A man-machine-environment collaborative intelligent operation and maintenance method and system, electronic equipment and storage medium

Abstract

The present application relates to a kind of man-machine-environment collaborative intelligent operation and maintenance method, system, electronic equipment and storage medium, solve the problem that prior art cannot simultaneously consider equipment fault diagnosis, personnel workload assessment and regional risk control.The method comprises: three-source data collaborative acquisition module synchronously acquires equipment operation data, personnel position data and regional environment data;Predictive maintenance model identifies fault equipment and generates fault diagnosis result;Spacetime intelligent model obtains real-time position and assesses workload state;Safety risk assessment model generates regional risk level;Based on fault diagnosis result, real-time position, workload state, regional risk level four-dimensional collaborative scheduling is carried out, and intelligent scheduling scheme is generated and pushed.The present application realizes equipment, personnel, environment collaborative perception and intelligent scheduling, significantly improves operation and maintenance efficiency and personnel safety.

Description

Technical Field

[0001] This invention relates to the field of industrial intelligent operation and maintenance technology, specifically to a human-machine-environment collaborative intelligent operation and maintenance method, system, electronic device and storage medium. Background Technology

[0002] Chemical enterprises are a crucial foundational industry of the national economy. Their production processes involve numerous complex equipment and hazardous media such as high-temperature, high-pressure, flammable, and explosive substances. High equipment density and stringent safety requirements are typical characteristics. With the continuous expansion of chemical enterprises and the improvement of their intelligent operation levels, traditional equipment operation and maintenance models are no longer sufficient to meet the demands of modern chemical production. Currently, the application level of information and digital technologies in equipment operation and maintenance management within the chemical industry varies considerably. Most enterprises still adopt a reactive, reactive maintenance model, where repairs are only carried out after equipment failure. This model suffers from significant problems such as delayed response and low maintenance efficiency.

[0003] The existing technology for the operation and maintenance management of chemical equipment mainly has the following problems: First, after equipment failure occurs, it is impossible to quickly locate the nearest maintenance personnel. Management personnel usually need to contact them one by one by phone or walkie-talkie, resulting in long response times and low efficiency. For example, although the existing equipment maintenance system WO2020193330 (British Telecom) can predict equipment maintenance, it is limited to equipment-level failure prediction and does not involve intelligent scheduling of maintenance personnel, thus failing to achieve coordinated linkage between failure response and personnel scheduling.

[0004] Secondly, the scheduling of maintenance personnel mainly relies on the experience and judgment of managers, lacking intelligent analysis of personnel workload, which easily leads to problems such as skill mismatch and uneven workload. Although the existing equipment fault detection system US20200057691 (IBM) can detect equipment faults, it only focuses on the equipment status and does not consider the workload status of maintenance personnel. This may result in maintenance tasks being assigned to fatigued maintenance personnel, affecting maintenance quality and safety.

[0005] Third, the equipment status monitoring system and the personnel location management system operate independently and cannot achieve coordinated operation, resulting in a lack of comprehensive information support for scheduling decisions. Although the existing Boeing patent CN105791379 can achieve personnel location, it is limited to personnel tracking and does not have the function of being combined with equipment operation and maintenance scheduling.

[0006] Fourth, the lack of a dynamic assessment mechanism for regional safety risks means that maintenance personnel may be dispatched to high-risk areas when equipment malfunctions occur, posing a safety hazard. Although Intel patent US10966059 involves location tracking technology, it has not been applied to operation and maintenance scheduling scenarios and cannot meet the safety assurance needs of chemical enterprises.

[0007] Fifth, although there are solutions involving risk assessment and equipment scheduling in the existing technology (such as CN117436651), these solutions lack assessment of personnel workload and cannot avoid assigning tasks to maintenance personnel who are already under heavy workload, resulting in excessive fatigue of personnel and affecting maintenance quality and safety.

[0008] In view of the above, this application is hereby submitted. Summary of the Invention

[0009] In a first aspect, the present invention provides a human-machine-environment collaborative intelligent operation and maintenance method, comprising: synchronously collecting equipment operation data, personnel location data, and regional environmental data through a three-source data collaborative acquisition module; analyzing the equipment operation data based on a predictive maintenance model to identify faulty equipment and generate fault diagnosis results; processing the personnel location data based on a spatiotemporal intelligent model to obtain the real-time location of personnel and assess their workload status; analyzing the regional environmental data based on a safety risk assessment model to generate a regional risk level; performing four-dimensional collaborative scheduling based on the fault diagnosis results, the real-time location of personnel, the workload status of personnel, and the regional risk level to generate an intelligent scheduling scheme; and pushing the intelligent scheduling scheme to the mobile terminal of the target maintenance personnel.

[0010] Preferably, the intelligent scheduling scheme includes: determining the fault type and urgency level based on the fault diagnosis results; matching candidate maintenance personnel with corresponding skills based on the fault type; calculating the spatial distance between each candidate maintenance personnel and the faulty equipment; obtaining the workload score of each candidate maintenance personnel; calculating the regional risk comprehensive score of the path for each candidate maintenance personnel to reach the faulty equipment; and performing weighted fusion based on the spatial distance, the workload score, the regional risk comprehensive score, and the urgency level to determine the optimal scheduling object and the optimal scheduling path.

[0011] Preferably, the method for calculating the workload score includes: acquiring personnel movement trajectory data within a preset time window; extracting personnel behavior feature vectors from the movement trajectory data based on a spatiotemporal behavior feature extraction model; and calculating the personnel workload score by combining the personnel's current task status and the behavior feature vectors.

[0012] Preferably, the method for calculating the regional risk comprehensive score includes: obtaining the historical risk level and real-time environmental parameters of each region traversed by the scheduling path; calculating the real-time risk value of each region based on the real-time environmental parameters and the historical risk level; and generating a regional risk comprehensive score based on the real-time risk value and the regional population density.

[0013] Preferably, the predictive maintenance model is generated based on multi-source operating data of the equipment, including: collecting vibration signals, temperature parameters and pressure parameters of the equipment; extracting time-domain features and frequency-domain features of the vibration signals; performing pattern matching on the time-domain features and frequency-domain features based on a fault diagnosis knowledge base to identify fault types; and generating fault diagnosis conclusions and handling suggestions.

[0014] Preferably, the spatiotemporal intelligent model is trained and generated based on historical trajectory data of personnel, and is used to extract personnel behavior patterns from real-time location data and assess workload status.

[0015] Preferably, the method further includes a self-evolution mechanism: receiving feedback information from maintenance personnel regarding the fault diagnosis conclusion; updating the fault diagnosis knowledge base based on the feedback information; and triggering incremental learning of the predictive maintenance model.

[0016] A second aspect of the present invention provides a human-machine-environment collaborative intelligent operation and maintenance system, comprising: a three-source data collaborative acquisition module for synchronously acquiring equipment operation data, personnel location data, and regional environmental data; a predictive maintenance module for analyzing the equipment operation data based on a predictive maintenance model, identifying faulty equipment, and generating fault diagnosis results; a spatiotemporal intelligence module for processing the personnel location data based on a spatiotemporal intelligence model, obtaining the real-time location of personnel, and assessing the personnel's workload status; a safety risk assessment module for analyzing the regional environmental data based on a safety risk assessment model and generating a regional risk level; a scheduling decision module for performing four-dimensional collaborative scheduling based on the fault diagnosis results, the real-time location of personnel, the personnel's workload status, and the regional risk level, generating an intelligent scheduling scheme; and a push module for pushing the intelligent scheduling scheme to the mobile terminal of the target maintenance personnel.

