CIR equipment running state simulation and diagnosis method and system

Abstract

The invention provides a CIR equipment operation state simulation and diagnosis method and system, and relates to the technical field of equipment state monitoring and fault diagnosis, and the method comprises the steps: obtaining multi-dimensional physical state data of target equipment in a current operation period; performing feature space mapping based on the equipment operation mechanism constraint and the historical degradation law to obtain a state feature vector; constructing a virtual operation environment, and executing time sequence deduction by taking the current state parameter as a simulation initial condition to obtain a simulation state sequence; performing space-time alignment on the simulation state sequence and the actual measurement state sequence, identifying an abnormal state interval by calculating deviation measurement, and tracing to an equipment physical structure hierarchy to realize abnormal source positioning; generating a diagnosis conclusion and an intervention strategy in combination with the operation constraint rule; and carrying out adaptive correction on evolution mechanism parameters of the virtual operation environment based on the intervened state response data. According to the invention, accurate simulation and abnormal traceability positioning of the equipment state can be realized, and the diagnosis accuracy and real-time performance are improved.

Description

Technical Field

[0001] This invention relates to the field of equipment condition monitoring and fault diagnosis technology, and in particular to CIR equipment operation status simulation and diagnosis methods and systems. Background Technology

[0002] With the deepening of industrial intelligent and digital transformation, the monitoring and fault diagnosis technology of complex equipment operation status has become a key link in ensuring production safety and improving equipment reliability. Traditional equipment condition monitoring mainly relies on periodic inspections and single parameter threshold alarms, which is difficult to meet the needs of modern industry for the whole life cycle health management of equipment.

[0003] As a critical device in chemical production, the operational stability of CIR (Chemical Injection and Recycling) equipment directly affects the safety and efficiency of the entire production line. Due to the complex chemical reactions, multiphase flows, and heat and mass transfer processes involved within CIR equipment, its failure modes exhibit diversity, concealment, and coupling characteristics. Traditional monitoring methods struggle to address the unique challenges of CIR equipment, such as progressive performance degradation due to catalyst poisoning, decreased heat transfer efficiency caused by reactor wall scaling, and uneven flow distribution due to ejector blockage—all complex failure scenarios.

[0004] In recent years, with the development of IoT technology, big data analytics, and the concept of digital twins, equipment condition simulation and diagnosis methods based on data-driven approaches and mechanistic model integration have gradually become a research hotspot. By constructing virtual operating models of CIR equipment, it is possible to achieve real-time tracking of its unique operating states, prediction of future states, and early identification of abnormal operating conditions, providing a scientific basis for equipment maintenance decisions. However, existing CIR equipment condition simulation and diagnosis technologies still face many challenges in practical applications, such as difficulties in multiphase reaction modeling, complex chemical-physical coupling mechanisms, and inaccurate fault feature extraction, necessitating the development of more accurate and effective dedicated diagnostic methods. Summary of the Invention

[0005] The present invention provides a method and system for simulating and diagnosing the operating status of CIR equipment, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a method for simulating and diagnosing the operating status of CIR equipment, comprising: Acquire multi-dimensional physical state data and current state parameters of the target device within the current operating cycle; Based on the constraints of equipment operation mechanism and historical degradation patterns, feature mapping is performed on the multi-dimensional physical state data to obtain a state feature vector that reflects the internal coupling relationship of the equipment. A virtual operating environment is constructed based on the state feature vector. By using the current state parameters of the device as the initial conditions for simulation, a time-series simulation is performed in the virtual operating environment to obtain a simulation state sequence containing evolution trajectories at multiple time scales. The simulated state sequence is spatiotemporally aligned with the actual measured state sequence of the device. Abnormal state intervals are identified by calculating the deviation metric between the two in the state space. The abnormal state intervals are then traced back to the physical structure level of the device based on the feature components corresponding to them, thus obtaining the abnormal source localization result. Based on the anomaly source location results and combined with equipment operation constraint rules, diagnostic conclusions and intervention strategies are generated. Based on the actual intervention effect of the intervention strategy, the device status is tracked and collected, and the evolution mechanism parameters in the virtual operating environment are adaptively corrected using the status response data after intervention.

[0007] Based on the constraints of equipment operation mechanisms and historical degradation patterns, feature mapping is performed on the multi-dimensional physical state data to obtain state feature vectors reflecting the internal coupling relationships of the equipment, including: Multidimensional physical state data is decomposed into multiple physical domain subsets according to physical domain attributes, and time-domain statistical features and frequency-domain energy distribution features are extracted from each physical domain subset. Based on the constraints of the device operation mechanism, a causal dependency graph between physical domains is constructed, and the interaction path between the sub-data sets of each physical domain is determined according to the causal dependency graph. By calculating the similarity between each physical domain subset and the samples of each degradation stage in the preset degradation trajectory library, the current degradation stage label of the device can be identified. Based on the action path in the causal dependency graph and the degradation stage label, coupling weight coefficients are assigned to each physical domain subset. The time-domain statistical features and frequency-domain energy distribution features of each physical domain subset are weighted and fused according to the coupling weight coefficients to obtain a state feature vector.

[0008] A virtual operating environment is constructed based on the state feature vector. By using the current state parameters of the device as the initial conditions for simulation, a time-series simulation is performed in the virtual operating environment to obtain a simulation state sequence containing multi-timescale evolution trajectories, including: Based on the preset physical process model library, the physical process models corresponding to each dimension component in the state feature vector are extracted, the physical process models are combined and the basic simulation framework of the virtual running environment is constructed, and the current state parameters of the device are mapped to the basic simulation framework to obtain the initial configuration of the virtual running environment. Based on the time scale characteristics of the equipment's operating conditions, the simulation time domain is divided into multiple time scale levels. A corresponding time step and simulation iteration cycle are assigned to each time scale level. The virtual operating environment is driven at each time scale level to perform state evolution calculations and obtain the evolution trajectory corresponding to each time scale level. Extract the state inflection points in the evolution trajectory at each time scale level, establish cross-scale synchronization constraint relationships based on the position of the state inflection points on the time axis, and perform time alignment and state consistency verification on the evolution trajectory at different time scale levels. The evolution trajectories at each time scale level, after time alignment and state consistency verification, are serialized and integrated to obtain the simulation state sequence.

[0009] The simulated state sequence is spatiotemporally aligned with the actual measured state sequence of the equipment. Abnormal state intervals are identified by calculating the deviation metric between the two in the state space. Based on the feature components corresponding to the abnormal state intervals, the source is traced back to the physical structure level of the equipment to obtain the anomaly source localization result, including: Based on the timestamp information of the simulated state sequence and the actual measured state sequence of the device, the time offset between the two sequences is identified, and a time axis alignment operation is performed on the two sequences. The simulated state sequence after time axis alignment and the actual measured state sequence of the device are projected onto a preset state space coordinate system to obtain the simulated state trajectory and the measured state trajectory. The point distance between the simulated state trajectory and the measured state trajectory at each corresponding time in the state space coordinate system is calculated, and a time series deviation measurement curve is constructed based on the point distance. The time series deviation measurement curve is subjected to threshold determination to identify the time period in which the deviation measurement exceeds the preset deviation threshold and mark it as an abnormal state interval. Calculate the deviation components of the simulated state trajectory and the measured state trajectory within the abnormal state interval in each dimension of the state space, and identify the dimension with the largest deviation component as the dominant deviation dimension. Based on the feature components corresponding to the dominant deviation dimension, a preset mapping relationship between the feature components and the physical structure level of the device is determined, and the physical components or physical processes corresponding to the feature components are calculated as the anomaly source location results.

