Electromechanical system component identification system based on multi-protocol fusion and network model driving

Abstract

The invention relates to the field of network model driving, and discloses an electromechanical system component identification system based on multi-protocol fusion and network model driving, which comprises the steps of constructing a multi-identification unit management framework through registration, indexing, loading and unloading of execution units, scheduling an identification unit set according to task types and working condition requirements, and identifying the components of the electromechanical system based on the multi-protocol fusion and the network model driving. Carrying out parameter instantiation and structure definition on the identification units, generating model parameters according to historical samples and feature mapping, detecting an input and output response relationship, calculating availability indexes through error analysis, response consistency and a time sequence stability algorithm, screening an identification unit set meeting precision requirements, carrying out multi-dimensional performance calculation, and carrying out multi-dimensional performance calculation. And a multi-working-condition simulation and performance analysis method is adopted to generate a performance index matrix, parameter instantiation and structure definition are performed on the identification units, model parameters are generated according to historical samples and feature mapping, and an updated executable identification unit set is output. The method has the advantage that the self-adaptability and the precision are improved.

Description

Technical Field

[0001] This invention relates to the field of network model-driven systems, specifically to an electromechanical system component identification system based on multi-protocol fusion and network model-driven systems. Background Technology

[0002] In the operation management and health monitoring of electromechanical systems, component identification is a crucial step in ensuring the reliability of system status assessment and fault diagnosis. Existing technologies are mostly based on network model-driven approaches, establishing topology models and signal mappings to identify component categories, operating states, and performance characteristics. They can also integrate data from communication buses such as CAN, Modbus, EtherCAT, and OPCUA to provide unified data support. However, this method still has limitations in complex operating conditions and variable environments. Under conditions such as high temperature and humidity, low temperature start-up, prolonged high load, or strong vibration, electrical and mechanical characteristics change significantly, and signals exhibit nonlinear and time-varying characteristics, making it difficult for fixed models to accurately reflect the system state. For example, the current harmonics of motors differ significantly between light and full loads, hydraulic system oil temperature changes cause pressure hysteresis, and CNC machine tool spindle vibration changes with machining conditions, all of which can easily lead to identification errors. Furthermore, existing models lack online learning and adaptive capabilities, failing to automatically switch to the optimal model based on real-time operating conditions, thus affecting component identification accuracy and response performance. Meanwhile, multi-protocol data vary significantly in format, synchronization, and sampling accuracy. Traditional systems rely on static rule fusion, which can easily lead to feature mismatch or data delays, affecting identification accuracy. Existing models lack online learning and adaptive capabilities, failing to automatically switch to the optimal model based on real-time operating conditions, thus impacting component identification accuracy and response performance. Therefore, it is essential to design an electromechanical system component identification system based on multi-protocol fusion and network model-driven approaches to improve adaptability and accuracy. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an electromechanical system component identification system based on multi-protocol fusion and network model driving, which has the advantages of improved adaptability and accuracy, and solves the problems mentioned in the background technology.

[0004] To achieve the aforementioned goals of improving adaptability and accuracy, this invention provides the following technical solution: an electromechanical system component identification system based on multi-protocol fusion and network model driving, comprising:

[0005] Unit Management Module: By performing unit registration, indexing, loading, and unloading, a multi-identification unit management framework is constructed. Based on modular containers, unit status is maintained, and the identification unit set is scheduled according to task type and working condition requirements.

[0006] Parameter filling module: Receives information from the unit management module, instantiates parameters and defines the structure of the identification unit, generates model parameters based on historical samples and feature mapping, adjusts the structure according to the equipment and working conditions, and outputs the executable identification unit to the unit filtering module.

[0007] Unit filtering module: Receives executable identification units and real-time running data, detects input-output response relationships, calculates availability indicators through error analysis, response consistency and timing stability algorithms, filters the set of identification units that meet the accuracy requirements, and outputs it to the performance evaluation module, while updating the unit management module;

[0008] Performance evaluation module: Receives the identification unit set and historical data, performs multi-dimensional performance calculations, uses multi-condition simulation and performance analysis methods to evaluate identification accuracy and response speed, generates a performance index matrix, selects the optimal identification unit, outputs the evaluation results to the adaptive update module, and writes them back to the unit management module.

[0009] The adaptive update module receives index information from the unit management module, instantiates parameters and defines the structure of the identification units, generates model parameters based on historical samples and feature mappings, and outputs the updated set of executable identification units.

[0010] Preferably, the process of constructing a multi-identification unit management framework is as follows:

[0011] By registering the execution unit, the information of each newly introduced identification unit is entered into the unit database, including the unit identifier, type, functional description and executable path;

[0012] During the indexing phase, an index table that can be quickly retrieved and scheduled is generated based on cell characteristics, historical execution records, and task categories;

[0013] During the loading phase, the specified unit is deployed to the modular container, and the unit parameters and status monitoring interface are initialized.

[0014] During the unloading phase, units that are no longer in use are released from the container based on task completion status or resource scheduling needs, and the index information is updated.

[0015] Preferably, the process of identifying the set of scheduling units based on task type and working condition requirements is as follows:

[0016] Each loaded identification unit is allocated an independent running environment in a modular container, and its status information, including running parameters, the most recent execution result, and availability metrics, is maintained in real time.