[0017] In summary, this application has the following beneficial effects: By simultaneously acquiring equipment operation data, personnel location data, and regional environmental data through three-source data collaborative acquisition technology, this invention achieves multi-dimensional collaborative perception of equipment status, personnel status, and regional environment. This breaks down the information silos of existing technologies where each system operates independently, providing comprehensive and real-time data support for intelligent scheduling. Compared to WO2020193330, which only addresses equipment maintenance prediction, and US20200057691, which only addresses equipment fault detection, this invention achieves the collaborative fusion of three-source data, laying a data foundation for intelligent scheduling.

[0018] By processing personnel location data through a spatiotemporal intelligent model, the system accurately assesses personnel workload status, providing personnel work status information for scheduling decisions. Compared with existing technologies (including CN117436651), this invention is the first to introduce personnel workload assessment into the field of operation and maintenance scheduling, effectively avoiding the assignment of maintenance tasks to already fatigued maintenance personnel, preventing personnel over-fatigue, and improving maintenance quality and safety.

[0019] This invention dynamically assesses the risk level of each region using a safety risk assessment model, comprehensively considering regional risk factors in scheduling decisions to achieve dynamic assessment and real-time monitoring of regional risks. Compared with existing technologies, this invention not only focuses on equipment failures but also on the safety of maintenance routes, effectively ensuring the personal safety of maintenance personnel.

[0020] By employing a four-dimensional collaborative scheduling mechanism, scheduling decisions are made by comprehensively considering four dimensions: fault type, personnel location, workload, and regional risk, thus achieving scientific and rational scheduling of maintenance personnel. Compared with existing technologies, the scheduling dimensions of this invention are more comprehensive, maximizing scheduling efficiency while ensuring safety.

[0021] By establishing a closed-loop system of diagnosis, scheduling, and feedback through a self-evolution mechanism, the system can continuously learn and evolve from actual operation, thereby continuously improving the accuracy of fault diagnosis and the rationality of scheduling decisions, and achieving a continuous improvement in the level of intelligence. Attached Figure Description

[0022] Figure 1 This is the overall system architecture diagram.

[0023] Figure 2 It is a flowchart of the method.

[0024] Figure 3 This is a flowchart of the four-dimensional collaborative scheduling process.

[0025] Figure 4 This is a flowchart for workload assessment.

[0026] Figure 5 This is a flowchart of the regional risk assessment process.

[0027] Explanation of reference numerals in the attached figures: 10-Three-source data collaborative acquisition module; 20-Predictive maintenance module; 30-Spatiotemporal intelligence module; 40-Safety risk assessment module; 50-Four-dimensional collaborative scheduling module; 60-Push module; 70-Self-evolution module; 101-Equipment operation data acquisition unit; 102-Personnel location data acquisition unit; 103-Regional environmental data acquisition unit; 201-Vibration sensor; 202-Temperature sensor; 203-Pressure sensor; 301-Location base station; 302-Location tag; 401-Environmental monitoring sensor. Detailed Implementation

[0028] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0029] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms "a," "the," and "the" as used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0030] To better understand the above technical solution, the following will provide a detailed explanation of the above technical solution in conjunction with specific implementation methods.

[0031] In some embodiments of this application, a human-machine-environment collaborative intelligent operation and maintenance method is provided. This method is an intelligent solution designed for equipment operation and maintenance scenarios in chemical enterprises. Through technologies such as collaborative acquisition of three-source data, intelligent fault diagnosis, workload assessment, dynamic risk assessment, and four-dimensional collaborative scheduling, it achieves rapid response to equipment failures and scientific and reasonable scheduling of maintenance personnel, effectively improving operation and maintenance efficiency and ensuring the safety of maintenance personnel.

[0032] The three-source data collaborative acquisition module is the basic data source of this invention, responsible for synchronously acquiring three types of key data: The first category of data is equipment operation data, including but not limited to various parameters closely related to the equipment's operating status, such as operating current, operating voltage, vibration signals, temperature parameters, pressure parameters, and flow parameters. Equipment operation data is collected through various sensors installed on the equipment, including vibration sensors, temperature sensors, pressure sensors, current sensors, and voltage sensors. The data acquisition frequency is differentiated according to the equipment type and its importance; a high-frequency acquisition mode is used for critical equipment to ensure that even minute changes in the equipment's state can be captured.

[0033] The second type of data is personnel location data, including the real-time location, historical trajectory, and area of ​​maintenance personnel. Personnel location data can be collected through various technologies, including but not limited to Bluetooth beacon positioning, WiFi positioning, UWB ultra-wideband positioning, RFID positioning, GPS positioning, and video analytics positioning. In chemical plant environments, due to the presence of numerous metal devices and complex electromagnetic environments, the positioning accuracy of a single positioning technology may be affected. Therefore, multi-source fusion positioning technology can be used to combine the advantages of multiple positioning technologies to improve positioning accuracy and reliability. Personnel location data includes not only the two-dimensional or three-dimensional coordinates of personnel but also information about the area where the personnel are located, and the time information for entering and leaving each area.

[0034] The third category of data is regional environmental data, which includes environmental parameters such as temperature, humidity, combustible gas concentration, toxic gas concentration, dust concentration, and noise level within the region, as well as information on the distribution of hazardous sources, personnel density, and equipment density. The collection of regional environmental data relies on various environmental monitoring sensors deployed in different areas, including gas detectors, temperature and humidity sensors, dust sensors, and noise monitors. These sensors monitor regional environmental parameters in real time and transmit the data to a central data processing system.

[0035] In practical implementation, the three-source data collaborative acquisition module adopts a unified data format and clock synchronization mechanism to ensure the consistency of the three types of data in the time dimension. The data acquisition module supports both wired and wireless transmission methods. Wired transmission methods include industrial Ethernet and fieldbus, suitable for fixed-installation equipment and scenarios with high real-time requirements. Wireless transmission methods include WiFi, LoRa, and NB-IoT, suitable for scenarios where wiring is difficult or flexible deployment is required. The acquired data is stored in a time-series database and a spatial database for easy historical querying and analysis.

[0036] By employing the above technical solution, comprehensive acquisition of equipment operating status, personnel location trajectories, and regional environmental information is achieved through three-source data collaborative acquisition technology, realizing multi-dimensional data collaborative perception. Compared with existing technologies, this invention breaks the information silo pattern of independent operation of equipment monitoring systems, personnel positioning systems, and environmental monitoring systems, providing comprehensive and real-time data support for subsequent intelligent scheduling. In particular, compared with existing technologies such as WO2020193330 and US20200057691 that only focus on a single data source, the three-source data collaborative acquisition of this invention has significant technical advantages.

[0037] In some embodiments of this application, predictive maintenance models are used to analyze equipment operation data, identify faulty equipment, and generate fault diagnosis results. The predictive maintenance model is one of the core technical components of this invention. It is built based on historical equipment operation data and machine learning algorithms, and can extract features reflecting the health status of the equipment from real-time collected equipment operation data. Based on these features, it can determine whether the equipment has malfunctioned and, if so, the type and severity of the malfunction.

