At the maintenance decision-making level, the lack of scientific quantitative basis makes it difficult for managers to find the optimal balance between maintenance costs, production efficiency, and risk control, frequently resulting in resource waste due to over-maintenance or frequent failures due to under-maintenance. Summary of the Invention

[0004] This invention provides a method and system for intelligent equipment operation and maintenance management based on industrial big data, in order to solve the defects existing in the prior art.

[0005] On the one hand, this invention provides an intelligent equipment operation and maintenance management method based on industrial big data, including:

[0006] Collect real-time operating status parameters and environmental condition parameters of industrial equipment to construct a multi-source heterogeneous operation and maintenance dataset.

[0007] Preprocessing, storing, and extracting time-series topological features from multi-source heterogeneous operation and maintenance datasets, and outputting standardized operation and maintenance feature data.

[0008] Based on standardized operation and maintenance characteristic data, preliminary operation and maintenance decision instructions are generated by combining a predefined operation and maintenance rule base with a multi-objective optimization algorithm.

[0009] By combining the collaborative constraints of the external equipment health management system, the preliminary operation and maintenance decision instructions are dynamically optimized and updated, and the final operation and maintenance decision instructions are output.

[0010] The final operation and maintenance decision instructions are sent to the target equipment operation and maintenance execution terminal, and the execution status is monitored and the execution results are fed back.

[0011] The intelligent equipment operation and maintenance management method based on industrial big data provided by this invention includes the following processes for preprocessing, storing, and extracting time-series topological features from multi-source heterogeneous operation and maintenance datasets:

[0012] The multi-source heterogeneous operation and maintenance datasets are cleaned, normalized, and feature fusion preprocessed to obtain preprocessed fused operation and maintenance data.

[0013] The preprocessed and integrated operation and maintenance data is stored in a time-series format and indexed to obtain structured operation and maintenance storage data.

[0014] Key feature retrieval is performed on structured operation and maintenance storage data to obtain core operation and maintenance feature data.

[0015] Perform time-series anomaly detection and topology correlation analysis on core operation and maintenance characteristic data, and output standardized operation and maintenance characteristic data.

[0016] According to the intelligent equipment operation and maintenance management method based on industrial big data provided by the present invention, the process of outputting standardized operation and maintenance characteristic data includes:

[0017] The LSTM-attention mechanism is used to perform time series modeling on core operation and maintenance feature data, identify abnormal feature points, and output time series anomaly detection results.

[0018] Perform topology spatial correlation analysis on the equipment to identify fault propagation paths between equipment and output the topology correlation analysis results.

[0019] The time-series anomaly detection results and topology correlation analysis results are fused in multiple dimensions and standardized in encoding to generate standardized operation and maintenance feature data.

[0020] According to the intelligent equipment operation and maintenance management method based on industrial big data provided by the present invention, the process of generating preliminary operation and maintenance decision instructions includes:

[0021] Based on a predefined operation and maintenance rule base, and matching standardized operation and maintenance feature data, the current health index of the equipment is calculated.

[0022] An initial set of maintenance actions is generated based on the health index.

[0023] A multi-objective optimization objective function is constructed, and the initial set of operation and maintenance actions is optimized and solved by improving the sparrow search algorithm to generate a Pareto optimal solution set.

[0024] Risk sensitivity screening and conflict resolution are performed on the Pareto optimal solution set, and preliminary operation and maintenance decision instructions are output.

[0025] According to the intelligent equipment operation and maintenance management method based on industrial big data provided by the present invention, the process of generating a Pareto optimal solution set includes:

[0026] Construct a multi-objective optimization objective function with the goals of minimizing total operation and maintenance costs, maximizing system availability, and minimizing failure risk.

[0027] Initialize the population size and number of iterations for the improved sparrow search algorithm.

[0028] An adaptive step size factor is introduced to optimize the discoverer-follower search mechanism, and a Pareto front candidate solution set is generated iteratively.

[0029] According to the intelligent equipment operation and maintenance management method based on industrial big data provided by this invention, the process of risk sensitivity screening of the Pareto optimal solution set includes:

[0030] Calculate the risk sensitivity index for each solution in the candidate solution set. The risk sensitivity index is defined as the rate of change of the failure risk value.

[0031] Set a preset threshold for risk sensitivity and remove solutions whose risk sensitivity index exceeds the preset threshold.

[0032] The retained solutions are encapsulated into instructions and conflict resolved to obtain preliminary operation and maintenance decision instructions.

[0033] According to the intelligent equipment operation and maintenance management method based on industrial big data provided by the present invention, the process of encapsulating instructions and resolving conflicts in the retained solution includes:

[0034] Establish a rule base for determining operational and maintenance decision conflicts, and identify conflict features of the screened candidate solutions from three dimensions: equipment operation and maintenance time conflicts, resource scheduling conflicts, and job permission conflicts.

[0035] A hierarchical priority ranking strategy is adopted to prioritize conflicting candidate decisions layer by layer according to the importance level of equipment, the urgency of the fault, and the constraints of operation and maintenance costs.