[0010] Based on the anomaly source location results and combined with equipment operation constraint rules, diagnostic conclusions and intervention strategies are generated, including: Based on the anomaly source location results, the corresponding equipment operation constraint rules are extracted, the constraint boundary conditions under normal operation are determined, and the deviation items of the state parameters that violate the constraints are identified by combining the actual measurement state sequence of the equipment within the abnormal state interval. Based on the deviation of the state parameters, the failure mode of the anomaly source location result is determined, and the matching fault type is retrieved from the preset fault knowledge base. The fault type is associated with the physical components or physical processes in the anomaly source location result to generate a diagnostic conclusion. Extract a set of candidate intervention actions corresponding to the fault type, evaluate the impact of each candidate intervention action on the equipment operation constraint rules, eliminate candidate intervention actions that cause other constraint boundary conditions to be violated, and select the remaining candidate intervention actions with the highest expected benefits as the intervention strategy.

[0011] Based on the actual intervention effect of the intervention strategy, the device status is tracked and collected, and the evolution mechanism parameters in the virtual operating environment are adaptively corrected using the post-intervention status response data, including: After the intervention strategy is executed, the device is continuously monitored to collect multi-dimensional physical state data of the device under the intervention, obtain the state response data after the intervention, and combine it with the historical state data before the intervention to construct a continuous state sequence containing the complete evolution process before and after the intervention. Based on the continuous state sequence, a new state feature vector is generated, and the new state feature vector is input into the virtual operating environment. The state at a certain moment after intervention is used as the simulation starting point to perform forward deduction and backward backtracking. Extract the forward extrapolation trajectory and the backward backtracking trajectory, and calculate the state convergence error at the intervention moment. When the state convergence error exceeds the error threshold, perform sensitivity analysis on the evolution mechanism parameters at each time scale level in the virtual operating environment. The evolution mechanism parameters refer to the key coefficients that control the change of physical state in the virtual operating environment over time. Based on the sensitivity analysis results, the evolution mechanism parameter that has the most significant impact on the state convergence error is determined, and the evolution mechanism parameter is iteratively adjusted in the opposite direction to the state convergence error.

[0012] A second aspect of the present invention provides a CIR equipment operation status simulation and diagnostic system, comprising: The data acquisition unit is used to acquire multi-dimensional physical state data and current state parameters of the target device during the current operating cycle. The vector generation unit is used to perform feature mapping on the multi-dimensional physical state data based on the constraints of the device's operating mechanism and historical degradation patterns, to obtain a state feature vector that reflects the internal coupling relationship of the device. The simulation and deduction unit is used to construct a virtual operating environment based on the state feature vector, and to perform time-series deduction in the virtual operating environment by using the current state parameters of the device as the initial conditions for simulation, so as to obtain a simulation state sequence containing evolution trajectories at multiple time scales. An anomaly localization unit is used to spatiotemporally align the simulated state sequence with the actual measured state sequence of the equipment, identify the abnormal state interval by calculating the deviation metric between the two in the state space, and trace the source to the physical structure level of the equipment based on the feature components corresponding to the abnormal state interval to obtain the anomaly source localization result. The strategy generation unit is used to generate diagnostic conclusions and intervention strategies based on the anomaly source location results and in combination with equipment operation constraint rules. The parameter correction unit is used to track and collect the device status based on the actual intervention effect of the intervention strategy, and to adaptively correct the evolution mechanism parameters in the virtual operating environment using the status response data after the intervention.

[0013] A third aspect of the present invention, An electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0014] Fourth aspect of the embodiments of the present invention, A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0015] The beneficial effects of this application are as follows: This invention acquires multi-dimensional physical state data of the device and performs feature space mapping, which can accurately capture the coupling relationship between various components inside the device. Compared with traditional diagnostic methods based on a single monitoring parameter, it can more comprehensively reflect the true operating status of the device, avoid misjudgment or omission caused by incomplete information, and improve the accuracy and reliability of status assessment.

[0016] This invention constructs a virtual operating environment and performs time-series deduction to generate a simulated state sequence. By comparing the simulation results with actual measurement data in a spatiotemporal alignment, it can accurately identify abnormal state intervals in the state space. Through the feature component tracing mechanism, the anomaly is located to the specific physical structure level, realizing accurate tracing from macroscopic abnormal phenomena to microscopic fault sources. This significantly improves the location accuracy and diagnostic efficiency of fault diagnosis, and provides clear guidance for equipment maintenance.

[0017] This invention establishes an adaptive correction mechanism based on intervention effect feedback. By tracking and collecting state response data after intervention, the evolution mechanism parameters in the virtual operating environment are dynamically optimized, enabling the simulation model to continuously evolve with changes in the actual operating conditions of the equipment. This ensures the long-term effectiveness and adaptability of the diagnostic system, avoids the problem of decreased diagnostic accuracy caused by equipment aging or changes in operating conditions in traditional static models, and achieves self-optimization and improvement of diagnostic capabilities. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the CIR equipment operation status simulation and diagnosis method according to an embodiment of the present invention; Detailed Implementation To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] Figure 1 This is a flowchart illustrating the CIR equipment operation status simulation and diagnosis method according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes: Acquire multi-dimensional physical state data and current state parameters of the target device within the current operating cycle; Based on the constraints of equipment operation mechanism and historical degradation patterns, feature mapping is performed on the multi-dimensional physical state data to obtain a state feature vector that reflects the internal coupling relationship of the equipment. A virtual operating environment is constructed based on the state feature vector. By using the current state parameters of the device as the initial conditions for simulation, a time-series simulation is performed in the virtual operating environment to obtain a simulation state sequence containing evolution trajectories at multiple time scales. The simulated state sequence is spatiotemporally aligned with the actual measured state sequence of the device. Abnormal state intervals are identified by calculating the deviation metric between the two in the state space. The abnormal state intervals are then traced back to the physical structure level of the device based on the feature components corresponding to them, thus obtaining the abnormal source localization result. Based on the anomaly source location results and combined with equipment operation constraint rules, diagnostic conclusions and intervention strategies are generated. Based on the actual intervention effect of the intervention strategy, the device status is tracked and collected, and the evolution mechanism parameters in the virtual operating environment are adaptively corrected using the status response data after intervention.

[0021] In one optional implementation, based on the constraints of the equipment's operating mechanism and historical degradation patterns, feature mapping is performed on the multi-dimensional physical state data to obtain a state feature vector reflecting the internal coupling relationships of the equipment, including: Multidimensional physical state data is decomposed into multiple physical domain subsets according to physical domain attributes, and time-domain statistical features and frequency-domain energy distribution features are extracted from each physical domain subset. Based on the constraints of the device operation mechanism, a causal dependency graph between physical domains is constructed, and the interaction path between the sub-data sets of each physical domain is determined according to the causal dependency graph. By calculating the similarity between each physical domain subset and the samples of each degradation stage in the preset degradation trajectory library, the current degradation stage label of the device can be identified. Based on the action path in the causal dependency graph and the degradation stage label, coupling weight coefficients are assigned to each physical domain subset. The time-domain statistical features and frequency-domain energy distribution features of each physical domain subset are weighted and fused according to the coupling weight coefficients to obtain a state feature vector.

[0022] The collected multi-dimensional physical state data is divided into physical domain attributes. Taking rotating equipment as an example, the raw data is decomposed into vibration, temperature, current, pressure, and acoustic sub-datasets. The vibration sub-dataset contains triaxial acceleration signals, with a sampling frequency of 10kHz and a sampling duration of 5s, resulting in 50,000 sampling points. The temperature sub-dataset contains measurement data for bearing temperature, winding temperature, and casing temperature, with a sampling frequency of 1Hz. The current sub-dataset records the instantaneous values ​​of three-phase currents, with a sampling frequency of 5kHz. The pressure sub-dataset contains lubricating oil pressure and cooling medium pressure, with a sampling frequency of 10Hz. The acoustic sub-dataset records the sound signals of the equipment operation, with a sampling frequency of 20kHz.