[0038] The process of building a predictive maintenance model includes the following steps: The first step is to collect historical operational data of the equipment as training samples. This historical operational data includes data from normal operation and data from various fault occurrences. Normal operation data is used to learn the normal behavior patterns of the equipment, while fault data is used to learn the characteristic patterns of various faults. The quality and completeness of the data directly affect the model's predictive performance; therefore, a strict data quality control mechanism is necessary.

[0039] The second step is to extract features reflecting the equipment's health status from the collected historical data. Feature extraction includes two aspects: time-domain feature extraction and frequency-domain feature extraction. Time-domain features are directly calculated from the time-domain signal and include mean, variance, standard deviation, peak value, peak-to-peak value, root mean square value, waveform factor, impulse factor, peak factor, kurtosis, etc. These features can reflect the statistical characteristics and distribution features of the signal. Frequency-domain features are extracted after converting the time-domain signal to the frequency domain using Fourier transform. These include the dominant frequency, frequency energy distribution, spectral peaks and their corresponding frequencies, sideband features, etc. Frequency-domain features can reflect the vibration sources and fault characteristics of the equipment.

[0040] The third step is to train a machine learning model using the extracted feature data. The model type can include support vector machines, random forests, neural networks, deep learning, etc. This invention does not specifically limit the model type, as long as it can achieve fault mode recognition and classification. The model training employs a supervised learning method, using the extracted feature vectors as input and the fault type as the label, to acquire fault mode recognition capabilities through learning.

[0041] The fourth step is to deploy the trained predictive maintenance model to the production environment for online analysis and fault diagnosis of the real-time collected equipment operation data. The model deployment adopts a hybrid architecture combining edge computing and cloud computing. Edge computing nodes are responsible for real-time data processing and fault early warning, while the cloud computing platform is responsible for model training and complex diagnostic calculations.

[0042] In its implementation, the predictive maintenance model's workflow includes the following steps: receiving equipment operation data in real time, preprocessing the raw data to extract features; inputting the extracted features into a trained fault classification model to output the fault type and confidence level; querying the fault diagnosis knowledge base based on the fault type to obtain typical manifestations, possible causes, and recommended handling suggestions for that type of fault; and generating a fault diagnosis conclusion based on the above information, including the faulty equipment name, fault occurrence time, fault type, fault severity, fault cause analysis, and handling suggestions.

[0043] The fault diagnosis knowledge base is a digital representation of domain expert experience, containing common fault types, fault characteristic patterns, fault causes, and maintenance and handling methods for various types of equipment. The knowledge base can be represented as a rule base, a case base, or a hybrid of both. The rule base uses if-then rules to represent expert knowledge, such as "If the vibration spectrum shows a peak at twice the rotational speed frequency and the vibration amplitude exceeds the threshold, then it is determined to be an imbalance fault." The case base uses a set of historical fault cases to represent expert knowledge; new faults are diagnosed by matching their similarity to historical cases. The knowledge base supports dynamic updates and expansion, continuously accumulating new fault cases and maintenance experience.

[0044] By employing the above technical solution, a predictive maintenance model enables intelligent identification and diagnosis of equipment faults. It automatically detects equipment anomalies and determines fault types without manual intervention, significantly improving the timeliness of fault detection and the accuracy of diagnosis. Compared to the prior art US20200057691, the predictive maintenance model of this invention incorporates a fault diagnosis knowledge base, enabling it to generate complete diagnostic conclusions including handling suggestions, providing a more comprehensive decision-making basis for subsequent maintenance scheduling.

[0045] In other implementations, fault diagnosis may employ one or more combinations of the following methods: (1) Intelligent fault identification based on predictive maintenance model. As mentioned above, machine learning model is used to analyze equipment operation data and automatically identify fault types.

[0046] (2) Pattern matching based on fault diagnosis knowledge base. The extracted equipment features are matched with the fault patterns stored in the knowledge base to identify the most similar fault types.

[0047] (3) Fault reasoning based on expert system rules. The knowledge of domain experts is represented as reasoning rules, and fault diagnosis reasoning is performed through a rule engine.

[0048] (4) Fault judgment based on equipment operating parameter thresholds. When the equipment operating parameters exceed the preset normal range thresholds, the equipment is judged to be abnormal.

[0049] The above-mentioned diagnostic methods can be used individually or in combination to improve the accuracy and reliability of fault diagnosis.

[0050] In some embodiments of this application, personnel location data is processed based on a spatiotemporal intelligent model to obtain real-time personnel locations and assess their workload status. The spatiotemporal intelligent model is another core technical component of this invention. It is constructed based on historical personnel trajectory data and deep learning algorithms, enabling it to learn behavioral patterns and activity patterns from personnel location trajectories and thereby assess their current workload status.

[0051] Personnel workload assessment is one of the key innovations that distinguishes this invention from existing technologies. Existing operation and maintenance scheduling systems (including CN117436651) primarily focus on equipment status and fault type, neglecting the workload status of maintenance personnel. This may lead to the assignment of maintenance tasks to fatigued personnel, affecting maintenance quality and safety. This invention introduces personnel workload assessment into the field of operation and maintenance scheduling for the first time, accurately assessing the current workload status of each maintenance worker through a spatiotemporal intelligent model, providing important reference for scheduling decisions.

[0052] The workflow of a spatiotemporal intelligent model includes the following steps: First, acquire the movement trajectory data of personnel within a preset time window. The length of the time window can be set according to actual management needs, preferably 2 to 8 hours. A time window that is too short may not accurately reflect the workload of personnel, while a time window that is too long may ignore changes in the current work status of personnel. The movement trajectory data includes all the locations the personnel pass through within this time window, the dwell time at each location, and the movement speed.

[0053] Then, a spatiotemporal behavioral feature extraction model is used to extract personnel behavioral feature vectors from the movement trajectory data. This model is constructed using deep learning algorithms, taking basic personnel trajectory data as input and outputting high-dimensional feature vectors reflecting personnel workload and behavioral characteristics. The model architecture can employ a hybrid structure combining convolutional neural networks (CNNs) and long short-term memory (LSTM) networks. The CNNs extract features from the spatial dimension of the trajectory data, while the LSTM networks extract features from the temporal dimension. The model is trained using a contrastive learning method, which brings trajectory samples with similar workload states closer together in the feature space and widens the distance between different samples, enabling the model to learn a more discriminative representation of workload features. The training data comes from historical personnel behavior data, and the labeled data includes subjective evaluations of personnel workload states by experts.

[0054] Finally, combining the personnel's current task status and extracted behavioral feature vectors, a workload score is calculated. The workload score is typically represented by a value from 0 to 100, with a higher score indicating a greater workload. The calculation of the workload score needs to comprehensively consider the following factors: basic indicators such as the personnel's movement distance, number of stops, and work duration within the time window; whether the personnel are currently performing a maintenance task, the estimated completion time of the task, and the complexity and urgency of the task. If a personnel is performing an urgent maintenance task, even if their workload index is not high, their current workload score should be set to a higher level, indicating that they cannot accept new assignments at this time. If a personnel is currently idle or about to complete their current task, their workload score can be determined based on their workload index.

[0055] In practical implementation, the spatiotemporal intelligent model also needs to consider the historical workload patterns of personnel. Different personnel have varying work efficiency and processing capabilities; some can maintain good performance under high workloads, while others may experience fatigue under lower workloads. Therefore, the calculation of workload scores needs to be personalized by incorporating historical data, establishing a workload baseline and threshold for each maintenance worker to improve the accuracy of the assessment.