[0036] For high-conflict decision-making, local variable fine-tuning and objective function quadratic fitting are performed to resolve coupling conflicts between multiple decisions without deviating from the Pareto optimal frontier.

[0037] The optimal decision after conflict resolution is encapsulated in standardized fields, and the operation and maintenance command format, operation and maintenance object identifier and execution timing parameters are unified to generate a well-organized preliminary operation and maintenance decision command.

[0038] The intelligent equipment operation and maintenance management method based on industrial big data provided by the present invention includes the following process for dynamic optimization and updating in conjunction with collaborative constraints:

[0039] It interacts with external device health management systems to obtain health management collaborative constraint data.

[0040] Based on collaborative constraint data, the execution window and priority of preliminary operation and maintenance decision instructions are dynamically adjusted.

[0041] Generate the final operation and maintenance decision instructions and convert the instructions into a compatible communication format for the equipment operation and maintenance execution terminal.

[0042] The intelligent equipment operation and maintenance management method based on industrial big data provided by the present invention further includes, after issuing the final operation and maintenance decision instruction:

[0043] During the execution of final maintenance decision instructions, the remaining service life of the equipment is predicted based on the equipment's historical health data.

[0044] Based on the predicted remaining useful life, dynamically update the equipment maintenance plan and spare parts inventory strategy.

[0045] On the other hand, the present invention also provides an intelligent equipment operation and maintenance management system based on industrial big data, including:

[0046] The data acquisition module is used to collect real-time operating status parameters and environmental condition parameters of industrial equipment.

[0047] The data processing module is used to preprocess, store, and extract time-series topological features from the collected data, and output standardized operation and maintenance feature data.

[0048] The decision generation module is used to generate preliminary operation and maintenance decision instructions based on standardized operation and maintenance characteristic data.

[0049] The collaborative optimization module is used to dynamically optimize and update the initial operation and maintenance decision instructions in conjunction with the collaborative constraints of the external device health management system, and output the final operation and maintenance decision instructions.

[0050] The instruction execution module is used to send the final operation and maintenance decision instructions to the target device operation and maintenance execution terminal, and to monitor the execution status and provide feedback on the execution results.

[0051] The intelligent equipment operation and maintenance management method and system based on industrial big data provided by this invention can keenly capture subtle deviations in equipment operation by deeply integrating and storing multi-source heterogeneous data in a time sequence, and extracting features by combining deep learning models. This transforms post-event maintenance into pre-event early warning, improves the accuracy and lead time of fault diagnosis, and solves the problem of lagging traditional manual inspection.

[0052] By introducing topology correlation analysis, fault propagation paths are identified based on the physical and logical connections between devices. Combined with the collaborative constraints of the external health management system, dynamic optimization is performed to avoid systemic chain collapses caused by local faults and improve the overall stability of the production line.

[0053] By using multi-objective optimization algorithms to automatically generate Pareto optimal solution sets and introducing risk sensitivity screening and conflict resolution mechanisms, operation and maintenance decisions can achieve a scientific balance between cost, availability and risk. This avoids the waste of resources due to over-maintenance and prevents the hidden dangers caused by insufficient maintenance, thus realizing a leap from experience-driven to data-driven decision-making.

[0054] By dynamically updating maintenance plans and spare parts strategies based on the remaining life prediction model, maintenance activities become more predictable and economical, reducing inventory backlog and losses from sudden downtime, and bringing economic benefits and improved management efficiency to enterprises. Attached Figure Description

[0055] The invention will now be further described with reference to the accompanying drawings.

[0056] Figure 1 This is a flowchart illustrating the intelligent equipment operation and maintenance management method based on industrial big data in this invention.

[0057] Figure 2 This is a flowchart illustrating the process of generating preliminary operation and maintenance decision instructions in this invention;

[0058] Figure 3 This is a schematic diagram of the intelligent equipment operation and maintenance management system based on industrial big data in this invention. Detailed Implementation

[0059] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0060] like Figures 1 to 3 As shown, the intelligent equipment operation and maintenance management method and system based on industrial big data provided in this embodiment of the invention can be executed by an intelligent equipment operation and maintenance management method based on industrial big data, the method including:

[0061] Equipment operating status parameters are collected through terminals such as vibration sensors, temperature sensors, pressure transmitters, current transformers, and speed sensors deployed at key parts of the equipment. These parameters include bearing vibration amplitude and frequency, stator winding temperature, inlet and outlet medium pressure, motor three-phase current and voltage, spindle speed, and equipment operating time. Environmental operating parameters are collected through environmental monitoring terminals. These parameters include ambient temperature, ambient humidity, dust concentration, corrosive gas concentration, and environmental vibration.

[0062] An edge computing gateway is used as a data aggregation node. It connects with various data acquisition terminals through industrial communication protocols such as Modbus RTU, OPCUA, and MQTT to perform millisecond-level time synchronization of the acquired data, with the synchronization error controlled within 10ms.