[0017] Based on the current task type and operating conditions, a set of candidate units is selected from the index information, and allocated according to priority, availability and historical performance using a scheduling algorithm;

[0018] During task execution, the container monitors the unit status in real time and triggers a rescheduling mechanism when an anomaly or resource conflict is detected.

[0019] Preferably, the process of instantiating parameters and defining the structure of the identification unit is as follows:

[0020] Obtain cell index information from the cell management module, including cell identifier, category, and previously executed parameter set;

[0021] Based on the unit function definition and algorithm requirements, an executable instance is generated for each unit, and the input / output interfaces, control parameters and computational structures are configured into the instance object;

[0022] The parameters are initialized, and preliminary estimates are obtained by fitting historical samples.

[0023] Preferably, the process of generating model parameters based on historical samples and feature mappings is as follows:

[0024] Key features are extracted from historical operating data and experimental samples, including input-output relationships, changes in operating conditions, and response behavior.

[0025] By using feature mapping, historical samples are mapped to the parameter space of the current unit instance to generate the initial parameter set of the model;

[0026] The parameters are fine-tuned using statistical regression or optimization algorithms to form model parameters that can be directly used for identification tasks.

[0027] Preferably, the process of calculating availability metrics using error analysis, response consistency, and time-series stability algorithms is as follows:

[0028] The error is calculated by comparing the output of the instantiated identification unit with the reference historical data, and the deviation is quantified and output.

[0029] Analyze the continuity of output over time, assess response consistency, and calculate time series stability metrics;

[0030] Availability scores are generated by combining error, response consistency, and timing stability, and the scores are fed back to the cell management module to update index information.

[0031] Preferably, the multidimensional performance calculation process is as follows:

[0032] The selected set of identification units is compared and analyzed with historical data to extract multi-dimensional performance indicators, including identification accuracy, response latency, computational efficiency, and adaptability under different working conditions.

[0033] Based on the weighted scores calculated from various performance indicators, the response characteristics of the unit under different input signals, load conditions and environmental changes are statistically analyzed.

[0034] By generating a performance matrix, the accuracy, response time, and robustness of each unit under various operating conditions are fully mapped.

[0035] Preferably, the process of generating the multidimensional performance index matrix is ​​as follows:

[0036] Through multi-condition simulation experiments, response data of the identification unit under different input signal types, amplitude changes, load conditions and environmental parameters were collected, including output signal waveform, response time, maximum deviation and average value under steady state.

[0037] Perform error statistical analysis on the collected response data to calculate the deviation, peak response, and average response time under each operating condition;

[0038] Based on the statistical results and input conditions under different working conditions, a multidimensional performance index matrix is ​​generated.

[0039] Preferably, the process of instantiating parameters and defining the structure of the identification unit is as follows:

[0040] The optimal identification unit selected from the performance index matrix is ​​used as the target, and the corresponding unit management module index information and historical feature data are obtained.

[0041] Based on the unit type and historical parameter distribution, parameter instantiation objects are generated, including input and output interface parameters, gain coefficients, time constants, and control constraints;

[0042] Define the internal structure of the identification unit, clarify the data flow and control relationship between each functional module, and establish the calling interface specification;

[0043] Set the initialization parameters for the unit's operation, and encapsulate the instantiated and structured identification unit into an executable object.

[0044] Preferably, the process of generating the updated set of executable identification units is as follows:

[0045] The set of instantiated identification units and their corresponding index information are obtained from the unit management module. Combined with historical running data, input-output feature mapping and the latest performance matrix, an updated set of model parameters is generated.

[0046] The updated model parameters are applied one by one to the instantiated recognition units to verify the parameter matching and interface specification consistency of each module within the unit.

[0047] Each unit is encapsulated, including initialization of runtime parameter settings, binding of call interfaces, and confirmation of execution logic, forming a complete executable identification unit object;

[0048] All updated executable units are aggregated to generate a set of identification units that can be directly deployed.

[0049] Compared with existing technologies, the present invention provides an electromechanical system component identification system based on multi-protocol fusion and network model driving, which has the following advantages:

[0050] This invention achieves closed-loop management of unit registration, indexing, loading, and unloading by modularizing the identification process of electromechanical system components. It combines historical samples and feature mapping for parameter instantiation and structural definition, enabling the screening, performance evaluation, and adaptive updating of identification units, thus forming a dynamically updatable set of executable units. The system performs multi-dimensional calculations and analyses of unit performance under various operating conditions, achieving optimized selection of identification units and adaptive parameter adjustment. This helps improve the overall identification accuracy of the system and enhances its adaptability under different operating conditions, thereby ensuring the reliability and dynamic response performance of electromechanical system component identification. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0052] 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.

[0053] Example 1: Please refer to Figure 1 As shown in the embodiment of the present invention, the electromechanical system component identification system based on multi-protocol fusion and network model driving includes:

[0054] Unit Management Module: By performing unit registration, indexing, loading and unloading, a multi-identification unit management framework is built. Based on modular containers, the unit status is maintained, and the identification unit set is scheduled according to task type and working condition requirements.