[0056] By employing the above technical solution, and through comprehensive analysis of personnel location trajectories and behavioral patterns using a spatiotemporal intelligent model, the workload status of personnel is accurately assessed, providing personnel work status information for intelligent scheduling. Compared with existing technologies, this invention introduces personnel workload assessment into the field of operation and maintenance scheduling for the first time, effectively avoiding the assignment of maintenance tasks to maintenance personnel with excessive workloads, preventing personnel fatigue, and improving the quality and safety of maintenance work. This is a significant technical advantage of this invention compared to existing technologies such as CN117436651.

[0057] In some embodiments of this application, regional environmental data is analyzed based on a safety risk assessment model to generate regional risk levels. Chemical plant production areas typically contain multiple hazards, including high-temperature and high-pressure equipment, flammable and explosive materials, toxic and harmful substances, and corrosive media. The safety risk levels vary significantly across different areas. Dynamic assessment of regional risk levels is another core innovation of this invention, providing safety risk information for scheduling decisions and effectively protecting the personal safety of maintenance personnel.

[0058] The input data for the safety risk assessment model includes the following categories: regional environmental parameters, including real-time monitoring data such as temperature, humidity, combustible gas concentration, toxic gas concentration, dust concentration, and noise level; hazard source distribution data, including the type, quantity, location, and risk level of various hazardous equipment in the region; personnel density data, including the current number and distribution of personnel in the region; and historical accident data, including records of safety accidents that have occurred in the region in the past and lessons learned.

[0059] The safety risk assessment model employs a multi-factor comprehensive evaluation method, combining expert knowledge and machine learning algorithms to determine the weight coefficients of each factor, and comprehensively quantifies and scores the regional safety risk. The model architecture can employ a weighted scoring model, an analytic hierarchy process (AHP) model, or a machine learning model, etc.; this invention does not specifically limit the model type. The scoring results are divided into different risk levels according to preset threshold ranges, typically categorized into four levels: low risk, medium risk, high risk, and extremely high risk. Different levels correspond to different control requirements and access conditions.

[0060] In practice, when regional environmental parameters change, such as the detection of increased combustible gas concentration or toxic gas leaks, the system automatically triggers a recalculation of the risk assessment model, updating the risk level of the area. Changes in risk levels are time-sensitive; after the expiration period, the system automatically reverts to the default risk level or reassesses. Furthermore, the system supports risk prediction, forecasting the evolution of regional risks based on historical data and current trends, providing forward-looking safety information for dispatching decisions.

[0061] By employing the above technical solution, a safety risk assessment model is used to dynamically assess and monitor the safety risks in various areas of a chemical enterprise in real time, enabling the timely detection of potential safety hazards. Compared with existing technologies, the regional risk assessment of this invention is real-time and dynamic, capable of adjusting risk levels promptly based on changes in environmental parameters, providing accurate safety risk information for maintenance personnel's scheduling decisions. This is a significant advantage of this invention over existing technologies that only focus on equipment status while neglecting environmental risks.

[0062] In some embodiments of this application, a four-dimensional collaborative scheduling system is implemented based on fault diagnosis results, real-time personnel location, personnel workload status, and regional risk level to generate an intelligent scheduling scheme. Intelligent scheduling is the core component of this invention. Its purpose is to comprehensively consider four dimensions—fault urgency, personnel skill matching, spatial distance, workload, and safety risk—to select the optimal scheduling target from all candidate maintenance personnel and plan the optimal scheduling path. The scheduling decision-making process employs a multi-objective optimization algorithm to maximize scheduling efficiency and maintenance quality while minimizing safety risks, all while satisfying various constraints.

[0063] The specific steps for generating the intelligent scheduling scheme include: The first step is to determine the fault type and urgency based on the fault diagnosis results. The fault type determines the required skills of the maintenance personnel, while the urgency determines the scheduling priority. For faults with high urgency, the system will prioritize scheduling, even at the expense of some other optimization objectives. The determination of urgency is based on factors such as the severity of the fault, the scope of its impact on production, and the potential secondary hazards it may cause.

[0064] The second step involves matching qualified maintenance personnel with the appropriate skills based on the fault type. The maintenance personnel database records each personnel's skill tags, education and training experience, certification status, and historical maintenance records. Skill tags employ a standardized classification system, covering various professional fields such as electrical, instrumentation, mechanical, and piping, as well as maintenance skills for various types of equipment. The system automatically filters out qualified personnel with the corresponding skills based on the fault type, generating a candidate list.

[0065] The third step is to calculate the spatial distance between each candidate maintenance personnel and the faulty equipment. In chemical plants, spatial distance includes not only straight-line distance but also factors such as building structure, access restrictions, and floor differences. Based on indoor map data of the chemical plant, the system uses a weighted shortest path algorithm to calculate the optimal reachable distance from each candidate personnel to the faulty equipment, taking into account factors such as travel distance, passage width, and staircase usage costs.

[0066] The fourth step is to obtain the workload score for each candidate maintenance worker. The workload score is calculated and updated in real time by the spatiotemporal intelligence model; a lower score indicates that the worker has more spare energy to accept new tasks. The system obtains the workload score for each candidate worker from the spatiotemporal intelligence module and makes a comprehensive judgment based on the worker's current task status.

[0067] The fifth step is to calculate the comprehensive regional risk score for each candidate maintenance personnel's path to the faulty equipment. The comprehensive regional risk score considers factors such as the risk level, personnel density, and hazard distribution of each area traversed by the path. First, the system identifies all areas traversed by the optimal path from the candidate personnel to the faulty equipment; then, it obtains the current risk level of each area; next, it calculates the real-time risk value of the area by combining the personnel density and hazard distribution; finally, it performs a weighted summation of the real-time risk values ​​of all areas traversed by the path to obtain the comprehensive regional risk score for the path.

[0068] Step 6 involves weighted fusion based on spatial distance, workload score, regional risk comprehensive score, and urgency level to determine the optimal scheduling target and optimal scheduling path. The fusion algorithm employs a multi-objective optimization method, comprehensively considering scores from multiple dimensions for decision-making. Weight coefficients can be dynamically adjusted according to fault type and urgency level; for example, the safety weight coefficient will be appropriately increased for high-risk faults, and the response speed weight coefficient will be increased for urgent faults. Through weighted fusion, a comprehensive score is calculated for each candidate, and the person with the highest comprehensive score is selected as the optimal scheduling target, with their optimal path used as the recommended route.

[0069] In practice, the scheduling decision module also supports the adjustment and optimization of scheduling schemes. When the optimal scheduling object cannot respond to scheduling instructions, the system will automatically select the suboptimal scheme; when there are temporary obstacles on the optimal path, the system will automatically replan the path; when the fault situation changes, the system will automatically update the scheduling scheme. The scheduling decision module supports a human-machine collaborative mode, that is, after the system generates a recommended scheme, it is executed after being confirmed or adjusted by the management personnel.

[0070] By adopting the above technical solution and employing a four-dimensional collaborative scheduling mechanism, scheduling decisions are made by comprehensively considering four dimensions: fault type, personnel location, workload, and regional risk, thereby achieving scientific and rational scheduling of maintenance personnel. Compared with existing technologies, the scheduling dimensions of this invention are more comprehensive, maximizing scheduling efficiency while ensuring safety. Compared with existing technologies such as WO2020193330, US20200057691, CN105791379, US10966059, and CN117436651, which only focus on one or a few dimensions, the four-dimensional collaborative scheduling of this invention has significant technical advantages.