[0063] The multi-source heterogeneous operation and maintenance dataset covers three categories: structured data (real-time sensor values, equipment operating parameters), semi-structured data (equipment operation logs, alarm records), and unstructured data (infrared thermal imaging images, equipment inspection videos). Taking the blast furnace blower unit of a steel plant as an example, a single device can generate about 200 different types of operation and maintenance data per second. The data acquisition frequency is dynamically adjusted according to the importance level of the equipment. The acquisition frequency for core equipment is 10Hz, for critical equipment it is 1Hz, and for general equipment it is 0.1Hz.

[0064] The multi-source heterogeneous operation and maintenance datasets are cleaned, normalized, and feature fusion preprocessed to obtain preprocessed fused operation and maintenance data.

[0065] The missing value handling adopts a hierarchical interpolation strategy. For short-term missing continuous operating parameters (missing duration less than 5 minutes), linear interpolation is used to fill in the missing data. For missing periodic operating parameters, the historical average of the same period is used to fill in the missing data. For data missing for a long time (missing duration greater than 30 minutes), the missing data is marked as invalid data and removed.

[0066] Outlier handling employs the 3σ principle combined with box plot method. First, outliers exceeding the normal distribution range are removed using the 3σ principle. Then, extreme outliers caused by sensor malfunctions are identified and removed using the box plot method. The outlier removal rate is controlled within 0.5% to avoid loss of valid data.

[0067] Duplicate value processing uses a unique identifier of "device ID + timestamp" to remove duplicates, ensuring that only one valid data record is retained for the same device at the same time.

[0068] The normalization process adopts a multidimensional adaptation strategy. For parameters with fixed value ranges (such as temperature 0-120℃ and pressure 0-1.6MPa), Min-Max normalization is used to map the data to the [0,1] interval. For parameters that follow a normal distribution (such as vibration acceleration and current fluctuation value), Z-score normalization is used to eliminate dimensional differences.

[0069] Feature fusion adopts a multimodal feature fusion method based on attention mechanism. The operating state features, environmental conditions features, and topological association features are respectively input into independent fully connected layers for feature mapping. The contribution weight of each modality feature is calculated through the attention layer, and the weighted fusion is used to obtain a global fusion feature with a dimension of 128. The temporal correlation and physical meaning of each feature are preserved during the fusion process.

[0070] The preprocessed and integrated operation and maintenance data is stored in a time-series format and indexed to obtain structured operation and maintenance storage data.

[0071] A hybrid storage architecture of "time-series database + relational database + distributed file system" is adopted. InfluxDB time-series database is used to store pre-processed fused operation and maintenance time-series data, MySQL relational database is used to store structured data such as basic device information, operation and maintenance rules, and decision records, and HDFS distributed file system is used to store unstructured data such as infrared thermal imaging images and inspection videos.

[0072] The data partitioning adopts a two-level partitioning strategy of "time partitioning + device hash partitioning". Time partitioning is performed by day, and hash partitioning is performed by device ID within each time partition. The data size of a single partition is controlled within 10GB to improve data read and write efficiency.

[0073] The index construction includes a timestamp primary key index, a device ID secondary index, and a feature dimension inverted index, supporting combined retrieval by time range, device number, and feature type, with retrieval response time controlled within 100ms.

[0074] A hot and cold data separation storage strategy is implemented. Hot data from the past three months is stored on SSD disks to meet real-time query requirements; cold data older than three months is automatically migrated to the HDFS distributed storage system, reducing storage costs by more than 60%; and cold data is loaded into memory on demand for offline analysis.

[0075] Key feature retrieval is performed on structured operation and maintenance storage data to obtain core operation and maintenance feature data.

[0076] First, based on the mutual information method and the random forest feature importance scoring method, the importance of all operation and maintenance features is quantitatively evaluated. The correlation score between each feature and equipment failure is calculated, and the top 20 core features with the highest correlation scores are selected. For example, the core features of the blast furnace blower unit include radial vibration amplitude, axial vibration amplitude, bearing housing temperature, motor stator temperature, outlet air pressure, and lubricating oil pressure.

[0077] An incremental retrieval mechanism is adopted, recording the timestamp of the last retrieval and only retrieving maintenance data added since the last retrieval, avoiding a full database scan and improving retrieval efficiency by over 80%. Data consistency checks are added during the retrieval process, comparing the continuity of timestamps with the uniqueness of device IDs, eliminating inconsistent data fragments, and ensuring the accuracy and completeness of core maintenance characteristic data.

[0078] For cross-device related feature retrieval, a pre-computed topological association index is used to directly obtain the corresponding feature data of the related devices without the need for cross-table joins.

[0079] Time-series anomaly detection and topology correlation analysis are performed on core operational and maintenance characteristic data to output standardized operational and maintenance characteristic data. Specifically, time-series anomaly detection is implemented using an LSTM-attention mechanism, and the attention weight calculation model is as follows:

[0080]

[0081]

[0082] In the formula, Indicates the first Attention weights corresponding to each temporal feature Indicates the first Attention score for each hidden state This represents the total time step of the time series feature, set to 60, corresponding to a 60-second historical data window. Represents the time step index of time series features. This indicates that the attention mechanism can learn the weight transpose vector. This represents the hidden state weight matrix. Indicates the LSTM network's... Hidden layer output state at time step This represents the attention mechanism bias term. This represents the hyperbolic tangent activation function. This represents the natural exponential function.