[0055] The process of building a multi-identification unit management framework in the unit management module is as follows:

[0056] By registering the execution unit, the information of each newly introduced identification unit is entered into the unit database, including the unit identifier, type, functional description and executable path;

[0057] When the system receives a new identification unit, it first generates a unique identifier for the unit, including unit number, type classification, and functional description information. At the same time, it records the unit's executable path and dependent file information. When each unit is entered into the database, it records the creation time, version number, and historical execution status. The database table structure includes unit identifier, unit type, functional description, path information, status field, and running record field. The system supports batch registration, ensures data consistency through transaction management, and records operation logs after each registration to ensure that unit information can be completely traced back and loaded when the system restarts or recovers from an anomaly.

[0058] During the indexing phase, an index table that can be quickly retrieved and scheduled is generated based on cell characteristics, historical execution records, and task categories;

[0059] After unit registration is complete, the system generates an index table based on unit characteristics, historical execution records, and task categories for rapid retrieval and scheduling. The index table contains unit identifiers, functional categories, task adaptation information, historical execution performance data, and recent usage time. It uses a hash mapping combined with a tree structure for storage to support multi-dimensional conditional retrieval, such as sorting by task category, precision requirements, and response time. During index table creation, historical execution data is analyzed, and statistics on unit execution success rate, average response latency, and resource consumption are statistically analyzed to generate sorting weights for scheduling, enabling the task allocation module to quickly select units that meet the criteria.

[0060] During the loading phase, the specified unit is deployed to the modular container, and the unit parameters and status monitoring interface are initialized.

[0061] During task scheduling, target units are selected through an index table and deployed into modular containers. The loading process includes copying executable files to the container runtime environment, initializing unit parameters such as input / output interfaces, control parameters, gain coefficients, time constants, and computational constraints, and starting a status monitoring interface to collect unit running status and performance data in real time. Containerized deployment provides runtime isolation, supports parallel execution of multiple unit instances, avoids resource conflicts and interference, and records initialization parameters, dependency library versions, startup time, and resource allocation information during loading, providing a data foundation for subsequent performance analysis and scheduling optimization.

[0062] During the unloading phase, units that are no longer in use are released from the container based on task completion status or resource scheduling needs, and the index information is updated.

[0063] When a task is completed or resource scheduling changes, the system releases unused units from the container, including shutting down running threads, releasing memory and computing resources, and stopping data collection from the status monitoring interface. During the unloading process, the index information of the unit management module is updated, and the unit status is marked as available or idle for the next scheduling call. The system records the timestamp, resource release amount, and final execution status of each unloading operation, and saves the running log and error information for subsequent analysis or model parameter adjustment.

[0064] The process of identifying and scheduling unit sets based on task type and working condition requirements in the unit management module is as follows:

[0065] Each loaded identification unit is allocated an independent running environment in a modular container, and its status information, including running parameters, the most recent execution result, and availability metrics, is maintained in real time.

[0066] In the modular container, each loaded identification unit is allocated an independent running environment, including independent memory space, processing threads, and input / output interfaces. The system collects the running parameters of each unit through a real-time monitoring mechanism, such as sampling period, processing latency, computational load, and memory usage. At the same time, it records the unit's most recent execution result, including input data summary, output data result, and running time. Availability indicators are calculated by statistically analyzing the unit's execution success rate, response latency, and resource usage, and are updated in memory in a time-series manner to provide real-time reference for scheduling.

[0067] Based on the current task type and operating conditions, a set of candidate units is selected from the index information, and allocated according to priority, availability and historical performance using a scheduling algorithm;

[0068] Based on the current task type and operating conditions, the scheduling module extracts a set of matching candidate units from the index information. The index information includes unit type, applicable operating conditions, historical execution performance, and recent call records. The system uses a priority scheduling algorithm to allocate tasks to the most suitable units. The scheduling strategy considers the real-time availability, historical execution accuracy, response time, and resource consumption of the units. During the allocation process, the scheduler calculates the task suitability score for each candidate unit and generates a scheduling list according to the score order to ensure that high-priority tasks get resources first.

[0069] During task execution, the container monitors the unit status in real time and triggers a rescheduling mechanism when an anomaly or resource conflict is detected.

[0070] During the task execution phase, the container monitors the running status of each unit in real time, including processing rate, error codes, abnormal interruptions, and resource conflicts. The system sets thresholds, such as the maximum allowable response latency, memory usage limit, and number of thread conflicts. When a unit exceeds the threshold or an abnormality occurs, a rescheduling mechanism is triggered. The rescheduling process includes reselecting a candidate unit that meets the conditions from the index table, pausing or migrating the current task data, updating the resource allocation within the container, and recording abnormal events and handling steps to provide data support for subsequent task allocation.

[0071] Parameter filling module: Receives information from the unit management module, instantiates parameters and defines the structure of the identification unit, generates model parameters based on historical samples and feature mapping, adjusts the structure according to the equipment and operating conditions, and outputs executable identification units to the unit filtering module.

