A device state self-adaptive evaluation method, edge computing device, system and storage medium

Abstract

The application relates to a device state adaptive evaluation method, an edge computing device, a system and a storage medium, the method comprising: acquiring a power source current signal of a target device; analyzing multi-dimensional equivalent physical characteristics representing the running conditions of the target device based on the current signal and constructing a multi-dimensional condition space; monitoring the feature distribution variation degree of the multi-dimensional equivalent physical characteristics in the multi-dimensional condition space; in response to the feature distribution variation degree meeting a preset reconstruction condition, performing reconstruction management on the current device state evaluation framework to obtain an updated evaluation framework; and mapping the current running condition to the multi-dimensional condition space based on the updated evaluation framework and outputting a state evaluation result. The application realizes dynamic adaptive adjustment of the evaluation framework with the change of the working condition, effectively overcomes the defect that the evaluation model is easy to fail under complex working conditions, significantly improves the evaluation accuracy and robustness without frequent manual intervention, and greatly reduces the operation and maintenance cost.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring and data analysis technology for industrial equipment, and in particular to an adaptive assessment method for equipment status, an edge computing device, a system, and a storage medium. Background Technology

[0002] With the increasing level of industrial automation, the demand for monitoring the operating status of power equipment such as motors is growing. Current technologies typically use sensors such as vibration and temperature sensors to collect data and perform equipment health assessments based on preset fixed models or thresholds. However, industrial equipment is diverse and operates under complex and variable conditions, making it difficult for fixed models to cover all scenarios. When equipment operating status shifts or new conditions emerge, the original assessment framework often becomes ineffective, leading to false alarms or missed alarms. Relying on manual recalibration or data filtering to update the model is not only costly and time-consuming but also struggles to handle the real-time processing demands of massive amounts of data. Therefore, there is an urgent need for an adaptive equipment status assessment solution that can adapt to complex operating conditions and automatically adjust the assessment framework without manual intervention. Summary of the Invention

[0003] The technical problem this application aims to solve is that existing equipment condition assessment models are prone to failure under complex and variable operating conditions, and manual maintenance costs are high. To address these shortcomings, this application provides an adaptive equipment condition assessment method, an edge computing device, a system, and a storage medium, enabling adaptive assessment based on current signal analysis of multi-dimensional physical characteristics and dynamic reconstruction of the assessment framework according to data variability.

[0004] To solve the above-mentioned technical problems, the technical solution adopted in this application is: This application provides an adaptive evaluation method for equipment status, characterized in that it includes: Acquire the power source current signal of the target device; Based on the current signal, multidimensional equivalent physical features characterizing the operating conditions of the target equipment are analyzed, and a multidimensional operating condition space is constructed based on the multidimensional equivalent physical features. Monitor the variability of the feature distribution of the multidimensional equivalent physical features in the multidimensional working condition space; In response to the feature distribution variability meeting the preset reconstruction conditions, the current device state assessment framework is reconstructed and reorganized to obtain the updated assessment framework. Based on the updated evaluation framework, the current operating conditions are mapped to a multi-dimensional operating condition space, and the status evaluation results of the target equipment are output.

[0005] Preferably, the step of resolving the multidimensional equivalent physical features characterizing the operating conditions of the target equipment based on the current signal includes: The current signal is decomposed in the time-frequency domain to extract signal component features; The signal component features are mapped to multidimensional physical quantities that characterize the action state of the target device, and these multidimensional physical quantities constitute multidimensional equivalent physical features.

[0006] Preferably, monitoring the variation of the feature distribution of the multidimensional equivalent physical features in the multidimensional working condition space includes: Calculate the distribution difference index between the newly acquired multidimensional equivalent physical features and historical benchmark features in the multidimensional working condition space; When the distribution difference index exceeds the preset threshold, the feature distribution variability is determined to meet the preset reconstruction conditions.

[0007] Preferably, the refactoring governance of the current device state assessment framework includes at least one of the following operations: Update the spatial boundary parameters of the multidimensional working space; Adjust the feature cluster centers within the multidimensional working condition space; Reconstruct the state classification decision model.

[0008] Preferably, acquiring the power source current signal of the target device includes: A current sensor equipped with an electromagnetic shielding layer is used to acquire multi-channel time-synchronized current signals at a preset high-frequency sampling rate. The current sensor acquires electromagnetic signals around the power cable of the target device based on the principle of electromagnetic induction, and filters out environmental interference signals through an electromagnetic shielding layer. The multi-channel time-synchronized current signal contains transient current waveform data of the target device at different operating phases, in order to preserve signal details that reflect the mechanical fluctuations of the target device.

[0009] Preferably, the step of mapping the current operating condition to a multi-dimensional operating condition space based on the updated evaluation framework and outputting the state evaluation result of the target equipment includes: Under the updated evaluation framework, the multi-dimensional working condition space is traversed to statistically analyze the probability of occurrence of anomaly characteristic values ​​of the target equipment within a specific working condition coordinate interval. Based on the probability of occurrence and combined with a pre-set degradation quantification model, the degree of mechanical degradation of the target equipment is calculated. Among them, the coordinate range of a specific working condition is jointly defined by the value range of at least two dimensions of the multidimensional physical quantity, and the anomaly characteristic value of the target equipment includes characteristic parameters that characterize mechanical wear, electrical faults or load abnormalities. The degradation quantification model is configured to calculate and output a continuous degradation degree value by weighted summation based on the difference between the occurrence probability and the preset probability threshold, as well as the weight coefficients corresponding to the coordinate interval of a specific working condition, so as to quantitatively reflect the degradation trend of the health status of the target equipment.