[0083] The process includes preprocessing, storing, and extracting time-series topology features from multi-source heterogeneous operation and maintenance datasets to output standardized operation and maintenance feature data.

[0084] The LSTM-attention mechanism is used to perform time series modeling on core operation and maintenance feature data, identify abnormal feature points, and output time series anomaly detection results.

[0085] The LSTM-attention mechanism model consists of two LSTM layers and one attention layer. The number of neurons in the hidden layer is set to 64. The input is a core feature sequence of length 60, and the output is the anomaly probability at each time step.

[0086] The model is trained using historical normal operating data of the equipment, with a training set to test set ratio of 8:2. Mean squared error is used as the loss function, and parameters are updated using the Adam optimizer with a learning rate of 0.001.

[0087] The detection process uses a sliding window method with a window size of 60 seconds and a step size of 10 seconds. The feature sequence within each window is predicted, and when the anomaly probability exceeds the adaptive threshold, it is determined to be an anomalous feature point.

[0088] The adaptive threshold is updated in real time using the kernel density estimation method. The threshold is calculated based on the normal data distribution of the most recent 7 days, and the false alarm rate is controlled within 5%.

[0089] The output of the time-series anomaly detection results includes information such as anomaly time, anomaly feature name, anomaly value, anomaly probability, and anomaly level.

[0090] Perform topology spatial correlation analysis on the equipment to identify fault propagation paths between equipment and output the topology correlation analysis results.

[0091] Based on the physical connection relationship of industrial equipment and the production process logic, a directed graph of equipment fault propagation is constructed. In the graph, nodes represent equipment, edges represent the fault propagation relationship between equipment, and the weight of the edges is obtained by statistical analysis of historical fault data, representing the probability of a fault propagating from upstream equipment to downstream equipment.

[0092] Taking a blast furnace blower unit as an example, the fault propagation directed graph includes nodes such as motor, coupling, gearbox, blower body, and bearing housing, with edge weights ranging from 0.1 to 0.9.

[0093] Based on the time-series anomaly detection results, a depth-first search algorithm is used to traverse the directed graph of fault propagation. Starting from the abnormal device node, all possible fault source nodes are traced, and the comprehensive propagation probability of each fault propagation path is calculated. The top 3 fault propagation paths with the highest probabilities are then output.

[0094] The output topology correlation analysis results include information such as the fault source device, fault propagation path, propagation probability, and a list of affected devices.

[0095] The time-series anomaly detection results and topology correlation analysis results are fused in multiple dimensions and standardized in encoding to generate standardized operation and maintenance feature data.

[0096] A feature splicing and fusion method is adopted to splice the numerical features of temporal anomaly detection with the semantic features of topological association analysis to form a unified feature vector.

[0097] Standardized encoding is achieved using JSON format, defining unified field specifications, including unique device identifier, anomaly occurrence time, anomaly feature set, fault propagation path, associated device information, data collection timestamp, etc. Each field has a clearly defined data type and value range to ensure that subsequent decision-making modules can resolve the data unambiguously.

[0098] The generation delay of standardized operation and maintenance characteristic data is controlled within 1 second to meet the needs of real-time operation and maintenance decision-making.

[0099] Based on standardized operation and maintenance characteristic data, preliminary operation and maintenance decision instructions are generated by combining a predefined operation and maintenance rule base with a multi-objective optimization algorithm. The process includes:

[0100] Based on a predefined operation and maintenance rule base, and matching standardized operation and maintenance feature data, the current health index of the equipment is calculated.

[0101] The predefined operation and maintenance rule base is built based on the ISO14224 equipment reliability and maintenance standard and the company's failure cases over the past 5 years. It uses production rules to represent rules, such as "IF bearing temperature > 90℃ AND vibration amplitude > 10mm / s THEN equipment health level is critical, and it is recommended to shut down and repair immediately".

[0102] The rule base consists of three levels: basic rules, industry-wide rules, and enterprise-customized rules, containing more than 1,200 rules. It supports the dynamic addition, modification, and deletion of rules, and the rule base is updated monthly based on new failure cases.

[0103] The rule matching uses a forward reasoning method, which matches standardized operation and maintenance feature data with the conditions in the rule base one by one, and triggers the rule with the highest matching degree.

[0104] An initial set of maintenance actions is generated based on the health index. A multi-objective optimization objective function is constructed, and the initial set of maintenance actions is optimized and solved using an improved sparrow search algorithm to generate a Pareto optimal solution set. The process includes:

[0105] Construct a multi-objective optimization objective function with the goals of minimizing total operation and maintenance costs, maximizing system availability, and minimizing failure risk.

[0106] Total maintenance cost includes labor cost, spare parts cost, and downtime loss cost; system availability is defined as the ratio of equipment uptime to total uptime; fault risk value is defined as the product of the probability of fault occurrence and the loss due to fault.

[0107] Initialize the population size and number of iterations for the improved sparrow search algorithm.