[0072] The process of instantiating parameters and defining the structure of the identification unit in the parameter filling module is as follows:

[0073] Obtain cell index information from the cell management module, including cell identifier, category, and previously executed parameter set;

[0074] The system first reads the index information of the registered identification units from the unit management module. The index information includes the unique identifier of each unit, the unit category, the function type, and its historical execution parameter set. During the reading process, the system queries the unit database to extract the historical input and output data, control parameters, and last execution results of the unit under different operating conditions, and forms a complete structured data object. This provides basic data support for the next step of parameter instantiation. In practical applications, this index information can be stored in the form of a database table or key-value mapping. Each record contains a timestamp and operating condition identifier to ensure traceability and data integrity.

[0075] Based on the unit function definition and algorithm requirements, an executable instance is generated for each unit, and the input / output interfaces, control parameters and computational structures are configured into the instance object;

[0076] Based on the functional definition and algorithm requirements of each identification unit, the system creates executable instance objects. During the generation process, the input and output interface parameters, control parameters, and internal calculation structures of each unit are configured into the instance one by one. For example, the motor identification unit will be configured with voltage and current input ports, and the output ports will correspond to torque and speed data. At the same time, PID control coefficients, filtering parameters, and sampling periods will be added to the internal structure. During the instance generation process, the system checks the interface consistency to ensure that the data type, signal range, and sampling frequency meet the algorithm requirements. For network model driving units, data flow paths and calling sequences will also be established according to the modular design to form a complete executable object.

[0077] The parameters are initialized, and preliminary estimates are obtained by fitting historical samples.

[0078] When initializing the parameters of the instantiated unit, the system uses historical samples for fitting to generate preliminary estimates. During the fitting process, input and output data under various operating conditions are collected, and statistical regression, least squares, or weighted fitting methods are used to calculate the initial values ​​of control parameters, such as gain coefficient, time constant, and bias. For nonlinear or time-varying units, piecewise linear fitting or window-based moving average fitting methods can be used to generate local parameter estimates from historical data under different load, temperature, or vibration conditions, forming a preliminary parameter vector. This vector is used as the initialization parameter in the instantiated unit to ensure that the unit can complete reasonable calculations and smoothly enter the real-time identification stage during its first execution.

[0079] The process of generating model parameters based on historical samples and feature mappings in the parameter filling module is as follows:

[0080] Key features are extracted from historical operating data and experimental samples, including input-output relationships, changes in operating conditions, and response behavior.

[0081] The system first extracts key features for identification from historical operating data and experimental samples. The data includes the input-output relationship, control signal changes, and system response behavior of each unit under different operating conditions. The input-output relationship includes continuous time series data, such as changes in motor voltage, current, and speed, and pressure and flow signals of the hydraulic system. Operating condition changes record environmental variables, load status, and operating modes. Response behavior includes system delay, transient process, and steady-state characteristics. During the extraction process, the data is organized by timestamp and operating condition label to form a feature matrix, ensuring that each sample data can be accurately mapped to a specific unit instance. Data preprocessing includes outlier removal, noise filtering, and normalization to ensure the accuracy and consistency of the features.

[0082] By using feature mapping, historical samples are mapped to the parameter space of the current unit instance to generate the initial parameter set of the model;

[0083] The extracted historical features are transformed into the parameter space of the current unit instance through feature mapping methods, establishing the correspondence between input and output features and model parameters. For example, for the motor identification unit, the current response curve and speed curve are mapped to the unit's gain coefficient, time constant, and filter coefficient; the hydraulic system pressure response is mapped to the elastic coefficient, damping parameter, and pipeline delay. During the mapping process, matrix operations and interpolation methods are used to align the historical feature samples with the initial structure of the current unit, generating the initial parameter set of the model to ensure that the parameter range is consistent with the unit design constraints.

[0084] The parameters are fine-tuned using statistical regression or optimization algorithms to form model parameters that can be directly used for identification tasks;

[0085] The generated initial parameter set is further fine-tuned through statistical regression or optimization algorithms to adapt to the specific instance characteristics of the current unit. Regression methods include least squares fitting, multiple linear regression, or weighted regression, which fit model parameters to historical input and output data. Optimization methods can employ gradient descent, Newton iteration, or genetic algorithms to perform global search and local fine-tuning of nonlinear or multimodal characteristic parameters. During the fine-tuning process, the system calculates the output error after each round of parameter adjustment and stops iteration based on the error convergence. Finally, a set of model parameters that can be directly used for real-time identification tasks is generated. These parameters include input-output mapping, key control coefficients, and execution weights, which can be directly called in the unit instance.

[0086] Unit filtering module: Receives executable identification units and real-time running data, detects input-output response relationships, calculates availability indicators through error analysis, response consistency and timing stability algorithms, filters the identification unit set that meets the accuracy requirements, and outputs it to the performance evaluation module, while updating the unit management module.

[0087] The process of calculating availability metrics in the unit selection module using error analysis, response consistency, and time-series stability algorithms is as follows:

[0088] The error is calculated by comparing the output of the instantiated identification unit with the reference historical data, and the deviation is quantified and output.