[0023] Time-domain statistical features were extracted from a subset of the vibration domain dataset. The mean of the vibration signal was calculated to be 0.15 m / s², and the standard deviation was 0.42 m / s². 2 The peak value is 3.8 m / s 2 The peak-to-peak value is 7.2 m / s 2 The root mean square value is 0.45 m / s 2The waveform factor (ratio of the absolute average value to the root mean square value of the vibration signal) is 1.38; the peak factor (ratio of the peak value to the root mean square value) is 8.44; the impulse factor (ratio of the peak value to the absolute average value of the vibration signal) is 11.63; and the margin factor (ratio of the peak value to the root mean square amplitude) is 14.2. Fourier transform of the vibration signal yields its frequency domain energy distribution characteristics. The spectrum is divided into three intervals: low frequency (0-500 Hz), mid frequency (500-2000 Hz), and high frequency (2000-5000 Hz). The energy proportions of each frequency band are calculated to be 48%, 35%, and 17%, respectively. The dominant frequency components are extracted as 320 Hz and 640 Hz, corresponding to amplitudes of 2.1 m / s². 2 and 1.3m / s 2 .

[0024] For the temperature domain subset, time-domain statistical features were extracted. The average bearing temperature was calculated to be 65°C, with a temperature rise rate of 2.3°C per hour, a temperature fluctuation range of 3.5°C, and a temperature change trend slope of 0.038°C / min. Spectral analysis of the temperature time series identified periodic fluctuation components, with a main period of 120 min and a corresponding temperature fluctuation amplitude of 1.8°C. For the current domain subset, the average three-phase currents were calculated to be 18.5A, 18.3A, and 18.7A, with an imbalance of 2.1%. Harmonic distortion was obtained through harmonic decomposition of the current signal, with the third harmonic content at 5.2% of the fundamental frequency and the fifth harmonic content at 3.1% of the fundamental frequency.

[0025] A dependency graph reflecting the causal relationships between physical domains is constructed. Based on the equipment's operating mechanism, the path of influence from the current domain to the temperature domain is determined. Since increased current leads to increased winding heating, the propagation delay for this path is set to 180 s. The path of influence from the temperature domain to the vibration domain is based on the thermal expansion effect; increased temperature causes changes in bearing clearance, resulting in altered vibration characteristics, with a propagation delay of 300 s. The path of influence from the vibration domain to the pressure domain is reflected in the increased vibration leading to fluctuations in lubricating oil pressure, with a propagation delay of 60 s. The path of influence from the pressure domain to the temperature domain is based on changes in lubrication conditions affecting frictional heat generation, with a propagation delay of 240 s. The acoustic domain, as a comprehensive representation domain, receives the combined effects of the vibration and temperature domains, with propagation delays of 10 s and 120 s, respectively. These paths constitute a closed-loop coupling structure, reflecting the multi-physics interactions within the equipment.

[0026] Sample data with labeled degradation stages were extracted from the historical operational database. Full lifecycle data from 20 similar devices were selected, each with an operating time between 8000 and 12000 hours. The degradation process was divided into four categories: normal operation, early degradation, intermediate degradation, and severe degradation. Time alignment was performed on samples from each stage, normalizing the root mean square (RMS) vibration values ​​of different devices to the same time scale. Zero-point calibration was performed based on the initial operating state of the device, with the first appearance of obvious degradation characteristics serving as the alignment reference point. When constructing the degradation trajectory library, the RMS values ​​for the normal operation stage in the vibration domain ranged from 0.3 to 0.5 m / s. 2 The early degradation stage is 0.5 to 0.8 m / s. 2 The intermediate degradation phase is 0.8 to 1.5 m / s. 2 The severe degradation stage is greater than 1.5 m / s 2 The bearing temperature ranges for each stage of the temperature range record are 50 to 60°C, 60 to 70°C, 70 to 85°C, and above 85°C, respectively.

[0027] The similarity between the current vibration domain subset and samples at each stage in the degradation trajectory database was calculated using the Euclidean distance metric. The average distance between the current vibration feature vector and samples from the normal operation stage was 0.08, the average distance with samples from the early degradation stage was 0.15, the average distance with samples from the mid-degradation stage was 0.32, and the average distance with samples from the severe degradation stage was 0.58. The stage corresponding to the smallest distance was selected as the preliminary judgment result. Simultaneously, the similarity of the temperature domain subset was calculated. The current temperature feature had a similarity of 0.87 with samples from the early degradation stage, and its similarity with other stages was below 0.6. Combining the similarity calculation results from the vibration and temperature domains, a voting mechanism was used to determine that the current device is in the early degradation stage, and the label value for this stage was set to 2.

[0028] Coupling weight coefficients are assigned based on the action path strength in the causal dependency graph and the label of the current degradation stage. In the early degradation stage, the vibration domain has the highest importance in representing the overall state, and is assigned a weight coefficient of 0.35. The temperature domain, as a secondary indicator domain, is assigned a weight coefficient of 0.25. The current domain reflects the driving state and is assigned a weight coefficient of 0.2. The pressure domain and acoustic domain are assigned weight coefficients of 0.1 and 0.1, respectively. The nine time-domain statistical features and three frequency band energy proportion features of the vibration domain are combined into a 12-dimensional feature sub-vector, which is multiplied by a weight coefficient of 0.35 to obtain the weighted features. The five time-domain statistical features and one frequency-domain periodic feature of the temperature domain are combined into a 6-dimensional feature sub-vector, which is multiplied by a weight coefficient of 0.25 to obtain the weighted features. The other physical domain subsets are weighted sequentially, and finally all weighted feature vectors are concatenated according to the physical domain order to obtain a 42-dimensional state feature vector, which fully represents the current multi-physical domain coupling state and degradation degree of the device.

[0029] In one optional implementation, a virtual operating environment is constructed based on the state feature vector. By using the current state parameters of the device as initial conditions for simulation, a time-series simulation is performed within the virtual operating environment to obtain a simulation state sequence containing multi-timescale evolution trajectories, including: Based on the preset physical process model library, the physical process models corresponding to each dimension component in the state feature vector are extracted, the physical process models are combined and the basic simulation framework of the virtual running environment is constructed, and the current state parameters of the device are mapped to the basic simulation framework to obtain the initial configuration of the virtual running environment. Based on the time scale characteristics of the equipment's operating conditions, the simulation time domain is divided into multiple time scale levels. A corresponding time step and simulation iteration cycle are assigned to each time scale level. The virtual operating environment is driven at each time scale level to perform state evolution calculations and obtain the evolution trajectory corresponding to each time scale level. Extract the state inflection points in the evolution trajectory at each time scale level, establish cross-scale synchronization constraint relationships based on the position of the state inflection points on the time axis, and perform time alignment and state consistency verification on the evolution trajectory at different time scale levels. The evolution trajectories at each time scale level, after time alignment and state consistency verification, are serialized and integrated to obtain the simulation state sequence.

[0030] The process of constructing a virtual operating environment based on state feature vectors begins with the selection and assembly of physical models. A physical process model library is pre-established, including thermodynamic, fluid dynamic, electromagnetic, and mechanical vibration models, each stored as a discretized state equation. Upon receiving a state feature vector, the system analyzes its various dimensional components, specifically including physical quantities such as temperature, pressure, rotational speed, and current. Assuming the state feature vector contains 12 dimensions, dimensions 1-4 represent the temperature field distribution, dimensions 5-7 represent the pressure field distribution, dimensions 8-10 represent rotational speed variations, and dimensions 11-12 represent current fluctuations. The system retrieves the heat conduction and convection heat transfer models based on the temperature-related dimensional components, the fluid continuity and momentum conservation models based on the pressure-related dimensional components, the rotational inertia and torque balance models based on the rotational speed-related dimensional components, and the Kirchhoff circuit and electromagnetic induction models based on the current-related dimensional components.

[0031] After obtaining the corresponding physical process models, these models are combined according to the actual physical connections of the equipment. Taking a rotating device as an example, the electromagnetic torque generated by its motor drives the rotor to rotate. The rotor rotation causes the fluid medium to flow, which in turn causes pressure changes. These pressure changes lead to changes in temperature distribution, which in turn affects the resistivity of the material and thus the current distribution. Input-output mapping relationships are established between the models. The torque output of the electromagnetic field model is used as the input of the moment of inertia model, the rotational speed output of the moment of inertia model is used as the boundary condition of the fluid dynamics model, the pressure output of the fluid dynamics model is used as the source term of the heat conduction model, and the temperature output of the heat conduction model is fed back to the electromagnetic field model to correct the resistance parameters. This combination of models forms a closed simulation loop, constituting the basic simulation framework of the virtual operating environment.