[0108] An adaptive step size factor is introduced to optimize the discoverer-follower search mechanism, and a Pareto front candidate solution set is generated iteratively.

[0109] The population size was set to 50, the maximum number of iterations was set to 100, and the discoverer ratio was set to 0.7.

[0110] The formula for calculating the adaptive step size factor is:

[0111]

[0112] in This is the initial step size factor, with a value of 0.5; This represents the current iteration number; This represents the maximum number of iterations. The coefficient of variation for population fitness is calculated based on the standard deviation of the current population fitness.

[0113] During the discoverer search phase, the step size gradually decreases with the number of iterations to improve the algorithm's local search capability. During the follower search phase, the step size is dynamically adjusted according to the population fitness dispersion. When the fitness dispersion is large, the step size is increased to expand the search range and avoid getting trapped in local optima. When the fitness dispersion is small, the step size is decreased to improve search accuracy.

[0114] The solutions generated iteratively are sorted using the non-dominated sorting method, and non-dominated solutions are retained to finally generate the Pareto optimal solution set.

[0115] The objective function vector for multi-objective optimization is represented as follows:

[0116]

[0117] In the formula, This represents the objective function vector for multi-objective optimization. This represents a vector of variables for optimizing operations and maintenance decisions, including operations and maintenance time, number of operations and maintenance personnel, number of spare parts used, and types of operations and maintenance. This represents the total cost of equipment operation and maintenance under the corresponding decision variable. This indicates the overall availability of the equipment operation and maintenance system under the corresponding decision variables. This represents the overall equipment failure risk value under the corresponding decision variable. This represents minimizing the optimization objective. This represents the feasible solution domain constraint space for the optimization variables, including maintenance time constraints (single maintenance time does not exceed 8 hours), maintenance personnel number constraints (no more than 5 people can participate in maintenance at the same time), spare parts inventory constraints (the required number of spare parts does not exceed the current inventory), and production plan constraints (maintenance time must not conflict with core production tasks), etc. This indicates the identifier for optimization constraints.

[0118] The Pareto optimal solution set is subjected to risk sensitivity screening and conflict resolution, and preliminary operation and maintenance decision instructions are output. The process includes:

[0119] Calculate the risk sensitivity index for each solution in the candidate solution set. The risk sensitivity index is defined as the rate of change of the failure risk value, and the calculation formula is as follows: ,in This represents the change in the fault risk value relative to the baseline solution. This represents the change in total operation and maintenance cost relative to the baseline solution.

[0120] Set a preset threshold for risk sensitivity and remove solutions with risk sensitivity indicators higher than the preset threshold. The preset threshold for risk sensitivity is set to 0.8. This threshold is obtained through statistical analysis of the enterprise's historical operation and maintenance data. Although solutions with operation and maintenance costs are lower than this threshold, their failure risk fluctuates greatly and their stability is poor, making them unsuitable as a basis for operation and maintenance decisions.

[0121] The retained solutions are encapsulated into instructions and conflict resolved to obtain preliminary operation and maintenance decision instructions.

[0122] The process of encapsulating instructions and resolving conflicts in the reserved solution includes:

[0123] Establish a rule base for determining operational and maintenance decision conflicts, and identify conflict features of the screened candidate solutions from three dimensions: equipment operation and maintenance time conflicts, resource scheduling conflicts, and job permission conflicts.

[0124] The rules for determining equipment maintenance time conflicts are: "If the same equipment ID and the maintenance time intervals overlap, then it is determined to be a time conflict"; the rules for determining resource scheduling conflicts are: "If the same maintenance resource ID (personnel, tools, vehicles) and the resource usage time intervals overlap, then it is determined to be a resource conflict"; and the rules for determining job permission conflicts are: "If the maintenance personnel's role permission level is less than the permission level required for equipment operation, then it is determined to be a permission conflict".

[0125] A hierarchical priority ranking strategy is adopted to prioritize conflicting candidate decisions layer by layer according to the importance level of equipment, the urgency of the fault, and the constraints of operation and maintenance costs.

[0126] Equipment importance is categorized into three levels: Level 1 (core equipment, failure will cause a complete production stoppage), Level 2 (critical equipment, failure will cause a partial production stoppage), and Level 3 (general equipment, failure does not affect production). Fault urgency is categorized into three levels: Level 1 (shutdown failure, equipment has stopped operating), Level 2 (performance degradation failure, equipment operating efficiency decreases by more than 30%), and Level 3 (potential failure, equipment parameters are abnormal but do not affect operation). Maintenance cost constraints are categorized into three levels: High (maintenance cost exceeds 100,000 RMB), Medium (maintenance cost 10,000-100,000 RMB), and Low (maintenance cost less than 10,000 RMB). Priority is assigned according to the principle of "equipment importance level > fault urgency level > maintenance cost constraint".

[0127] For high-conflict decision-making, local variable fine-tuning and objective function quadratic fitting are performed to resolve coupling conflicts between multiple decisions without deviating from the Pareto optimal frontier.