[0071] In other implementations, the four-dimensional collaborative scheduling can generate an intelligent scheduling scheme based on at least three dimensions: fault diagnosis results, real-time personnel location, personnel workload status, and regional risk level. For example, in some application scenarios, scheduling can be based on three dimensions: fault diagnosis results, real-time personnel location, and personnel workload status, without considering the regional risk level; or in other scenarios, scheduling can be based on these three dimensions, without considering the personnel workload status. The system can flexibly select the combination of scheduling dimensions according to the actual application scenario and business needs, reducing system complexity while ensuring scheduling effectiveness.

[0072] In some preferred embodiments of this application, the intelligent dispatching scheme is pushed to the mobile terminals of the target maintenance personnel. The push module is responsible for timely and accurate transmission of dispatching instructions to the selected maintenance personnel, ensuring the effective execution of dispatching decisions. Multiple push methods are supported, including but not limited to mobile application push notifications, SMS notifications, instant messaging, and voice broadcasts. Push content includes faulty equipment information, fault type, urgency level, recommended route, on-site photos or videos, handling suggestions, and safety precautions. After receiving the dispatching instructions via their mobile terminals, maintenance personnel can view detailed information, confirm receipt, plan action routes, and provide feedback on execution status.

[0073] In its implementation, the push module also supports a closed-loop feedback mechanism. During maintenance tasks, personnel can provide feedback on fault diagnosis conclusions, including evaluations of diagnostic accuracy, adoption status of suggested solutions, and supplementary fault information. This feedback is uploaded to the system via mobile devices, and the system updates its fault diagnosis knowledge base based on this information, triggering incremental learning of the predictive maintenance model. This closed-loop feedback mechanism enables the system to continuously learn and optimize from actual operation, improving diagnostic accuracy and scheduling efficiency.

[0074] By adopting the above technical solution, the push module enables timely transmission of scheduling instructions and closed-loop collection of execution feedback, ensuring the effective execution of scheduling decisions and continuously optimizing system performance through the feedback mechanism.

[0075] In some preferred embodiments of this application, the method for calculating workload scores specifically includes the following steps: First, the movement trajectory data of personnel within a preset time window is acquired. The length of the preset time window can be set according to actual management needs, preferably 2 to 8 hours. A time window that is too short may not accurately reflect the workload of personnel, while a time window that is too long may ignore changes in the current work status of personnel. The movement trajectory data includes all locations traversed by personnel within the time window, the coordinates of each location, the entry and exit times, and changes in movement speed. This data is collected in real time by the personnel positioning system and transmitted to the spatiotemporal intelligence module for processing.

[0076] Then, a spatiotemporal behavioral feature extraction model is used to extract personnel behavior feature vectors from the movement trajectory data. This model is constructed using deep learning algorithms, taking the original trajectory data as input and automatically learning and extracting high-dimensional features reflecting personnel's workload and behavioral patterns. The model architecture employs an encoder-decoder structure. The encoder consists of a multi-layer spatiotemporal convolutional network and a recurrent neural network. The spatiotemporal convolutional network extracts features from the spatial dimension of the trajectory data, while the recurrent neural network extracts features from the temporal dimension. The decoder consists of a fully connected network that maps the feature vectors output by the encoder to the workload feature space. The model is trained using a contrastive learning method, which brings trajectory samples with similar workload states closer together in the feature space and widens the distance between different samples, enabling the model to learn a more discriminative representation of workload features.

[0077] Finally, the workload score is calculated by combining the employee's current task status and the extracted behavioral feature vector. The calculation of the workload score requires comprehensive consideration of both the behavioral feature vector and the current task status. The behavioral feature vector reflects the employee's historical activity patterns; for example, frequent movement and long working hours usually indicate high workload. The current task status reflects the employee's current working situation, such as whether a task is being performed, the estimated completion time of the task, and the complexity of the task. The final workload score comprehensively reflects the feasibility and suitability of the employee to accept a new task at a specific time.

[0078] By adopting the above technical solution, the workload status of personnel can be accurately assessed through multi-dimensional data analysis and deep learning models, providing a scientific basis for scheduling decisions, effectively avoiding excessive fatigue of personnel, and improving the quality and safety of maintenance work.

[0079] In some preferred embodiments of this application, the method for calculating the regional risk comprehensive score specifically includes the following steps: First, the historical risk levels and real-time environmental parameters of each area traversed by the scheduling path are obtained. The risk level of each area is calculated and maintained in real time by the safety risk assessment model, including information such as the coordinate boundaries of each area, risk type, historical risk level, and current real-time environmental parameters. Real-time environmental parameters include real-time data reflecting the safety status of the area, such as current gas concentration, temperature, humidity, and personnel density.

[0080] Then, real-time risk values ​​for each region are calculated based on real-time environmental parameters and historical risk levels. The calculation of real-time risk values ​​employs a multi-factor weighted model, using environmental parameters and historical risk levels as inputs, and summing them using preset weighting coefficients to obtain a quantified real-time risk score. These weighting coefficients can be set and adjusted based on historical accident data and expert experience to reflect the degree of influence of different factors on regional risk.

[0081] Finally, a comprehensive regional risk score is generated based on real-time risk values ​​and regional population density. Population density is a crucial factor influencing regional risk; higher density increases the potential for casualties and impacts in the event of an accident. Therefore, the comprehensive regional risk score needs to consider both real-time risk values ​​and population density, and can be calculated using a product model or a weighted summation model. A higher comprehensive regional risk score indicates a greater risk for the route.

[0082] In practical implementation, the calculation of the regional risk comprehensive score also needs to consider path planning factors. When multiple paths exist from candidate personnel to faulty equipment, the system calculates the regional risk comprehensive score for each path and selects the path with the lowest score as the recommended path. If it is necessary to pass through a high-risk area, the system will issue a risk warning, and scheduling decision-makers can decide whether to adjust the scheduling plan based on the actual situation. Classic path planning algorithms such as A* algorithm, Dijkstra's algorithm, and fast random search tree algorithm can be used. A risk penalty term is introduced into the cost function of path search to guide the algorithm to prioritize lower-risk paths.

[0083] By adopting the above technical solution, the regional risk comprehensive score of the scheduling path is intelligently calculated by comprehensively considering factors such as regional risk level, real-time environmental parameters and personnel density, providing safety risk information for scheduling decisions and effectively ensuring the personal safety of maintenance personnel during the performance of maintenance tasks.

[0084] In some preferred embodiments of this application, the specific implementation of the predictive maintenance model includes the following steps: First, multi-source data, including vibration signals, temperature parameters, and pressure parameters, are collected from the equipment. Vibration signals are a crucial indicator of the operating status of rotating equipment. By collecting vibration signals through accelerometers installed on the bearing housings, faults such as bearing wear, imbalance, and misalignment can be detected promptly. Temperature is a key parameter reflecting the thermal state of the equipment, including surface temperature, bearing temperature, and winding temperature. Abnormal temperature increases are usually a sign of equipment overload or cooling system failure. Pressure parameters include hydraulic system pressure, gas pipeline pressure, and container internal pressure. Abnormal pressure may indicate pipeline blockage, leakage, or valve malfunction. The fusion of multi-source data can reflect the operating status of the equipment from different perspectives, improving the comprehensiveness and accuracy of fault diagnosis.