[0032] The current state parameters of the equipment include real-time measured values ​​of physical quantities such as temperature, pressure, rotational speed, and current. Assuming the currently measured bearing temperature is 85°C, the inlet pressure is 3.2 MPa, the spindle speed is 3000 rpm, and the stator current is 42 A, these measured values ​​are mapped to the corresponding state variable positions in the basic simulation framework. The initial temperature field of the heat conduction model is assigned a spatial distribution of 85°C, the inlet boundary condition of the fluid dynamics model is assigned a pressure value of 3.2 MPa, the initial angular velocity of the moment of inertia model is assigned the value in radians per second corresponding to 3000 rpm, and the stator winding current of the electromagnetic field model is assigned 42 A. In addition to directly measured parameters, a comprehensive initialization configuration of the virtual operating environment is completed based on the equipment's geometric parameters and material property parameters. Geometric parameters include fixed dimensional information such as rotor diameter, number of blades, and bearing spacing, while material property parameters include physical constants such as density, thermal conductivity, and resistivity.

[0033] The multi-scale partitioning of the simulation time domain is based on the characteristic time differences of different physical processes during equipment operation. The characteristic time magnitudes of electromagnetic field changes are milliseconds, rotational speed fluctuations are seconds, temperature field evolution is minutes, and material fatigue accumulation is hours. The system divides the overall simulation time domain into four levels: fast-scale, medium-scale, slow-scale, and ultra-slow-scale. The time step of the fast-scale level is set to 0.001s to capture transient fluctuations in electromagnetic torque and high-frequency oscillations in current, with a simulation iteration cycle of 0.1s. The time step of the medium-scale level is set to 0.1s to track dynamic adjustments in rotational speed and pulsating changes in pressure, with a simulation iteration cycle of 10s. The time step of the slow-scale level is set to 1s to simulate the diffusion process of the temperature field and the cumulative effect of thermal stress, with a simulation iteration cycle of 60s. The time step of the ultra-slow scale level is set to 60s to evaluate the degradation trend of material properties and the long-term drift of system efficiency, and its simulation iteration cycle is 3600s.

[0034] When driving the virtual operating environment at the fast-scale level, the state variables of the electromagnetic field model and the circuit model are updated at 0.001s intervals. Assuming the initial stator current is 42A and the electromagnetic torque is 120N·m, after 0.001s, the current increases to 42.5A due to power supply voltage fluctuations, and the torque increases accordingly to 121.5N·m. 100 iterations are performed continuously, covering an iteration cycle of 0.1s, yielding the current evolution trajectory and torque evolution trajectory containing 100 state points. At the medium-speed scale level, the state variables of the rotational inertia model and the hydrodynamic model are updated at 0.1s intervals. The average torque value calculated at the fast-scale level is used as the driving input for the medium-speed scale level; assuming this average torque is 120.8N·m, combined with the rotor's rotational inertia value of 0.5kg·m... 2 The angular velocity increment after 0.1 s was calculated, and the rotational speed increased from 3000 rpm to 3006 rpm. The fluid dynamics model updated the flow field distribution based on the new rotational speed value, and the outlet pressure changed from the initial 2.8 MPa to 2.85 MPa. The intermediate-speed scale was iterated 100 times over a 10 s period to generate the rotational speed evolution trajectory and the pressure evolution trajectory.

[0035] The slow-scale hierarchy uses the average pressure and average rotational speed values ​​from the medium-speed hierarchy as boundary conditions. Assuming an average pressure of 2.83 MPa and an average rotational speed of 3003 rpm over 10 seconds, the heat transfer model calculates the convective heat transfer between the fluid and the solid wall based on these conditions. The current bearing temperature is 85°C, and the fluid temperature is 70°C. The heat flow driven by the temperature difference causes the bearing temperature to drop by 0.2°C to 84.8°C within 1 second. The slow-scale hierarchy iterates 60 times, covering a 60-second period, and the temperature field gradually approaches a new equilibrium state, with the bearing temperature eventually stabilizing at 83.5°C. The ultra-slow-scale hierarchy assesses material performance degradation based on the average temperature value from the slow-scale hierarchy. Assuming an average bearing temperature of 83°C over 1 hour, the fatigue damage accumulation rate of the material at this temperature is 0.002 h. - ¹The dimensionless damage value is used to update the material property parameters, reducing the elastic modulus by 0.1% from the initial value.

[0036] State inflection points are extracted by monitoring the rate of change of state variables in the evolution trajectory. In the current evolution trajectory at the fast-scale level, when the absolute value of the current change rate exceeds the threshold of 100 A / s, the system marks that moment as a fast inflection point. For example, if the current suddenly jumps from 42.5 A to 45 A at 0.032 s, with a change rate of 2500 A / s, this point is marked as a fast inflection point. In the rotational speed evolution trajectory at the medium-speed scale, when the absolute value of the rotational speed change rate exceeds the threshold of 50 rpm / s, it is marked as a medium-speed inflection point. For example, if the rotational speed rapidly increases from 3006 rpm to 3120 rpm at 3.6 s, this point is marked as a medium-speed inflection point. In the slow-speed scale, locations where the temperature change rate exceeds 0.5°C / s are marked as slow inflection points.

[0037] When establishing cross-scale synchronization constraints, the system uses the timestamp of the fast inflection point as a reference, checking for corresponding state responses at the intermediate-speed and slow-speed scale levels near that timestamp. The fast inflection point occurs at 0.032s. The system searches for state changes between 0.03s and 0.04s at the intermediate-speed scale level, finding a peak torque response at 0.035s, thus determining a causal relationship. The system establishes time synchronization constraints, requiring that the physical effects triggered by the fast inflection point must manifest within a delay of no more than three time steps at the intermediate-speed scale level. For the speed jump at 3.6s at the intermediate-speed inflection point, the system checks the temperature response between 3s and 4s at the slow-speed scale level, finding an upward trend in temperature starting at 3.8s. The time difference between the two is 0.2s, satisfying the physical rationality constraint for cross-scale propagation.

[0038] Time alignment mapping maps evolution trajectories at different timescales onto a unified time axis. The 100 state points at the fast-scale level are distributed between 0 and 0.1 s, the 100 state points at the medium-scale level are distributed between 0 and 10 s, and the 60 state points at the slow-scale level are distributed between 0 and 60 s. The system establishes a time interpolation operator to take values ​​for the fast-scale trajectory at the medium-scale time nodes. At medium-speed time nodes such as 0.1 s and 0.2 s, the corresponding state values ​​are obtained by weighted averaging of adjacent points on the fast trajectory. For the current value at 0.1 s, the system linearly weights the current values ​​at the two nearest points at 0.099 s and 0.101 s in the fast trajectory with weights of 0.5 and 0.5 respectively, to obtain the current aligned value at 0.1 s.

[0039] The state consistency check verifies whether the state values ​​of different scale levels at the same moment meet the physical constraints. At 1 second, the fast-scale level gives a current alignment value of 43A, the medium-scale level gives a rotational speed of 3008 rpm, and the slow-scale level gives a temperature of 84.5°C. The system verifies whether the current value can generate the torque required to maintain the rotational speed based on the relationship between electromagnetic torque and current, and verifies whether the rotational speed can lead to the observed temperature value based on the relationship between rotational speed and heat generation power. Assuming that the torque corresponding to the 43A current is 123 N·m, the mechanical power required to maintain the 3008 rpm rotational speed is 38 kW. The steady-state temperature obtained after heat dissipation calculations from the heat loss generated by the 38 kW power is 84.3°C, which differs from the 84.5°C given by the slow-scale level by 0.2°C. The difference is within the allowable error range of 0.5°C, and the state consistency check passes.