[0128] First, adjust the maintenance time window, postponing the maintenance time of low-priority tasks; second, adjust the maintenance personnel configuration, reassigning backup maintenance personnel with the same qualifications; finally, adjust the spare parts usage plan, adopting alternative spare parts or temporary repair solutions. By performing a quadratic fitting of the objective function, ensure that the adjusted decision remains near the Pareto optimal frontier, with the deviation of the objective function value controlled within 5%.

[0129] The optimal decision after conflict resolution is encapsulated in standardized fields, and the operation and maintenance command format, operation and maintenance object identifier and execution timing parameters are unified to generate a well-organized preliminary operation and maintenance decision command.

[0130] The standardized operation and maintenance instructions include 18 fields such as instruction ID, device ID, device name, operation and maintenance type, operation and maintenance time, list of operation and maintenance personnel, list of required spare parts, operation steps, safety precautions, and expected completion time. They are stored in XML format and are compatible with various operation and maintenance execution terminals.

[0131] The calculation model for the equipment health index is as follows:

[0132]

[0133] In the formula, Indicates the first Taiwanese industrial equipment at all times The device health index ranges from [0,1]. A value closer to 1 indicates a better device health. These represent the weight coefficients corresponding to the dimensions of operating status, environmental conditions, and topological association, respectively. They are determined using the Analytic Hierarchy Process (AHP) combined with the entropy weight method. The subjective weights for each dimension are determined using AHP, and the objective weights are determined using the entropy weight method. Finally, the subjective and objective weights are weighted and averaged to obtain the final combined weight. For example, the weight coefficients for the blast furnace blower unit are as follows: , , , Indicates time Normalized score of equipment operating status Indicates time Normalized score of equipment time-series operating conditions Indicates time Inter-device topology association failures affect the normalized score. This indicates the industrial equipment number index. Indicates the current moment.

[0134] By combining the collaborative constraints of the external equipment health management system, the preliminary operation and maintenance decision instructions are dynamically optimized and updated, and the final operation and maintenance decision instructions are output. The process includes:

[0135] It interacts with external device health management systems to obtain health management collaborative constraint data.

[0136] The OPCUA protocol is used to achieve bidirectional data interaction with the production scheduling system, ERP system, human resources system, and safety management system. Data transmission is encrypted using TLS 1.3 to ensure data security.

[0137] The acquired collaborative constraint data includes: equipment production plans and downtime windows from the production scheduling system, spare parts inventory status and procurement cycle from the ERP system, shift schedules and qualification information of maintenance personnel from the human resources system, and work permit approval status from the safety management system.

[0138] Based on collaborative constraint data, the execution window and priority of preliminary operation and maintenance decision instructions are dynamically adjusted.

[0139] If the production schedule shows that the equipment has a core production task during the execution period of the initial maintenance decision, the maintenance time will be adjusted to the nearest production gap; if spare parts inventory is insufficient and the procurement cycle exceeds 7 days, a temporary repair solution will be prioritized and a spare parts procurement request will be automatically generated; if the maintenance personnel have other tasks during the execution period, a backup maintenance personnel with the same qualifications will be reassigned; if the safety management system has not approved the corresponding work permit, the maintenance instruction will be marked as pending approval and will be issued after the permit is approved.

[0140] Generate the final operation and maintenance decision instructions and convert the instructions into a compatible communication format for the equipment operation and maintenance execution terminal.

[0141] Based on the communication protocols of different maintenance execution terminals, the standardized XML format instructions are converted into the corresponding protocol formats. For example, for PLC-controlled devices, they are converted into Modbus RTU protocol format; for handheld maintenance terminals, they are converted into HTTP protocol format; and for automated maintenance robots, they are converted into ROS protocol format.

[0142] The final operation and maintenance decision instructions are sent to the target equipment operation and maintenance execution terminal, and the execution status is monitored and the execution results are fed back.

[0143] A message queue mechanism is used to ensure reliable instruction delivery, guaranteeing that instructions are not lost or duplicated.

[0144] It receives real-time feedback information from the operation and maintenance execution terminal, including command reception status, execution progress, execution results, and abnormal situations, with the execution status updated once per minute. If an abnormality occurs during execution, the emergency response mechanism is immediately triggered, notifying operation and maintenance management personnel for manual intervention and automatically generating an exception handling plan.

[0145] During the execution of final maintenance decision instructions, the remaining service life of the equipment is predicted based on historical health data. The maintenance plan and spare parts inventory strategy are then dynamically updated according to the predicted remaining service life.

[0146] The prediction model for the remaining useful life is as follows:

[0147]

[0148] In the formula, Indicates the first The equipment is at all times The remaining service life, in hours. This indicates the rated full lifespan of the equipment as designed at the time of manufacture, provided by the equipment manufacturer. Indicates time The corresponding equipment performance degradation rate coefficient is calculated using the maximum likelihood estimation method based on historical equipment failure data and accelerated life test data. Different types of equipment have different degradation rate coefficients; for example, the degradation rate coefficient for rolling bearings is approximately [missing value]. , Indicates the first The equipment is at all times Health index, Represents the time variable of integration. This represents the natural exponential function.