[0089] The output signal of the instantiated identification unit under the current operating condition is compared point by point with the corresponding reference data in the historical records. The output deviation is quantified by calculating the absolute error, relative error and mean square error of each sampling point. The historical reference data includes typical output curves under different load, temperature and vibration conditions. The sampling frequency is consistent with the current unit output to ensure that the error calculation is completed at the same time resolution. The error calculation process is implemented in the software layer through matrix operations. Each input channel is processed separately to generate an error vector for subsequent analysis.

[0090] Analyze the continuity of output over time, assess response consistency, and calculate time series stability metrics;

[0091] A continuous analysis is performed on the time-series signal output by the identification unit. By calculating the correlation between the first derivative of the output signal and the derivatives of historical data, the consistency of the unit's response over continuous time is evaluated. Weights are assigned to rapidly changing and slowly drifting signals. Sliding window averaging and median filtering are used to process abnormal fluctuations to reduce the impact of occasional noise. During the analysis, the output response time, amplitude deviation, and phase delay are recorded after each input change. The time-series stability indices, including the mean response delay, delay variance, and number of overshoots, are calculated using the autocorrelation function, power spectral density, and hysteresis characteristics of the output signal. By comparing the standard time-series response of similar units under historical operating conditions, it is determined whether the current unit exhibits unstable oscillations or abnormal delays. This process combines time-series statistical analysis and frequency domain analysis methods to independently calculate the time-series stability matrix for each key output channel.

[0092] Availability scores are generated by combining error, response consistency, and timing stability, and the scores are fed back to the cell management module to update index information.

[0093] Error, response consistency, and timing stability results are combined according to preset weights to generate a comprehensive availability score. The score is recorded in numerical form and associated with the unit identifier, task category, and execution time to form a traceable record. The generated score results are simultaneously fed back to the unit management module. By updating the unit index information, including historical performance records, recent availability scores, and status identifiers, the unit management database is updated in real time, providing data support for scheduling and subsequent task selection.

[0094] Performance evaluation module: Receives the identification unit set and historical data, performs multi-dimensional performance calculations, uses multi-condition simulation and performance analysis methods to evaluate identification accuracy and response speed, generates a multi-dimensional performance index matrix, selects the optimal identification unit, outputs the evaluation results to the adaptive update module, and writes them back to the unit management module.

[0095] The multidimensional performance calculation process in the performance evaluation module is as follows:

[0096] The selected set of identification units is compared and analyzed with historical data to extract multi-dimensional performance indicators, including identification accuracy, response latency, computational efficiency, and adaptability under different working conditions.

[0097] The selected identification unit set is aligned with historical data, and multi-dimensional performance indicators are extracted. For each unit to be evaluated and each predefined working condition, the real-time output sequence is first aligned with the historical reference sequence by timestamp. If the sampling frequency is inconsistent, resampling is performed uniformly. The resampling method adopts linear interpolation or sample-preserving method. The default sampling frequency is 100 samples per second or the value configured according to the task. After alignment, the absolute error sequence is calculated point by point, and the root mean square error and normalized root mean square error are calculated based on the sequence. The normalization scale uses the mean of the historical samples of the unit under the same working condition as the divisor. The response delay is determined by detecting the appearance of a clear input. The difference between the point of change and the point when the output reaches 90% of its steady state is measured. The latency is statistically analyzed in milliseconds, and the mean and standard deviation are output. The computational efficiency is quantified by recording the total processing time, average single-sample processing time, and average number of samples processed per second for each identification task. At the same time, the peak processor utilization and memory usage during runtime are collected as resource utilization indicators. For the adaptability under different working conditions, the aforementioned indicators are calculated separately on each working condition sample set and stored with the working condition identifier as the index. All raw calculation results are written to a temporary performance table according to the unit identifier, working condition identifier, and time window for subsequent summarization.

[0098] Based on the weighted scores calculated from various performance indicators, the response characteristics of the unit under different input signals, load conditions and environmental changes are statistically analyzed.

[0099] The raw indicators obtained in the first step are first normalized by interval. The normalization interval is determined based on the 5th percentile and 95th percentile of the historical samples. Values ​​outside the interval are truncated at the boundary and recorded as outliers. After normalization, each indicator is linearly weighted and summed according to the weight vector defined in the configuration file to generate a comprehensive score of the unit under a single working condition. The source and effective time of the weight configuration are recorded together. For multiple score sequences under the same working condition, the sample size, arithmetic mean, variance, median, and confidence interval estimated by the bootstrap sampling method are calculated. When estimating the confidence interval, the number of samplings and the confidence level are recorded. When processing samples with different input signal types and different load levels, the samples are grouped according to input characteristics and load levels. The score distribution within each group is calculated and the group statistics table is output. The differences between groups can be quantified by the analysis of variance method and the test results and significance thresholds are recorded.

[0100] By generating a performance matrix, the accuracy, response time, and robustness of each unit under various operating conditions are fully mapped.