[0040] The serialization and integration process merges the validated evolution trajectories at various scales into a unified simulation state sequence. State points at all scale levels are arranged chronologically, and a complete state vector containing all state variables, including current, torque, speed, pressure, and temperature, is assembled at each time point. At 0.1s, the state vector contains a current of 43.2A and a torque of 124 N·m at the fast-speed scale, a speed of 3006 rpm and a pressure of 2.85 MPa at the medium-speed scale, and a temperature of 84.9°C at the slow-speed scale. At 1s, the state vector is updated to a current of 43A, a torque of 123 N·m, a speed of 3008 rpm, a pressure of 2.87 MPa, and a temperature of 84.5°C. The integrated simulation state sequence, containing state vectors from all time points, forms a continuous multi-dimensional time series data structure. This data structure fully records the evolution of the equipment from its initial state across multiple time scales, providing comprehensive simulation data support for subsequent fault prediction and performance evaluation.

[0041] In one optional implementation, the simulated state sequence is spatiotemporally aligned with the actual measured state sequence of the device. Abnormal state intervals are identified by calculating the deviation metric between the two in the state space. The abnormal state intervals are then traced back to the device's physical structure level based on their corresponding feature components to obtain the anomaly source localization result, including: Based on the timestamp information of the simulated state sequence and the actual measured state sequence of the device, the time offset between the two sequences is identified, and a time axis alignment operation is performed on the two sequences. The simulated state sequence after time axis alignment and the actual measured state sequence of the device are projected onto a preset state space coordinate system to obtain the simulated state trajectory and the measured state trajectory. The point distances at corresponding times in the state space coordinate system are calculated, and a time deviation measurement curve is constructed based on the point distances. The time series deviation measurement curve is subjected to threshold determination to identify the time period in which the deviation measurement exceeds the preset deviation threshold and mark it as an abnormal state interval. Calculate the deviation components of the simulated state trajectory and the measured state trajectory within the abnormal state interval in each dimension of the state space, and identify the dimension with the largest deviation component as the dominant deviation dimension. Based on the feature components corresponding to the dominant deviation dimension, a preset mapping relationship between the feature components and the physical structure level of the device is determined, and the physical components or physical processes corresponding to the feature components are calculated as the anomaly source location results.

[0042] In the specific implementation of spatiotemporal alignment between the simulated state sequence and the actual measured state sequence of the equipment, after acquiring the simulated state sequence and the actual measured state sequence, the timestamp information of the two sequences is extracted. Each state value in the simulated state sequence carries a timestamp. For example, the simulated state sequence of a rotating machine records the vibration amplitude and rotational speed data at times 0s, 0.1s, 0.2s, etc., while the actual measured state sequence records the corresponding parameter data and its timestamp. Since there is a difference between the simulation start time and the actual measurement start time, the time offset is calculated by identifying the time position of characteristic peaks or characteristic abrupt change points in the two sequences. Specifically, the first significant vibration peak is identified in the simulated state sequence at time point 5.3s, while the corresponding vibration peak is identified in the actual measured state sequence at time point 5.8s. Thus, the time offset is determined to be 0.5s. This offset is then subtracted from all timestamps in the simulated state sequence to complete the time axis alignment operation, ensuring that the corresponding events in the two sequences occur at the same time point.

[0043] After aligning the time axis, the aligned simulated state sequence and the actual measured state sequence of the equipment are projected onto a preset state space coordinate system. This state space coordinate system is a multi-dimensional coordinate system, with the number of dimensions corresponding to the number of state characteristic parameters. Taking a power transmission device as an example, the state space coordinate system includes four dimensions: vibration amplitude, rotational speed, temperature, and load current. The system projects the state data of the simulated state sequence at time point 6.0s onto the state space. The coordinates of this state point are vibration amplitude 2.3mm, rotational speed 1450rpm, temperature 68°C, and load current 15.2A, forming a point on the simulated state trajectory. Similarly, the system projects the data of the actual measured state sequence of the equipment at the same time point onto the same state space. The coordinates of this measured state point are vibration amplitude 2.8mm, rotational speed 1445rpm, temperature 72°C, and load current 15.8A. By performing the same projection operation at each corresponding moment, the system obtains the complete simulated state trajectory and the measured state trajectory.

[0044] The system calculates the point distances between the simulated and measured state trajectories at corresponding moments in the state space coordinate system. At time 6.0s, the differences between the simulated and measured points in each dimension are calculated: 0.5mm for vibration amplitude, 5rpm for rotational speed, 4°C for temperature, and 0.6A for load current. To eliminate the influence of different dimensional dimensions, the differences are normalized. The vibration amplitude difference is divided by the characteristic scale of 10mm for that dimension, yielding a normalized value of 0.05; the rotational speed difference is divided by the characteristic scale of 100rpm, yielding a normalized value of 0.05; the temperature difference is divided by the characteristic scale of 20°C, yielding a normalized value of 0.2; and the load current difference is divided by the characteristic scale of 5A, yielding a normalized value of 0.12. The system sums the squares of the normalized differences in each dimension and takes the square root to obtain the point distance at that moment as 0.25. By performing the same calculations at all times, a time series deviation metric curve is constructed. This curve, with time on the horizontal axis and distance between points on the vertical axis, reflects the degree of deviation between the two trajectories over the entire time series.

[0045] A threshold determination is performed on the constructed time-series deviation measurement curve. Based on historical operating data and statistical analysis results, a preset deviation threshold of 0.35 is set. All data points on the time-series deviation measurement curve are traversed to identify time periods where the distance between points exceeds 0.35. In one analysis, the interval from time point 18.5s to time point 32.7s was identified, where the distance between points consistently exceeded the preset deviation threshold, reaching a maximum of 0.68. This interval was marked as an abnormal state interval. Another abnormal state interval was also identified, from time point 45.2s to time point 48.9s, where the maximum distance between points was 0.52. The system records and stores all identified abnormal state intervals for subsequent analysis.

[0046] For the identified abnormal state intervals, the deviation components of the simulated and measured state trajectories in each dimension of the state space were calculated. The abnormal state interval from 18.5s to 32.7s was selected for detailed analysis. Simulated and measured state data for all moments within this interval were extracted, and the deviation in each dimension was calculated. In the vibration amplitude dimension, the difference in vibration amplitude for all moments within the interval was calculated, and the average deviation component was obtained as 0.82mm after taking the absolute value. In the speed dimension, the average deviation component was 12rpm. In the temperature dimension, the average deviation component was 8.5°C. In the load current dimension, the average deviation component was 1.3A. To facilitate comparison of the degree of deviation in different dimensions, the average deviation components in each dimension were normalized. The normalized deviation component for the vibration amplitude dimension was 0.082, for the speed dimension it was 0.12, for the temperature dimension it was 0.425, and for the load current dimension it was 0.26. The normalized deviation component for the temperature dimension was identified as the largest, and the temperature dimension was determined as the dominant deviation dimension.

[0047] Based on the feature components corresponding to the dominant deviation dimension, the mapping relationship between the preset feature components and the physical structure level of the equipment is queried. This mapping relationship is stored in a mapping relationship database, which records the correspondence between each state feature parameter and physical components or physical processes. For the dominant deviation dimension of temperature, the physical structure information associated with this feature component is retrieved from the mapping relationship database. The database records show that the equipment temperature feature component is mainly affected by three physical components: the cooling system, the bearing friction unit, and the lubricating oil circulation loop. Further analysis of the temperature change trend within the abnormal state range reveals that the temperature shows a continuous upward trend within this range, and the heat dissipation rate is significantly lower than the simulation expectation.

[0048] Based on the operating characteristics of the cooling system, the abnormal temperature deviation was determined to be primarily due to a decrease in cooling system efficiency. Calculations of the physical process parameters corresponding to the cooling system within this abnormal state range, including indicators such as cooling medium flow rate and effective utilization rate of radiator surface area, revealed that the cooling medium flow rate decreased by 35% compared to normal values. The specific physical processes of the cooling system and its insufficient cooling medium flow were identified as the anomaly source location, and a detailed report was generated, including the abnormal component identification, the abnormal time interval, and the quantified value of the deviation degree. The report indicated that the cooling system experienced an anomaly between 18.5s and 32.7s, with temperature as the dominant deviation dimension, and a normalized deviation component reaching 0.425. It recommended inspection and maintenance of the cooling medium supply pipeline and flow control valves.