[0149] The remaining lifespan prediction is updated hourly, and the prediction results are dynamically adjusted based on real-time collected health data, with the prediction error controlled within 10%.

[0150] The dynamic update method for the maintenance plan is as follows: when the predicted remaining service life is less than 30 days, a Level 1 preventive maintenance plan is generated and a comprehensive overhaul is scheduled for the nearest production downtime day; when the predicted remaining service life is between 30 and 90 days, a Level 2 preventive maintenance plan is generated and a key inspection is scheduled for monthly maintenance; when the predicted remaining service life is greater than 90 days, the original daily maintenance plan is maintained.

[0151] The spare parts inventory strategy adopts the dynamic safety stock method, and the formula for calculating the safety stock level is as follows: ,in This is the service level coefficient (usually taken as 1.65, corresponding to 95% service level). The standard deviation of the remaining useful life prediction error. This allows for lead time for spare parts procurement. Safety stock levels are dynamically adjusted based on changes in the remaining useful life prediction error. When the prediction error increases, safety stock is increased to avoid spare parts shortages; when the prediction error decreases, safety stock is decreased to reduce inventory capital tied up.

[0152] In summary, this embodiment provides an intelligent equipment operation and maintenance management method based on industrial big data. By deeply integrating and storing multi-source heterogeneous data in a time-series manner, and combining it with deep learning models for feature extraction, it can keenly capture subtle deviations in equipment operation, transforming post-event maintenance into pre-event early warning, improving the accuracy and lead time of fault diagnosis, and solving the problem of lagging traditional manual inspection.

[0153] By introducing topology correlation analysis, fault propagation paths are identified based on the physical and logical connections between devices. Combined with the collaborative constraints of the external health management system, dynamic optimization is performed to avoid systemic chain collapses caused by local faults and improve the overall stability of the production line.

[0154] By using multi-objective optimization algorithms to automatically generate Pareto optimal solution sets and introducing risk sensitivity screening and conflict resolution mechanisms, operation and maintenance decisions can achieve a scientific balance between cost, availability and risk. This avoids the waste of resources due to over-maintenance and prevents the hidden dangers caused by insufficient maintenance, thus realizing a leap from experience-driven to data-driven decision-making.

[0155] By dynamically updating maintenance plans and spare parts strategies based on the remaining life prediction model, maintenance activities become more predictable and economical, reducing inventory backlog and losses from sudden downtime, and bringing economic benefits and improved management efficiency to enterprises.

[0156] Based on the same general inventive concept, this invention also protects an intelligent equipment operation and maintenance management system based on industrial big data. The intelligent equipment operation and maintenance management system based on industrial big data provided by this invention will be described below. The intelligent equipment operation and maintenance management system based on industrial big data described below can be referred to in correspondence with the intelligent equipment operation and maintenance management method and system based on industrial big data described above.

[0157] The intelligent equipment operation and maintenance management system based on industrial big data includes a data acquisition module, a data processing module, a decision generation module, a collaborative optimization module, and an instruction execution module.

[0158] The data acquisition module is used to collect real-time operating status parameters and environmental condition parameters of industrial equipment;

[0159] The data processing module is used to preprocess, store, and extract time-series topological features from the collected data, and output standardized operation and maintenance feature data.

[0160] The decision generation module is used to generate preliminary operation and maintenance decision instructions based on the standardized operation and maintenance characteristic data;

[0161] The collaborative optimization module is used to combine the collaborative constraints of the external device health management system to dynamically optimize and update the preliminary operation and maintenance decision instructions, and output the final operation and maintenance decision instructions.

[0162] The instruction execution module is used to send the final operation and maintenance decision instruction to the target device operation and maintenance execution terminal, and to monitor the execution status and provide feedback on the execution results.

[0163] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent device operation and maintenance management method based on industrial big data, characterized in that, include: Collect real-time operating status parameters and environmental condition parameters of industrial equipment to construct a multi-source heterogeneous operation and maintenance dataset; The multi-source heterogeneous operation and maintenance dataset is preprocessed, stored, and time-series topology features are extracted to output standardized operation and maintenance feature data. Based on the standardized operation and maintenance characteristic data, preliminary operation and maintenance decision instructions are generated by combining a predefined operation and maintenance rule base with a multi-objective optimization algorithm. By combining the collaborative constraints of the external device health management system, the preliminary operation and maintenance decision instructions are dynamically optimized and updated, and the final operation and maintenance decision instructions are output. The final operation and maintenance decision instruction is sent to the target device operation and maintenance execution terminal, and the execution status is monitored and the execution results are fed back.