[0101] Based on the comprehensive score and statistics from the second step, a performance index matrix is ​​constructed for all candidate units. Rows in the matrix correspond to unit identifiers, and columns correspond to predefined operating condition sets and performance dimensions. Each unit's operating condition cross-cell contains fields, including the normalized comprehensive score, the original root mean square error, the mean response delay, the average single-sample processing time, and the anomaly trigger count. The performance matrix is ​​stored in a relational database table format. The database table fields include unit identifier, operating condition identifier, statistical time window start time, statistical time window end time, original index values, normalized values, and comprehensive score. The matrix is ​​also exported as a comma-separated file for offline analysis. After the matrix is ​​generated, an integrity check is performed. The check items include cell fill rate checks, index value range checks, and timestamp continuity checks. The check results, along with the metadata used for calculation, are written to the log. The metadata includes the sampling frequency, statistical time window length, and weight configuration version, so that downstream adaptive update and scheduling modules can perform subsequent processing based on this matrix.

[0102] The process of generating a multidimensional performance index matrix in the performance evaluation module is as follows:

[0103] Through multi-condition simulation experiments, response data of the identification unit under different input signal types, amplitude changes, load conditions and environmental parameters were collected, including output signal waveform, response time, maximum deviation and average value under steady state.

[0104] Based on a predefined set of operating conditions, a multi-condition simulation experiment is performed on each identification unit. The operating conditions include three types of input signals: step signals, sinusoidal signals, and random disturbance signals. Input amplitudes are tested in stages from 50% to 150% of the nominal value. Load conditions consist of several levels from no-load to rated load. Environmental parameters cover the temperature range of 0°C to 50°C and the relative humidity range of 20% to 90%. The experiment duration for each operating condition is typically set to 60 seconds, and the sampling frequency is set to 100 Hz by default. The collected data includes the original output signal waveform, input signal waveform, timestamp, and temperature and humidity values ​​recorded synchronously during operation. The experiment is repeated at least ten times to obtain statistical samples. Data acquisition uses a time synchronization mechanism to ensure input and output alignment. The acquisition equipment needs to record sampling accuracy and clock deviation. All raw data is written into the raw data table using the unit identifier, operating condition identifier, and test number as indexes for subsequent batch processing.

[0105] Perform error statistical analysis on the collected response data to calculate the deviation, peak response, and average response time under each operating condition;

[0106] The acquired raw waveforms are preprocessed, including removing DC bias, low-pass filtering to remove high-frequency noise exceeding the bandwidth, and outlier removal. Finite impulse response filters are used, and the filter order and cutoff frequency parameters are recorded. Then, key statistics are calculated within each test window, mainly including maximum deviation, root mean square error, peak response, and steady-state average. The response delay is calculated from the moment when the input changes significantly to the time required for the output to reach 90% of its steady state. The steady-state period is defined as the last ten seconds after the input stops changing. The steady-state average is the arithmetic mean of the samples within this period. The computational efficiency is quantified by recording the data processing time and processor usage percentage for each test. All statistical results include the start and end times of the time window and the test number, and are written into an error statistics table for summary analysis.

[0107] Based on the statistical results and input conditions under different working conditions, a multidimensional performance index matrix is ​​generated.

[0108] The statistics obtained in the second step are summarized according to the unit identifier and operating condition identifier to construct a performance index matrix. Each row of the matrix corresponds to an identification unit, and each column corresponds to a predefined operating condition and performance dimension intersection item. The intersection cell contains field items, including the original value of the maximum deviation, the original value of the root mean square error, the mean response delay, the steady-state average value, the average single sample processing time, and the number of test samples. The performance matrix is ​​persistently stored in the form of a relational database table. The table structure includes the unit identifier, operating condition identifier, statistical time window start time, statistical time window end time, original index values, and sample counts. The metadata used during the calculation is recorded during matrix generation, including sampling frequency, filtering parameters, steady-state judgment rules, and statistical window size. After generation, integrity verification is performed. Verification items include cell fill rate, index value boundary check, and time window consistency. The verification results are written to the log along with the matrix file and archived for subsequent use.

[0109] The adaptive update module receives index information from the unit management module, instantiates parameters and defines the structure of the identification units, generates model parameters based on historical samples and feature mappings, and generates an updated set of executable identification units.

[0110] The process of parameter instantiation and structure definition of the identification unit in the adaptive update module is as follows:

[0111] The optimal identification unit selected from the performance index matrix is ​​used as the target, and the corresponding unit management module index information and historical feature data are obtained.

[0112] The system selects the optimal identification unit that meets the various performance standards from the performance index matrix and uses it as the current instantiation target. The system queries the characteristic data of the unit in the historical task execution through the unit management module, including the input and output signal types, signal amplitude range, processing delay, computational consumption and previous parameter distribution. The data is stored in two forms: time series and discrete parameters, with an accompanying operating condition label, to ensure that subsequent parameter instantiation can be based on complete historical data.

[0113] Based on the unit type and historical parameter distribution, parameter instantiation objects are generated, including input and output interface parameters, gain coefficients, time constants, and control constraints;

[0114] Based on the unit type and historical parameter distribution, the system generates parameter instantiation objects for each identification unit. The objects include input interface parameters and output interface parameters, and clearly define the signal type, voltage or current amplitude range, sampling frequency and data format of each interface. The gain coefficient and time constant parameters are obtained through statistical analysis and regression calculation based on historical operating data. Control constraints include maximum allowable delay, minimum sampling period and power or load limits. The generation process is implemented through scripted templates, so that parameter objects of different types of units can be automatically created and directly used for simulation and actual operation.