[0049] In one optional implementation, based on the anomaly source location results and combined with equipment operation constraint rules, a diagnostic conclusion and intervention strategy are generated, including: Based on the anomaly source location results, the corresponding equipment operation constraint rules are extracted, the constraint boundary conditions under normal operation are determined, and the deviation items of the state parameters that violate the constraints are identified by combining the actual measurement state sequence of the equipment within the abnormal state interval. Based on the deviation of the state parameters, the failure mode of the anomaly source location result is determined, and the matching fault type is retrieved from the preset fault knowledge base. The fault type is associated with the physical components or physical processes in the anomaly source location result to generate a diagnostic conclusion. Extract a set of candidate intervention actions corresponding to the fault type, evaluate the impact of each candidate intervention action on the equipment operation constraint rules, eliminate candidate intervention actions that cause other constraint boundary conditions to be violated, and select the remaining candidate intervention actions with the highest expected benefits as the intervention strategy.

[0050] After receiving the anomaly source location result, the equipment anomaly diagnosis system immediately extracts all operational constraint rules related to the anomaly source from the equipment operation rule database. Assuming the anomaly source location result points to the cooling pump assembly of the cooling circulation system, the system will read the constraint rule set corresponding to that assembly. This rule set includes the following: the cooling pump speed should be maintained between 1500 rpm and 3000 rpm; the coolant flow rate should be maintained between 800 L / h and 1200 L / h; the cooling pump inlet and outlet pressure difference should be controlled between 0.3 MPa and 0.7 MPa; and the cooling pump motor temperature should not exceed 80°C. The system stores these numerical ranges as constraint boundary conditions under normal operating conditions.

[0051] The system retrieves the actual measurement sequence of the equipment within the abnormal state interval. This sequence records all monitoring data from the start to the end of the abnormality. Using timestamps as indexes, the system compares each actual measurement value with the constraint boundary conditions. In a specific case, the system found that in the first 5 minutes of the abnormal interval, the coolant pump speed remained within the normal range of 2400 rpm, but the coolant flow rate gradually decreased from the normal 1000 L / h to 650 L / h, significantly lower than the lower limit threshold. Simultaneously, the pressure difference between the coolant pump inlet and outlet increased from the normal 0.5 MPa to 0.9 MPa, exceeding the upper limit threshold. The motor temperature continuously climbed from 72°C to 95°C, also violating the temperature constraint. The system marks these three parameters as state parameter deviations, recording them as negative flow rate deviation, positive pressure difference deviation, and positive temperature deviation, respectively.

[0052] The diagnostic engine determines the failure mode based on the combined characteristics of deviations in state parameters. Analysis revealed that the decrease in flow rate accompanied by an increase in differential pressure indicates an abnormal increase in flow resistance, while the increase in motor temperature reflects abnormal heat generation due to increased equipment load. This combination of characteristics is defined as a blockage-type failure mode. The diagnostic engine accesses a pre-set fault knowledge base, which stores various fault types and their characteristic descriptions indexed by failure modes. The fault type with the highest matching degree to the blockage-type failure mode is a partial blockage in the pipeline. Typical characteristics of this fault include a 20% to 40% decrease in flow rate, a 30% to 60% increase in differential pressure, and a 15% to 30% increase in motor temperature. Current actual measurement data shows a 35% decrease in flow rate, an 80% increase in differential pressure, and a 32% increase in temperature, which highly matches this fault type.

[0053] The fault type was correlated with the physical components in the anomaly source location results. The anomaly source location results indicated that the anomaly occurred in the cooling pump assembly of the cooling circulation system. The blockage in the piping could involve specific physical components such as the cooling pump inlet filter, the cooling pump impeller channel, or connecting pipes. Further analysis of the differential pressure change curve revealed that the pressure rise mainly occurred on the cooling pump inlet side. Therefore, the blockage location was precisely pinpointed to the cooling pump inlet filter. The system's diagnostic conclusion was that the inlet filter of the cooling pump assembly in the cooling circulation system was blocked, with a blockage degree of approximately 40%, resulting in insufficient coolant flow, abnormally increased cooling pump load, and motor overheating.

[0054] The intervention strategy generation module extracts a set of candidate intervention actions corresponding to partial pipeline blockage faults from the intervention action knowledge base. This set includes five candidate actions: the first is to immediately shut down the system and clean the inlet filter; the second is to reduce the main operating power of the equipment to decrease cooling demand; the third is to start the standby cooling pump for switching operation; the fourth is to increase the cooling pump speed to increase flow rate; and the fifth is to adjust the coolant ratio to reduce viscosity. The system evaluates the impact of each of these candidate intervention actions on the equipment operating constraints.

[0055] The evaluation program simulated the execution of the first action, calculating that the main system temperature would rise to 110°C during the shutdown, exceeding the 100°C upper limit of the main system temperature constraint. Therefore, this action would violate the main system overheat constraint. The simulation results for the second action showed that reducing the main operating power to 60% of the rated power could reduce the cooling demand to a level that the current flow rate could meet, but would lead to a 40% decrease in production efficiency, affecting production schedule constraints. The feasibility check for the third action found that the standby cooling pump was currently under maintenance and could not be immediately activated, making this action unfeasible. The simulation calculation for the fourth action showed that increasing the speed to 3500 rpm could increase the flow rate, but would further raise the cooling pump motor temperature to 110°C, severely violating the motor temperature constraint and posing a risk of burnout. The impact assessment for the fifth action showed that adjusting the coolant ratio required draining the existing coolant and refilling it, a process that required a 40-minute shutdown, which would also violate the main system temperature constraint.

[0056] The expected benefits of each remaining candidate intervention action are calculated. While the second action impacts production efficiency, it maintains safe equipment operation without shutdown; therefore, its expected benefit is set at 0.6. All other actions are eliminated due to constraint violations. The second action is selected as the immediate intervention strategy, and a planned maintenance recommendation is generated, suggesting that the inlet filter be cleaned during the next planned downtime window to completely resolve the blockage. The intervention strategy output module encapsulates this strategy into executable instructions, including specific parameter settings for adjusting the main operating power to 60% of the rated power, and a task instruction to submit a filter cleaning work order to the maintenance system, ensuring that the equipment can maintain its current safe operation while also receiving fundamental repair at the appropriate time.

[0057] In one optional implementation, based on the actual intervention effect of the intervention strategy, the device status is tracked and collected, and the evolution mechanism parameters in the virtual operating environment are adaptively corrected using the post-intervention status response data, including: After the intervention strategy is executed, the device is continuously monitored to collect multi-dimensional physical state data of the device under the intervention, obtain the state response data after the intervention, and combine it with the historical state data before the intervention to construct a continuous state sequence containing the complete evolution process before and after the intervention. Based on the continuous state sequence, a new state feature vector is generated, and the new state feature vector is input into the virtual operating environment. The state at a certain moment after intervention is used as the simulation starting point to perform forward deduction and backward backtracking. Extract the forward extrapolation trajectory and the backward backtracking trajectory, and calculate the state convergence error at the intervention moment. When the state convergence error exceeds the error threshold, perform sensitivity analysis on the evolution mechanism parameters at each time scale level in the virtual operating environment. The evolution mechanism parameters refer to the key coefficients that control the change of physical state in the virtual operating environment over time. Based on the sensitivity analysis results, the evolution mechanism parameter that has the most significant impact on the state convergence error is determined, and the evolution mechanism parameter is iteratively adjusted in the opposite direction to the state convergence error.