2. The intelligent equipment operation and maintenance management method based on industrial big data according to claim 1, characterized in that, The process of preprocessing, storing, and extracting time-series topological features from multi-source heterogeneous operation and maintenance datasets includes: The multi-source heterogeneous operation and maintenance dataset is cleaned, normalized, and preprocessed with feature fusion to obtain preprocessed fused operation and maintenance data. The preprocessed and fused operation and maintenance data is then stored in a time-series format and indexed to obtain structured operation and maintenance storage data. Key feature retrieval is performed on the structured operation and maintenance storage data to obtain core operation and maintenance feature data; Perform time-series anomaly detection and topology correlation analysis on the core operation and maintenance characteristic data, and output standardized operation and maintenance characteristic data. 3.The industrial big data-based intelligent device operation and maintenance management method of claim 2, wherein, The process of outputting standardized operation and maintenance characteristic data includes: The LSTM-attention mechanism is used to perform time series modeling on core operation and maintenance feature data, identify abnormal feature points, and output time series anomaly detection results. Perform topology spatial correlation analysis on the equipment to identify fault propagation paths between equipment and output the topology correlation analysis results; The time-series anomaly detection results and the topology correlation analysis results are fused in multiple dimensions and standardized encoding to generate the standardized operation and maintenance feature data. 4.The industrial big data-based intelligent device operation and maintenance management method of claim 1, wherein, The process of generating preliminary operation and maintenance decision instructions includes: Based on a predefined operation and maintenance rule base, the standardized operation and maintenance feature data is matched to calculate the current health index of the equipment. An initial set of maintenance actions is generated based on the health index. A multi-objective optimization objective function is constructed, and the initial set of operation and maintenance actions is optimized and solved by an improved sparrow search algorithm to generate a Pareto optimal solution set; Risk sensitivity screening and conflict resolution are performed on the Pareto optimal solution set, and the preliminary operation and maintenance decision instructions are output. 5.The industrial big data-based intelligent device operation and maintenance management method of claim 4, wherein, The process of generating the Pareto optimal solution set includes: Construct a multi-objective optimization objective function that aims to minimize total operating costs, maximize system availability, and minimize failure risk. Initialize the population size and number of iterations for the improved sparrow search algorithm; An adaptive step size factor is introduced to optimize the discoverer-follower search mechanism, and a Pareto front candidate solution set is generated iteratively. 6.The industrial big data-based intelligent device operation and maintenance management method of claim 5, wherein, The process of risk sensitivity screening for Pareto optimal solutions includes: Calculate the risk sensitivity index for each solution in the candidate solution set, whereby the risk sensitivity index is defined as the rate of change of the fault risk value; Set a preset threshold for risk sensitivity, and remove solutions whose risk sensitivity index is higher than the preset threshold; The retained solution is encapsulated into instructions and conflict resolved to obtain the preliminary operation and maintenance decision instructions. 7.The industrial big data-based intelligent device operation and maintenance management method of claim 6, wherein, The process of encapsulating instructions and resolving conflicts in the retained solution includes: Establish a rule base for determining operational and maintenance decision conflicts, and identify conflict features of the screened candidate solutions from three dimensions: equipment operation and maintenance time conflicts, resource scheduling conflicts, and job permission conflicts. A hierarchical priority ranking strategy is adopted to prioritize conflicting candidate decisions layer by layer according to the importance level of equipment, the urgency of the fault, and the constraints of operation and maintenance costs. For high-conflict decision-making, local variable fine-tuning and objective function quadratic fitting are performed to resolve the coupling conflict between multiple decisions without deviating from the Pareto optimal frontier. The optimal decision after conflict resolution is encapsulated in standardized fields, and the operation and maintenance command format, operation and maintenance object identifier and execution timing parameters are unified to generate a well-organized preliminary operation and maintenance decision command. 8.The industrial big data-based intelligent device operation and maintenance management method of claim 1, wherein, The process of dynamically optimizing and updating by combining collaborative constraints includes: It interacts with the health management system of external devices to obtain health management collaborative constraint data; Based on the collaborative constraint data, the execution window and priority of the preliminary operation and maintenance decision instructions are dynamically adjusted; The final operation and maintenance decision instruction is generated, and the instruction is converted into a compatible communication format of the equipment operation and maintenance execution terminal. 9.The industrial big data-based intelligent device operation and maintenance management method of claim 1, wherein, Following the issuance of the final operation and maintenance decision instruction, the following is also included: During the execution of the final maintenance decision instruction, the remaining service life of the equipment is predicted based on the equipment's historical health data. Based on the predicted remaining useful life, dynamically update the equipment maintenance plan and spare parts inventory strategy.

10. An intelligent device operation and maintenance management system based on industrial big data, the intelligent device operation and maintenance management system adopting any one of the intelligent device operation and maintenance management methods of claims 1-9, characterized in that, The intelligent device operation and maintenance management system includes: The data acquisition module is used to collect real-time operating status parameters and environmental condition parameters of industrial equipment; The data processing module is used to preprocess, store, and extract time-series topological features from the collected data, and output standardized operation and maintenance feature data. The decision generation module is used to generate preliminary operation and maintenance decision instructions based on the standardized operation and maintenance characteristic data; The collaborative optimization module is used to dynamically optimize and update the preliminary operation and maintenance decision instructions in conjunction with the collaborative constraints of the external device health management system, and output the final operation and maintenance decision instructions. The instruction execution module is used to send the final operation and maintenance decision instruction to the target device operation and maintenance execution terminal, and to monitor the execution status and provide feedback on the execution results.

Images