[0115] Define the internal structure of the identification unit, clarify the data flow and control relationship between each functional module, and establish the calling interface specification;

[0116] The system defines the internal structure of each unit, including functional module division, data flow path and control logic. The input and output interfaces, calculation order and processing priority of each module are clearly recorded. The calling interfaces between modules are connected through standardized communication protocols to ensure accurate data transmission and synchronous execution between modules. The structure definition is stored using graphical data structures and matrix description. At the same time, the system records the system resource requirements of each module, including memory usage, processing cycle and thread priority, to achieve structured management and dynamic scheduling.

[0117] Set the initialization parameters for the unit's operation, and encapsulate the instantiated and structured identification unit into an executable object;

[0118] For the instantiated and structured identification unit, the system initializes the operating parameters, including the initial input signal state, timer reset, cache clearing and control flag setting. All parameters and structural information are encapsulated into an executable object, which can be directly loaded into the modular runtime container for execution. During the encapsulation process, the system generates object identifier, version number and execution path, enabling the unit to be quickly deployed under different tasks and working conditions, while ensuring compatibility with the unit management module and scheduling system, and realizing standardized calling and operation management.

[0119] The process of generating the updated set of executable identification units in the adaptive update module is as follows:

[0120] The set of instantiated identification units and their corresponding index information are obtained from the unit management module. Combined with historical running data, input-output feature mapping and the latest performance matrix, an updated set of model parameters is generated.

[0121] The system reads the set of instantiated identification units from the unit management module, and simultaneously obtains the index information of each unit, including unit identifier, type, previously used parameter set, execution history, and operating condition records. Combining historical operating data and input / output feature mapping, the system analyzes the unit's response under different signal inputs, load changes, and environmental conditions. The system also compares the newly generated performance matrix with historical data to identify the performance deviations and potential parameter adjustment space of each unit under various operating conditions, providing data support for generating updated model parameters. All data is stored in time series and multidimensional matrix form and corresponds one-to-one with unit identifiers. Based on the collected historical features, input / output mapping, and performance matrix information, the system calculates a new set of model parameters for each identification unit. The calculation process includes parameter regression analysis, weighted fitting, and multi-condition optimization to determine the optimal values ​​of input / output interface parameters, gain coefficients, time constants, and control constraints. Statistical analysis is performed on the response data under different operating conditions. By calculating the mean, variance, and extreme values, the parameters are adjusted to ensure stable operation throughout the entire operating range. The generated parameter set is stored in an object-oriented manner, allowing each unit to be directly applied to the instantiated object.

[0122] The updated model parameters are applied one by one to the instantiated recognition units to verify the parameter matching and interface specification consistency of each module within the unit.

[0123] The system loads the updated model parameters one by one into the instantiated identification units. At the same time, it performs parameter matching verification on each module within the unit to ensure that the input / output interfaces, control logic, and calculation structure are consistent with the new parameters. During the verification process, the system checks the signal channel matching, whether the parameter value range exceeds the limit, and whether the module calling order is consistent with the interface specification. Any mismatch or potential conflict is automatically marked and logged for the system to correct or regenerate parameters. This process ensures that each unit can maintain the integrity of its internal structure and the standardization of its operation after the parameters are updated.

[0124] Each unit is encapsulated, including initialization of runtime parameter settings, binding of call interfaces, and confirmation of execution logic, forming a complete executable identification unit object;

[0125] For each unit that has undergone parameter updates and consistency checks, the system sets initial operating parameters, including input signal reset, timer clearing, cache clearing, and control flag configuration. The internal structure, parameter set, and calling interface of the unit are bound into a complete executable object, and an object identifier, version number, and execution path are generated. The encapsulated unit can run directly in a modular container and supports interface communication with the scheduling system and unit management module to achieve rapid deployment and operation monitoring.

[0126] All updated executable units are aggregated to generate a set of identification units that can be directly deployed;

[0127] The system aggregates all updated executable identification units into a complete unit set. The set information includes the identifier, parameter version, execution path, and working condition adaptation record of each unit. The set can be directly loaded into the runtime environment or scheduling system to achieve unified management and batch deployment. The system also records update logs and parameter adjustment history to provide data support for subsequent performance evaluation and adaptive optimization, ensuring that each unit maintains parameter consistency and executability under different tasks and working conditions.

[0128] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0129] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for identification of components of an electromechanical system based on multi-protocol fusion and network model driving, characterized in that, Comprise: Unit management module: by executing unit registration, indexing, loading and unloading, build multi-recognition unit management framework, maintain unit state based on modular container, and dispatch recognition unit set according to task type and working condition demand; Parameter filling module: receive unit management module information, parameter instantiation and structure definition for recognition unit set, generate model parameters according to historical samples and feature mapping, adjust structure according to equipment and working condition, output executable recognition unit to unit screening module; Unit screening module: receive executable recognition unit and real-time running data, detect input-output response relationship, calculate availability index through error analysis, response consistency and timing stability algorithm, screen recognition unit set meeting accuracy requirement, and output to performance evaluation module, while updating unit management module; Performance evaluation module: receive recognition unit set and historical data, perform multi-dimensional performance calculation, evaluate recognition accuracy and response speed by using multi-working condition simulation and performance analysis method, generate multi-dimensional performance index matrix, select optimal recognition unit, and output evaluation result to adaptive update module, and write back to unit management module; Adaptive update module: receive indexing information of unit management module, parameter instantiation and structure definition for recognition unit, and generate updated executable recognition unit set according to historical samples and feature mapping.