[0085] Then, the time-domain and frequency-domain features of the vibration signal are extracted. Time-domain features are calculated directly from the time-domain signal and include mean, variance, standard deviation, peak value, peak-to-peak value, root mean square (RMS), waveform factor, impulse factor, peak factor, and kurtosis. These features reflect the statistical characteristics and distribution features of the signal; for example, the RMS value is related to vibration energy, and kurtosis is related to the impact component. Frequency-domain features are extracted after converting the time-domain signal to the frequency domain using Fourier transform. These include the dominant frequency, frequency energy distribution, spectral peaks and their corresponding frequencies, and sideband features. Frequency-domain features reflect the vibration source and fault characteristics of the equipment; for example, bearing faults are usually manifested as spectral peaks at specific frequencies, and modulation phenomena are manifested as sidebands. The feature extraction process typically uses a sliding window method, segmenting the continuously acquired signal for feature extraction to form a feature time series.

[0086] Next, pattern matching is performed on time-domain and frequency-domain features based on the fault diagnosis knowledge base to identify fault types. The fault diagnosis knowledge base is a structured representation of domain expert experience and historical fault cases, containing typical feature patterns, causes, and maintenance methods for various equipment faults. The pattern matching process calculates the similarity between the extracted features and the fault patterns stored in the knowledge base. When the similarity exceeds a preset threshold, the corresponding fault type is determined. The knowledge base can be represented as a rule base, a case base, or a hybrid of both, supporting dynamic updates and expansion.

[0087] Finally, fault diagnosis conclusions and handling recommendations are generated. The diagnosis conclusions include information such as faulty equipment identification, fault occurrence time, fault type, fault severity, fault location, and fault cause analysis. The handling recommendations include recommended handling methods, required tools and spare parts, estimated repair time, and safety precautions. The diagnosis conclusions and handling recommendations are output in a structured format to facilitate subsequent scheduling decisions and maintenance execution.

[0088] By adopting the above technical solution, through multi-source data acquisition and intelligent feature extraction, combined with expert knowledge base and machine learning model, automatic identification and diagnosis of equipment faults can be achieved, significantly improving the timeliness of fault detection and the accuracy of diagnosis.

[0089] In some preferred embodiments of this application, the spatiotemporal intelligent model is trained based on historical personnel trajectory data and used to extract personnel behavior patterns from real-time location data and assess workload status. The training process of the spatiotemporal intelligent model employs a combination of supervised and unsupervised learning methods. Supervised learning utilizes labeled data to train the model, including historical personnel trajectories and corresponding expert-evaluated workload levels. Unsupervised learning utilizes unlabeled trajectory data to learn personnel behavior patterns, improving the model's generalization ability. After model training is complete, new trajectory data can be analyzed in real time, outputting workload scores and behavior pattern analysis results.

[0090] In practical implementation, the spatiotemporal intelligent model also needs to consider the impact of individual differences. Different personnel have varying work efficiencies and processing capabilities; the same amount of activity may mean different workload levels for different individuals. Therefore, the model needs to establish a personalized baseline model for each maintenance worker, adjusting evaluation parameters based on their historical data to improve the accuracy of the assessment. Furthermore, the model also needs to consider the differences in the type of work task; for example, performing complex maintenance tasks has different impacts on workload than performing simple inspection tasks, requiring the introduction of task type as an auxiliary input into the model.

[0091] By adopting the above technical solution, the spatiotemporal intelligent model can realize the intelligent assessment of personnel workload, provide personnel status information for scheduling decisions, and effectively avoid personnel over-fatigue.

[0092] In other implementations, the assessment of personnel workload status may be carried out using one or more of the following methods in combination: (1) Evaluation based on the number of work orders currently pending by the personnel. The system obtains the number of work orders that the maintenance personnel have not yet completed. When the number of work orders pending exceeds the preset threshold, it is determined that the personnel are in a high-load state and should not be assigned new maintenance tasks.

[0093] (2) Evaluation based on personnel movement trajectory data within a preset time window. As described above, the spatiotemporal intelligent model method assesses workload by analyzing personnel movement trajectories and behavioral characteristics.

[0094] (3) Evaluation based on the personnel's historical work order completion efficiency and / or working time. The system counts the number of work orders completed by the personnel within a preset time window, the average completion time, the cumulative working time, etc., and compares them with the personnel's historical efficiency benchmark to determine their current workload status.

[0095] (4) Evaluation based on personnel fatigue detection data. Collect personnel's physiological indicators (such as heart rate and skin temperature) through wearable devices (such as smart bracelets and smart helmets), or detect personnel's mental state through video analysis technology. When fatigue characteristics are detected, it is determined that the personnel are in a high-load state.

[0096] The above-mentioned assessment methods can be used individually or in combination to comprehensively judge the workload status of personnel and improve the accuracy of the assessment.

[0097] In some preferred embodiments of this application, the method further includes a self-evolution mechanism, which receives feedback from maintenance personnel on fault diagnosis conclusions, updates the fault diagnosis knowledge base based on the feedback information, and triggers incremental learning of the predictive maintenance model. The self-evolution mechanism is a key mechanism for ensuring continuous system optimization and adaptation to change. It enables the system to learn from actual maintenance experience, continuously accumulate new fault modes and maintenance knowledge, and improve the accuracy and adaptability of diagnosis.

[0098] Feedback from maintenance personnel on fault diagnosis conclusions includes multiple dimensions: evaluation of diagnostic accuracy, i.e., the maintenance personnel's assessment of whether the fault type diagnosis given by the system is correct, which can be a qualitative evaluation such as correct, basically correct, or incorrect, or a specific accuracy rate score; adoption of handling suggestions, i.e., the degree to which maintenance personnel adopt the handling suggestions given by the system, including information such as complete adoption, partial adoption, non-adoption, and modification suggestions; and supplementary fault information, i.e., fault characteristics or fault causes that maintenance personnel discover during actual maintenance that were not detected by the system or that were not identified. This information is of great value for improving the fault diagnosis knowledge base.

[0099] When updating the fault diagnosis knowledge base based on feedback information, the system first analyzes and assesses the quality of the feedback information, selecting valuable information for knowledge base updates. For cases where the diagnostic accuracy is evaluated as incorrect, the system analyzes the cause of the error. If the knowledge base lacks pattern knowledge of this type of fault, it adds the case as a new fault. If the model judgment is incorrect, the relevant features are optimized and adjusted. For fault information supplemented by maintenance personnel, the system compares and analyzes it with the original diagnostic conclusions, extracting valuable new features or fault patterns and adding them to the knowledge base. The knowledge base is updated incrementally, supporting online updates without affecting the normal operation of the system.

[0100] When incremental learning of the predictive maintenance model is triggered, the system uses the updated knowledge base as training data to incrementally train the predictive maintenance model. Incremental learning methods can update model parameters using new samples without retraining the entire model, enabling the model to learn new fault modes while maintaining its learning of existing knowledge. Commonly used incremental learning methods include online learning, transfer learning, and incremental support vector machines. After incremental learning is completed, the model's diagnostic capabilities are improved, enabling accurate identification of newly added fault types.