[0058] During the status tracking and parameter correction process after the actual execution of the intervention strategy, continuous monitoring of the equipment's operating status throughout its entire lifecycle is required. Taking a large industrial piece of equipment as an example, after implementing a temperature control intervention strategy, a distributed sensor network is used to collect real-time data on multiple physical parameters of the equipment, including temperature, pressure, vibration frequency, and energy consumption. The sampling frequency is set to 100Hz, and the continuous data collection spans 72 hours after the intervention. The collected data includes the complete temperature change curve from 85°C before the intervention to 62°C after the intervention, the dynamic process of pressure stabilizing from 2.3MPa to 1.8MPa, and the vibration acceleration from 15m / s². 2 Reduced to 8m / s 2 The decay trajectory. These real-time collected data constitute the post-intervention state response dataset, which needs to be spliced ​​and fused with the historical state data of 168 hours before the intervention to form a continuous state sequence with a time span of 240 hours.

[0059] When constructing the continuous state sequence, the time axis is divided into several fixed time windows, each 10 minutes long. Statistical features are extracted for each physical parameter within each window. Taking temperature as an example, within a 10-minute window, the average temperature is extracted to be 73.5°C, the temperature fluctuation range is 2.1°C, the rate of temperature change is -0.3°C / min, and the acceleration of temperature change is reflected in the rate of change itself slowing down at a rate of 0.02°C / min. After performing the same feature extraction operation on all physical parameters, the statistical features of each window are arranged in chronological order, forming a feature sequence matrix containing 1440 time windows. The feature vector corresponding to each window has a 48-dimensional dimension, covering various statistical quantities such as mean, variance, rate of change, peak value, and trough value for four major categories of parameters: temperature, pressure, vibration, and energy consumption.

[0060] After generating the new state feature vector, the equipment state at 36 hours after the intervention was executed was selected as the simulation starting point. At this time, the equipment temperature was 68°C, the pressure was 1.9 MPa, and the vibration acceleration was 10 m / s². 2The state feature vector at that moment is input into the virtual operating environment, which contains multiple subsystems, including a thermodynamic evolution module, a mechanical dynamics evolution module, and an electrical system evolution module. In the thermodynamic evolution module, the thermal conductivity coefficient is set to 0.45, the heat capacity parameter to 1200 J / (kg·°C), and the heat dissipation efficiency coefficient to 0.72. In the mechanical dynamics evolution module, the damping coefficient is set to 0.38, the stiffness coefficient to 5600 N / m, and the inertia coefficient to 450 kg·m. 2 .

[0061] During the forward simulation, starting from the state at time 36, the system simulates the evolution of the equipment state towards future time according to the parameter settings of each evolution mechanism in the virtual operating environment. The simulation time step is set to 1 minute, and the simulation continues for 36 hours until time 72. The predicted state at time 72 obtained from the forward simulation is: temperature 59°C, pressure 1.75 MPa, and vibration acceleration 7.2 m / s². 2 Simultaneously, a backward backtracking operation was performed, starting from the state starting point at 36 hours ago. Following the reverse process of the evolution mechanism, the evolution trajectory of the equipment state back in time was traced, with a backtracking time step of 1 minute, continuing for 36 hours until the initial moment of intervention. The inferred state at the initial moment of intervention obtained from the backward backtracking was: temperature 87°C, pressure 2.35 MPa, and vibration acceleration 15.8 m / s². 2 .

[0062] The forward-engineered trajectory and the backward-following trajectory were compared at the moment of intervention, and the state convergence error at that moment was calculated. The temperature at the intervention moment, calculated from the forward-engineered trajectory, was 83°C, while the temperature at the intervention moment, calculated from the backward-following trajectory, was 87°C, a difference of 4°C. For the pressure parameter, the forward-engineered backward value was 2.25 MPa, and the backward-following forward value was 2.35 MPa, a difference of 0.1 MPa. For the vibration acceleration parameter, the forward-engineered backward value was 14.5 m / s². 2 The backward retrospective forward value is 15.8 m / s 2 The difference is 1.3 m / s 2 After normalizing the differences in each parameter, a comprehensive error index was obtained, yielding a state convergence error value of 0.18. The system's preset error threshold is 0.12. The current convergence error exceeds this threshold, indicating a deviation in the evolution mechanism parameters within the virtual operating environment, requiring adaptive correction.

[0063] Sensitivity analysis was performed on evolution mechanism parameters at different time scales within the virtual operating environment. The time scales were divided into three levels: a microscopic rapid layer, a mesoscopic transition layer, and a macroscopic slow layer. The microscopic rapid layer included 12 evolution parameters ranging from seconds to minutes, such as transient heat flux density factor and instantaneous vibration response coefficient. The mesoscopic transition layer included 18 evolution parameters ranging from minutes to hours, such as temperature diffusion coefficient and pressure balance regulation coefficient. The macroscopic slow layer included 8 evolution parameters ranging from hours to days, such as equipment aging rate coefficient and long-term performance degradation coefficient. A perturbation test was performed on each parameter, increasing the parameter value by 5% from its original value. Forward and backward regression were then re-executed, and the change in state convergence error was recorded. For example, increasing the thermal conductivity coefficient from 0.45 to 0.4725 resulted in a change in state convergence error from 0.18 to 0.15, a change of 0.03. All 38 evolution mechanism parameters were perturbed one by one, and the influence of each parameter on the state convergence error was calculated.

[0064] Sensitivity analysis results show that the temperature diffusivity coefficient of the meso-level transition layer has the most significant impact on the state convergence error; a 5% perturbation of this parameter results in a 0.08 change in error, accounting for 44% of the overall weight. The transient heat flux density factor of the micro-level fast layer accounts for 23% of the overall weight, while the long-term performance degradation coefficient of the macro-level slow layer accounts for 18%. The temperature diffusivity coefficient is determined to be the evolution mechanism parameter that needs priority correction; its current set value is 0.00032m. 2 / s. Based on the reverse direction of the state convergence error, the positive value of 0.18 indicates that the evolution speed of the virtual operating environment is faster than that of the actual device, necessitating a reduction in the temperature diffusion coefficient to slow down the temperature change rate in the virtual environment. An iterative adjustment strategy is adopted, reducing the temperature diffusion coefficient by 3% in each iteration, resulting in an adjusted parameter value of 0.00031m. 2 / s, re-execute the forward derivation and backward backtracking, and calculate the new state convergence error as 0.14. Continue with the second iteration, further adjusting the parameter values ​​to 0.0003m. 2 / s, the state convergence error decreased to 0.10, which is below the error threshold of 0.12, completing the parameter correction process, and the evolution mechanism parameters of the virtual running environment were adaptively optimized.

[0065] The CIR equipment operation status simulation and diagnostic system of this invention includes: The data acquisition unit is used to acquire multi-dimensional physical state data and current state parameters of the target device during the current operating cycle. The vector generation unit is used to perform feature mapping on the multi-dimensional physical state data based on the constraints of the device's operating mechanism and historical degradation patterns, to obtain a state feature vector that reflects the internal coupling relationship of the device. The simulation and deduction unit is used to construct a virtual operating environment based on the state feature vector, and to perform time-series deduction in the virtual operating environment by using the current state parameters of the device as the initial conditions for simulation, so as to obtain a simulation state sequence containing evolution trajectories at multiple time scales. An anomaly localization unit is used to spatiotemporally align the simulated state sequence with the actual measured state sequence of the equipment, identify the abnormal state interval by calculating the deviation metric between the two in the state space, and trace the source to the physical structure level of the equipment based on the feature components corresponding to the abnormal state interval to obtain the anomaly source localization result. The strategy generation unit is used to generate diagnostic conclusions and intervention strategies based on the anomaly source location results and in combination with equipment operation constraint rules. The parameter correction unit is used to track and collect the device status based on the actual intervention effect of the intervention strategy, and to adaptively correct the evolution mechanism parameters in the virtual operating environment using the status response data after the intervention.