2. The multi-protocol converged and network model driven electromechanical system component recognition system based on claim 1, wherein, The process of building multi-recognition unit management framework is: By executing unit registration, the information of each newly introduced recognition unit is entered into the unit database, including unit identification, type, function description and executable path; In the indexing stage, generate index table that can be quickly searched and scheduled according to unit characteristics, historical execution records and task categories; In the loading stage, deploy the specified unit to the modular container, and initialize the unit parameters and state monitoring interface; In the unloading stage, release the unit from the container that is no longer used according to the task completion or resource scheduling demand, and update the index information.

3. The multi-protocol fusion and network model driven electromechanical system component recognition system of claim 2, wherein, The process of dispatching recognition unit set according to task type and working condition demand is: In the modular container, allocate independent running environment for each loaded recognition unit, and maintain its state information in real time, including running parameters, the latest execution result and availability index; According to the current task type and working condition, select candidate unit set from the index information, and allocate according to priority, availability and historical performance by using scheduling algorithm; During task execution, the container monitors the unit state in real time, and triggers the rescheduling mechanism when detecting abnormality or resource conflict.

4. The multi-protocol fusion and network model driven electromechanical system component recognition system of claim 3, wherein, The process of parameter instantiation and structure definition for recognition unit is: Get unit indexing information from unit management module, including unit identification, category and previously executed parameter set; According to unit function definition and algorithm requirement, generate executable instance for each unit, and configure input-output interface, control parameter and calculation structure to instance object; Initialize parameters, and get preliminary estimated value by fitting historical samples.

5. The multi-protocol fusion and network model driven based electromechanical system component recognition system of claim 4, wherein, The process of generating model parameters according to historical samples and feature mapping is: Extract key features from historical running data and experimental samples, including input-output relationship, working condition change and response behavior; The historical samples are mapped to the parameter space of the current unit instance by a feature mapping method to generate an initial parameter set of the model; The parameters are fine-tuned by statistical regression or optimization algorithms to form model parameters that can be directly used for identification tasks.

6. The multi-protocol fusion and network model driven based electromechanical system component recognition system of claim 5, wherein, The availability index process is calculated by error analysis, response consistency, and timing stability algorithms: Calculate the error between the output of the instantiated identification unit and the reference historical data to quantify the output deviation. Analyze the continuity of the output over time, evaluate the response consistency, and calculate the timing stability index. Combine the error, response consistency, and timing stability to generate an availability score, and feed the score back to the unit management module to update the index information.

7. The multi-protocol fusion and network model driven electromechanical system component recognition system of claim 6, wherein, The multi-dimensional performance calculation process is: Compare the selected identification unit set with the historical data to extract multi-dimensional performance indicators, including identification accuracy, response delay, computational efficiency, and adaptability under different working conditions. Based on the weighted score of each performance indicator, statistical analysis is performed on the response characteristics of the unit under different input signals, load conditions, and environmental changes. Generate a performance matrix to map the accuracy, response time, and robustness of each unit under each working condition.

8. The multi-protocol fusion and network model driven electromechanical system component recognition system of claim 7, wherein, The multi-dimensional performance indicator matrix generation process is: Collect response data of the identification unit under different input signal types, amplitude changes, load conditions, and environmental parameters through multi-condition simulation experiments, including output signal waveform, response time, maximum deviation, and average value in stable state. Perform error statistical analysis on the collected response data to calculate the deviation, peak response, and average response time under each working condition. Based on the statistical results and input conditions under different working conditions, generate a multi-dimensional performance indicator matrix.

9. The multi-protocol fusion and network model driven based electromechanical system component recognition system of claim 8, wherein, The parameter instantiation and structure definition process for the identification unit is: Select the optimal identification unit from the performance indicator matrix as the target, and obtain the corresponding unit management module index information and historical feature data. Based on the unit type and historical parameter distribution, generate parameter instantiation objects, including input / output interface parameters, gain coefficients, time constants, and control constraints. Define the internal structure of the identification unit to clarify the data flow and control relationship between the functional modules, and establish the calling interface specification. Set the running initialization parameters of the unit, and encapsulate the instantiated and structured identification unit as an executable object.

10. The multi-protocol fusion and network model driven based electromechanical system component recognition system of claim 9, wherein, The process of generating an updated executable identification unit set is: Obtain the instantiated identification unit set and corresponding index information from the unit management module, and generate an updated model parameter set based on historical running data, input / output feature mapping, and the latest performance matrix. Apply the updated model parameters to the instantiated identification unit one by one to verify the parameter matching and interface specification consistency of each module within the unit. Perform encapsulation processing for each unit, including initialization and running parameter setting, calling interface binding, and execution logic confirmation, to form a complete executable identification unit object. Summarize all updated executable units to generate a directly deployable identification unit set.

Images