[0101] By adopting the above technical solution, a closed-loop system of diagnosis-scheduling-feedback-optimization is established through a self-evolution mechanism, enabling the system to continuously learn and evolve from actual operation, thereby continuously improving the accuracy of fault diagnosis and the rationality of scheduling decisions, and achieving continuous improvement in the level of intelligence.

[0102] In some embodiments of this application, a human-machine-environment collaborative intelligent operation and maintenance system is provided. This system comprises six core components: a three-source data collaborative acquisition module, a predictive maintenance module, a spatiotemporal intelligence module, a security risk assessment module, a scheduling decision module, and a push module. These components are connected through standardized data interfaces to achieve data sharing and functional collaboration.

[0103] The three-source data collaborative acquisition module is responsible for collecting equipment operation data, personnel location data, and regional environmental data, serving as the foundational data source for the entire system. This module employs a distributed architecture, deploying corresponding sensors and data acquisition devices in various equipment and areas, and transmitting the data to the central data server via an industrial communication network. The data acquisition module supports multiple communication protocols and data formats, adapting to the data acquisition needs of different types of equipment.

[0104] The predictive maintenance module is responsible for identifying faulty equipment based on the predictive maintenance model and generating fault diagnosis results, thus achieving intelligent diagnosis of equipment faults. This module adopts a hybrid architecture combining edge computing and cloud computing. Edge computing nodes are responsible for real-time data processing and fault early warning, while the cloud computing platform is responsible for model training and complex diagnostic calculations.

[0105] The spatiotemporal intelligence module is responsible for acquiring real-time personnel location and workload status based on a spatiotemporal intelligence model, providing personnel status information for scheduling decisions. This module integrates with the positioning infrastructure, receiving personnel location data in real time and calculating workload scores.

[0106] The safety risk assessment module is responsible for generating regional risk levels based on the safety risk assessment model, providing safety risk information for scheduling decisions. This module interfaces with the enterprise's safety management information system to obtain regional risk information and hazard source data.

[0107] The scheduling decision module is the core module of the system, responsible for generating the optimal scheduling scheme by integrating information from various aspects. This module receives data and information from other modules and generates a scheduling scheme through a multi-objective optimization algorithm.

[0108] The push module is responsible for promptly delivering dispatch plans to the mobile terminals of target maintenance personnel, ensuring the effective transmission of dispatch instructions. This module pushes dispatch instructions to maintenance personnel's smart terminals via mobile networks, supporting multiple push methods such as SMS, APP push, and instant messaging.

[0109] By adopting the above technical solution, a modular system architecture is used to achieve collaborative perception and intelligent scheduling of equipment, personnel and environment. Data sharing and functional collaboration among modules jointly support intelligent operation and maintenance of human-machine-environment collaboration in chemical industry scenarios.

[0110] In some preferred embodiments of this application, the scheduling decision module includes five sub-units: a skill matching unit, a distance calculation unit, a load assessment unit, a risk assessment unit, and an optimal decision unit. These sub-units work collaboratively to achieve comprehensive optimization of maintenance scheduling across multiple factors.

[0111] The skills matching unit is responsible for matching candidate maintenance personnel based on the equipment type and fault type of the faulty equipment. The unit maintains skill files for each maintenance worker, recording their professional background, education and training experience, certification status, professional skills, and historical maintenance records. When a fault occurs, the system automatically matches maintenance personnel with the corresponding skills as candidates based on the fault type.

[0112] The distance calculation unit is responsible for calculating the spatial distance between candidate maintenance personnel and the faulty equipment. Based on the indoor map data of the chemical plant, the distance calculation unit uses a weighted shortest path algorithm to calculate the optimal reachable distance from each candidate personnel to the faulty equipment, taking into account factors such as straight-line distance, floor differences, passage restrictions, and travel time.

[0113] The load assessment unit is responsible for calculating the workload score of candidate maintenance personnel. The unit obtains the workload score for each candidate from the spatiotemporal intelligence module; this score is updated in real time to reflect the personnel's current working status.

[0114] The risk assessment unit is responsible for calculating the comprehensive risk score of the area along the path from candidate maintenance personnel to the faulty equipment. The risk assessment unit obtains the real-time risk level of each area from the safety risk assessment module and calculates the comprehensive risk score of the area traversed by the candidate personnel's path to the faulty equipment.

[0115] The optimal decision-making unit is responsible for determining the optimal scheduling scheme based on a weighted fusion of spatial distance, workload score, and regional risk comprehensive score. The optimal decision-making unit weights and fuses the evaluation results from the above four dimensions, selecting the candidate with the highest comprehensive score as the optimal scheduling object, and the corresponding optimal path is recommended as the maintenance route. The weighting coefficients can be dynamically adjusted according to the fault type and urgency.

[0116] By adopting the above technical solution, multi-unit collaborative work is used to achieve comprehensive optimization of maintenance scheduling by taking into full account factors such as skill matching, spatial distance, workload and safety risks, so as to ensure the scientific, rational and safe nature of scheduling decisions.

[0117] In some preferred embodiments of this application, the system further includes a self-evolution module for receiving feedback from maintenance personnel and updating the fault diagnosis knowledge base, triggering incremental learning of the predictive maintenance model. The self-evolution module is a key component for the system to achieve continuous optimization and adaptive learning. By establishing a closed loop of user feedback, it enables the system to continuously accumulate new knowledge, improve performance, and adapt to environmental changes.

[0118] In practice, the self-evolution module receives feedback from maintenance personnel via mobile terminals, including diagnostic accuracy evaluations, adoption status of handling suggestions, and supplementary fault information. The module parses, assesses the quality of, and extracts knowledge from this feedback, transforming valuable information into structured knowledge entries. For cases with negative diagnostic accuracy evaluations, the module analyzes the causes of the diagnostic errors, identifies deficiencies in the knowledge base, and adds new fault characteristic patterns or fault cases to the knowledge base. For fault information supplemented by maintenance personnel, the module performs correlation analysis with existing knowledge, extracts valuable new knowledge, and updates the knowledge base. After the knowledge base is updated, the module triggers the incremental learning process of the predictive maintenance model, retraining the model parameters using the updated knowledge base to improve the model's diagnostic capabilities. The self-evolution module supports both periodic batch learning and online incremental learning modes, allowing for flexible selection based on system load.

[0119] By adopting the above technical solution, a self-learning and continuous optimization mechanism is established through the self-evolution module, enabling the system to continuously grow from actual operating experience, and continuously improve diagnostic accuracy and intelligence level.

[0120] In some embodiments of this application, an electronic device is provided, including a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the steps of the human-machine-environment collaborative intelligent operation and maintenance method described above. The electronic device can be a computing device such as a server, workstation, or mobile terminal. The processor is responsible for executing the program or instructions, and the memory is responsible for storing the program instructions and runtime data.

[0121] In specific implementation, for example, the electronic device can be a mobile electronic device such as a mobile phone, tablet computer, laptop computer, PDA, in-vehicle electronic device, wearable device, super mobile personal computer, personal digital assistant, etc., or a non-mobile electronic device such as a server, network attached storage, personal computer, television, ATM, or self-service machine, etc. This application embodiment does not impose specific limitations. The electronic device used for human-machine-environment collaborative intelligent operation and maintenance in chemical industry scenarios in this application embodiment can be an electronic device with an operating system. This operating system can be Android, iOS, or other possible operating systems, and this application embodiment does not impose specific limitations.