[0066] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0067] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0068] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0069] 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 or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for simulating and diagnosing the operating status of CIR equipment, characterized in that, include: Acquire multi-dimensional physical state data and current state parameters of the target device within the current operating cycle; Based on the constraints of equipment operation mechanism and historical degradation patterns, feature mapping is performed on the multi-dimensional physical state data to obtain a state feature vector that reflects the internal coupling relationship of the equipment. A virtual operating environment is constructed based on the state feature vector. By using the current state parameters of the device as the initial conditions for simulation, a time-series simulation is performed in the virtual operating environment to obtain a simulation state sequence containing evolution trajectories at multiple time scales. The simulated state sequence is spatiotemporally aligned with the actual measured state sequence of the device. Abnormal state intervals are identified by calculating the deviation metric between the two in the state space. The abnormal state intervals are then traced back to the physical structure level of the device based on the feature components corresponding to them, thus obtaining the abnormal source location result. Based on the anomaly source location results and combined with equipment operation constraint rules, diagnostic conclusions and intervention strategies are generated. Based on the actual intervention effect of the intervention strategy, the device status is tracked and collected, and the evolution mechanism parameters [1][2] in the virtual operating environment are adaptively corrected using the status response data after intervention.

2. The method according to claim 1, characterized in that, Based on the constraints of equipment operation mechanisms and historical degradation patterns, feature mapping is performed on the multi-dimensional physical state data to obtain state feature vectors reflecting the internal coupling relationships of the equipment, including: Multidimensional physical state data is decomposed into multiple physical domain subsets according to physical domain attributes, and time-domain statistical features and frequency-domain energy distribution features are extracted from each physical domain subset. Based on the constraints of the device operation mechanism, a causal dependency graph between physical domains is constructed, and the interaction path between the sub-data sets of each physical domain is determined according to the causal dependency graph. By calculating the similarity between each physical domain subset and the samples of each degradation stage in the preset degradation trajectory library, the current degradation stage label of the device can be identified. Based on the action path in the causal dependency graph and the degradation stage label, coupling weight coefficients are assigned to each physical domain subset. The time-domain statistical features and frequency-domain energy distribution features of each physical domain subset are weighted and fused according to the coupling weight coefficients to obtain a state feature vector.

3. The method according to claim 1, characterized in that, A virtual operating environment is constructed based on the state feature vector. By using the current state parameters of the device as the initial conditions for simulation, a time-series simulation is performed in the virtual operating environment to obtain a simulation state sequence containing multi-timescale evolution trajectories, including: Based on the preset physical process model library, the physical process models corresponding to each dimension component in the state feature vector are extracted, the physical process models are combined and the basic simulation framework of the virtual running environment is constructed, and the current state parameters of the device are mapped to the basic simulation framework to obtain the initial configuration of the virtual running environment. Based on the time scale characteristics of the equipment's operating conditions, the simulation time domain is divided into multiple time scale levels. A corresponding time step and simulation iteration cycle are assigned to each time scale level. The virtual operating environment is driven at each time scale level to perform state evolution calculations and obtain the evolution trajectory corresponding to each time scale level. Extract the state inflection points in the evolution trajectory at each time scale level, establish cross-scale synchronization constraint relationships based on the position of the state inflection points on the time axis, and perform time alignment and state consistency verification on the evolution trajectory at different time scale levels. The evolution trajectories at each time scale level, after time alignment and state consistency verification, are serialized and integrated to obtain the simulation state sequence.

4. The method according to claim 1, characterized in that, The simulated state sequence is spatiotemporally aligned with the actual measured state sequence of the equipment. Abnormal state intervals are identified by calculating the deviation metric between the two in the state space. Based on the feature components corresponding to the abnormal state intervals, the source is traced back to the physical structure level of the equipment to obtain the anomaly source localization result, including: Based on the timestamp information of the simulated state sequence and the actual measured state sequence of the device, the time offset between the two sequences is identified, and a time axis alignment operation is performed on the two sequences. The simulated state sequence after time axis alignment and the actual measured state sequence of the device are projected onto a preset state space coordinate system to obtain the simulated state trajectory and the measured state trajectory. The point distance between the simulated state trajectory and the measured state trajectory at each corresponding time in the state space coordinate system is calculated, and a time series deviation measurement curve is constructed based on the point distance. The time series deviation measurement curve is subjected to threshold determination to identify the time period in which the deviation measurement exceeds the preset deviation threshold and mark it as an abnormal state interval. Calculate the deviation components of the simulated state trajectory and the measured state trajectory within the abnormal state interval in each dimension of the state space, and identify the dimension with the largest deviation component as the dominant deviation dimension. Based on the feature components corresponding to the dominant deviation dimension, a preset mapping relationship between the feature components and the physical structure level of the device is determined, and the physical components or physical processes corresponding to the feature components are calculated as the anomaly source location results.

5. The method according to claim 1, characterized in that, Based on the anomaly source location results and combined with equipment operation constraint rules, diagnostic conclusions and intervention strategies are generated, including: Based on the anomaly source location results, the corresponding equipment operation constraint rules are extracted, the constraint boundary conditions under normal operation are determined, and the deviation items of the state parameters that violate the constraints are identified by combining the actual measurement state sequence of the equipment within the abnormal state interval. Based on the deviation of the state parameters, the failure mode of the anomaly source location result is determined, and the matching fault type is retrieved from the preset fault knowledge base. The fault type is associated with the physical components or physical processes in the anomaly source location result to generate a diagnostic conclusion. Extract a set of candidate intervention actions corresponding to the fault type, evaluate the impact of each candidate intervention action on the equipment operation constraint rules, eliminate candidate intervention actions that cause other constraint boundary conditions to be violated, and select the remaining candidate intervention actions with the highest expected benefits as the intervention strategy.

6. The method according to claim 1, characterized in that, Based on the actual intervention effect of the intervention strategy, the device status is tracked and collected, and the evolution mechanism parameters in the virtual operating environment are adaptively corrected using the post-intervention status response data, including: After the intervention strategy is executed, the device is continuously monitored to collect multi-dimensional physical state data of the device under the intervention, obtain the state response data after the intervention, and combine it with the historical state data before the intervention to construct a continuous state sequence containing the complete evolution process before and after the intervention. Based on the continuous state sequence, a new state feature vector is generated, and the new state feature vector is input into the virtual operating environment. The state at a certain moment after intervention is used as the simulation starting point to perform forward deduction and backward backtracking. Extract the forward extrapolation trajectory and the backward backtracking trajectory, and calculate the state convergence error at the intervention moment. When the state convergence error exceeds the error threshold, perform sensitivity analysis on the evolution mechanism parameters at each time scale level in the virtual operating environment. The evolution mechanism parameters refer to the key coefficients that control the change of physical state in the virtual operating environment over time. Based on the sensitivity analysis results, the evolution mechanism parameter that has the most significant impact on the state convergence error is determined, and the evolution mechanism parameter is iteratively adjusted in the opposite direction to the state convergence error.

7. A CIR equipment operation status simulation and diagnostic system, used to implement the method as described in any one of claims 1-6, characterized in that, include: The data acquisition unit is used to acquire multi-dimensional physical state data and current state parameters of the target device during the current operating cycle. The vector generation unit is used to perform feature mapping on the multi-dimensional physical state data based on the constraints of the device's operating mechanism and historical degradation patterns, to obtain a state feature vector that reflects the internal coupling relationship of the device. The simulation and deduction unit is used to construct a virtual operating environment based on the state feature vector, and to perform time-series deduction in the virtual operating environment by using the current state parameters of the device as the initial conditions for simulation, so as to obtain a simulation state sequence containing evolution trajectories at multiple time scales. An anomaly localization unit is used to spatiotemporally align the simulated state sequence with the actual measured state sequence of the equipment, identify the abnormal state interval by calculating the deviation metric between the two in the state space, and trace the source to the physical structure level of the equipment based on the feature components corresponding to the abnormal state interval to obtain the anomaly source localization result. The strategy generation unit is used to generate diagnostic conclusions and intervention strategies based on the anomaly source location results and in combination with equipment operation constraint rules. The parameter correction unit is used to track and collect the device status based on the actual intervention effect of the intervention strategy, and to adaptively correct the evolution mechanism parameters in the virtual operating environment using the status response data after the intervention.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images