[0122] By adopting the above technical solution, the intelligent operation and maintenance method of the present invention is implemented through electronic devices, enabling the method to be deployed and run in software form, which has the advantages of flexible deployment, controllable cost, and easy upgrade and maintenance.

[0123] In some embodiments of this application, a readable non-transitory storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, it implements the steps of the human-machine-environment collaborative intelligent operation and maintenance method described above. The readable non-transitory storage medium includes computer-readable non-transitory storage media, such as computer read-only memory, random access memory, magnetic disk, or optical disk.

[0124] In practical implementation, programs or instructions stored on readable non-transitory storage media, when executed by a processor, can control the processor to execute various steps in the method of this invention, realizing functions such as equipment operation data acquisition, fault diagnosis, workload assessment, regional risk assessment, intelligent scheduling scheme generation and push. Readable non-transitory storage media have the characteristics of large storage capacity, long data retention time, and fast read / write speed, making them suitable for storing and deploying the intelligent operation and maintenance software system of this invention.

[0125] By adopting the above technical solution, the method of the present invention can be software-borne through a readable non-transitory storage medium, which facilitates the distribution, installation and deployment of software products and expands the application scope of the present invention.

[0126] This invention, through the aforementioned technical solution, achieves collaborative perception and intelligent scheduling of equipment failures, personnel locations, and regional risks in chemical engineering scenarios. The system comprehensively acquires equipment operating status, personnel location trajectories, and regional environmental information through three-source data collaborative acquisition technology. It achieves intelligent diagnosis of equipment failures and accurate assessment of personnel workload through predictive maintenance models and spatiotemporal intelligent models, dynamically monitors regional risks through a safety risk assessment model, and scientifically and rationally schedules maintenance personnel through a four-dimensional collaborative scheduling algorithm. Compared with existing technologies, this invention introduces personnel workload assessment into the field of operation and maintenance scheduling for the first time. Through the collaborative fusion of three-source data and the four-dimensional collaborative scheduling mechanism, it achieves multi-dimensional collaborative perception and intelligent scheduling of equipment, personnel, and the environment, effectively solving the shortcomings of existing technologies in personnel workload assessment and improving operation and maintenance efficiency and the safety of maintenance personnel. By establishing a closed-loop system of diagnosis-scheduling-feedback through a self-evolutionary mechanism, the system can continuously learn and optimize, significantly improving the level of intelligent operation and maintenance of equipment in chemical enterprises.

[0127] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A human-machine-environment collaborative intelligent operation and maintenance method, characterized in that, include: The three-source data collaborative acquisition module synchronously collects equipment operation data, personnel location data, and regional environmental data. Based on a predictive maintenance model, the equipment operation data is analyzed to identify faulty equipment and generate fault diagnosis results; based on a spatiotemporal intelligent model, the personnel location data is processed to obtain the real-time location of personnel and assess their workload status; based on a safety risk assessment model, the regional environmental data is analyzed to generate a regional risk level. Based on the fault diagnosis results, the real-time location of the personnel, the workload status of the personnel, and the risk level of the area, four-dimensional collaborative scheduling is performed to generate an intelligent scheduling scheme. The intelligent scheduling scheme is pushed to the mobile terminal of the target maintenance personnel.

2. The method according to claim 1, characterized in that, The intelligent scheduling scheme includes: determining the fault type and urgency based on the fault diagnosis results; matching candidate maintenance personnel with corresponding skills based on the fault type; calculating the spatial distance between each candidate maintenance personnel and the faulty equipment; obtaining the workload score of each candidate maintenance personnel; calculating the regional risk comprehensive score of the path for each candidate maintenance personnel to reach the faulty equipment; and performing weighted fusion based on the spatial distance, the workload score, the regional risk comprehensive score, and the urgency to determine the optimal scheduling object and the optimal scheduling path.

3. The method according to claim 2, characterized in that, The method for calculating the workload score includes: acquiring personnel movement trajectory data within a preset time window; extracting personnel behavior feature vectors from the movement trajectory data based on a spatiotemporal behavior feature extraction model; and calculating the personnel workload score by combining the personnel's current task status and the behavior feature vectors.

4. The method according to claim 2, characterized in that, The method for calculating the comprehensive regional risk score includes: obtaining the historical risk level and real-time environmental parameters of each region traversed by the scheduling path; calculating the real-time risk value of each region based on the real-time environmental parameters and the historical risk level; and generating a comprehensive regional risk score based on the real-time risk value and the regional population density.

5. The method according to claim 1, characterized in that, The predictive maintenance model is generated based on multi-source operating data of the equipment, including: collecting vibration signals, temperature parameters and pressure parameters of the equipment; extracting time-domain features and frequency-domain features of the vibration signals; performing pattern matching on the time-domain features and frequency-domain features based on a fault diagnosis knowledge base to identify fault types; and generating fault diagnosis conclusions and handling suggestions.

6. The method according to claim 1, characterized in that, The spatiotemporal intelligent model is trained and generated based on historical trajectory data of personnel, and is used to extract personnel behavior patterns from real-time location data and assess workload status.

7. The method according to claim 1, characterized in that, It also includes a self-evolution mechanism: receiving feedback from maintenance personnel on the fault diagnosis conclusions; updating the fault diagnosis knowledge base based on the feedback information; and triggering incremental learning of the predictive maintenance model.

8. The method according to claim 1, characterized in that, The assessment of personnel workload status includes at least one of the following methods: assessment based on the number of work orders currently pending; assessment based on personnel movement trajectory data within a preset time window; The evaluation is based on the staff's historical work order completion efficiency and / or working hours; the evaluation is based on staff fatigue detection data.

9. The method according to claim 1, characterized in that, The identification of faulty equipment and generation of fault diagnosis results includes at least one of the following methods: fault identification based on a predictive maintenance model; pattern matching based on a fault diagnosis knowledge base; fault reasoning based on expert system rules; and fault judgment based on equipment operating parameter thresholds.

10. The method according to claim 1, characterized in that, The four-dimensional collaborative scheduling includes generating an intelligent scheduling scheme based on at least three dimensions: the fault diagnosis results, the real-time location of personnel, the workload status of personnel, and the regional risk level.

11. A human-machine-environment collaborative intelligent operation and maintenance system, characterized in that, include: The three-source data collaborative acquisition module is used to simultaneously collect equipment operation data, personnel location data, and regional environmental data. The predictive maintenance module is used to analyze the equipment operation data based on the predictive maintenance model, identify faulty equipment, and generate fault diagnosis results; the spatiotemporal intelligence module is used to process the personnel location data based on the spatiotemporal intelligence model, obtain the real-time location of personnel, and assess the personnel workload status. The safety risk assessment module is used to analyze the regional environmental data based on the safety risk assessment model and generate the regional risk level. The scheduling decision module is used to perform four-dimensional collaborative scheduling based on the fault diagnosis results, the real-time location of the personnel, the workload status of the personnel, and the risk level of the area, and generate an intelligent scheduling scheme. The push module is used to push the intelligent scheduling scheme to the mobile terminal of the target maintenance personnel.

12. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein when the program or instructions are executed by the processor, they implement the steps of the human-machine-environment collaborative intelligent operation and maintenance method as described in any one of claims 1 to 10.

13. A readable non-transitory storage medium, characterized in that, The storage medium stores a program or instructions, which, when executed by a processor, implement the steps of the human-machine-environment collaborative intelligent operation and maintenance method as described in any one of claims 1 to 10.

Images