[0010] Preferably, the method further includes: Real-time monitoring of the cumulative status of target equipment's operating data; In response to the cumulative status of running data reaching a preset time length threshold, or in response to the target equipment completing a preset production cycle, a rolling update mechanism for historical status data is automatically triggered. Under the rolling update mechanism, a sliding time window is used to capture the latest cycle of running data and replace the expired data in the historical status data in order to update the historical baseline features. Based on the updated historical benchmark features, the probability of occurrence of anomaly feature values ​​of target equipment within a specific working condition coordinate range is recalculated, and the calculation benchmark for the degree of mechanical deterioration is updated simultaneously. The rolling update mechanism is executed asynchronously in the background without manual intervention, enabling unsupervised adaptive iteration of target device status assessment and eliminating model aging errors caused by long-term operation of the target device.

[0011] This application also provides an edge computing device, including: The data acquisition interface is configured to connect a current sensor to obtain the power source current signal of the target device. Memory, configured to store computer programs; The processor is configured to execute a computer program in the memory to implement the device state adaptive evaluation method as described above.

[0012] This application also provides a device status monitoring system, including: Edge computing devices as described above; And an external system platform that communicates with the edge computing device; The edge computing device is configured to execute an adaptive evaluation method for device status locally and transmit the quantified output of the degree of mechanical degradation and the operation log to an external system platform; The external system platform is configured to receive and display the degree of mechanical degradation and operation logs.

[0013] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the device state adaptive evaluation method as described above.

[0014] The beneficial effects of this application are as follows: By acquiring the power source current signal of the target equipment and analyzing its multi-dimensional equivalent physical characteristics to construct a multi-dimensional operating condition space, the abstract current waveform is transformed into a representation of the equipment's operating state with clear physical meaning, overcoming the deficiency of insufficient information in single-dimensional data. Based on this, by monitoring the variability of feature distribution and automatically triggering the reconstruction and governance of the evaluation framework when conditions are met, the evaluation model can dynamically and adaptively adjust to changes in the actual operating conditions of the equipment. This solves the problem of fixed models easily failing under complex and variable operating conditions, avoids frequent manual recalibration, and significantly reduces operation and maintenance costs. Simultaneously, by combining high-precision current signal acquisition with state mapping in multi-dimensional space, it can capture minute mechanical fluctuations and electrical anomalies that are difficult to detect by traditional vibration or temperature sensors. Furthermore, the degree of mechanical degradation output can be used to intuitively reflect the equipment's health trend, improving the accuracy and timeliness of equipment condition assessment and providing reliable technical support for the intelligent and low-cost operation and maintenance of large-scale industrial equipment groups. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the present application will be further described below in conjunction with the accompanying drawings and embodiments. The drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 This is a flowchart illustrating the device state adaptive evaluation method according to an embodiment of this application. Figure 2 This is a flowchart illustrating step S200 in an embodiment of this application; Figure 3 This is a schematic diagram of the edge computing device and monitoring system according to an embodiment of this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, a clear and complete description will be provided below in conjunction with the technical solutions in the embodiments of this application. Obviously, the described embodiments are some embodiments of this application, but not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this application.

[0017] Embodiment 1 of this application provides an adaptive device status assessment method, mainly applied in edge computing device 5, for realizing intelligent status monitoring of power equipment such as motors. Figure 1 As shown, it includes the following steps: Step S100: Obtain the power source current signal of the target device.

[0018] Specifically, to ensure the accuracy of subsequent analysis, a current sensor 1 equipped with an electromagnetic shielding layer is used to acquire multi-channel time-synchronized current signals at a preset high-frequency sampling rate to read the effective value of the current. The current sensor 1 is typically installed non-destructively on the power cable of the target equipment using a snap-on design. It acquires the electromagnetic signals scattered around the power cable based on the principle of electromagnetic induction and transmits them to the data acquisition module 6 inside the edge computing device 5 via signal cable 2. During this process, the electromagnetic shielding layer effectively filters out complex environmental interference signals in the industrial environment, preventing weak mechanical fluctuation signals from being drowned out by background noise. Simultaneously, the data acquisition module 6 acquires data at a preset high-frequency sampling rate (e.g., not less than 10kHz) and ensures strict synchronization of multi-channel signals on the time axis, thereby completely preserving the transient current waveform data of the target equipment at different operating phases. This high-precision signal acquisition method is a physical prerequisite for subsequent analysis of subtle mechanical characteristics. Without the two hardware conditions of high-frequency synchronization and shielding interference immunity, subsequent algorithm analysis will be unable to effectively identify early signs of mechanical degradation due to data source distortion.

[0019] Step S200: Based on the current signal, analyze the multidimensional equivalent physical features that characterize the operating conditions of the target equipment, and construct a multidimensional operating condition space based on the multidimensional equivalent physical features.

[0020] Specifically, in order to transform abstract current waveforms into physical quantities with clear engineering significance, the original current signal is processed to extract feature components that reflect the operating states of the equipment, such as load, speed, and torque. These components are then combined to form multidimensional equivalent physical features, thereby establishing a multidimensional operating condition space that can cover the normal operating range of the equipment. This enables the mapping from the electrical signal domain to the mechanical physical domain, providing a standardized analytical benchmark for subsequent condition assessment.

[0021] Step S300: Monitor the variation of the feature distribution of multidimensional equivalent physical features in the multidimensional working condition space.

[0022] Specifically, the system calculates in real time the distribution of newly collected data points within the constructed multi-dimensional operating condition space, determining whether they deviate from the clustering area of ​​historical benchmark data. Through a monitoring mechanism that detects changes in equipment operating modes, the variability of the feature distribution increases significantly when equipment operating conditions drift or malfunction, thus providing a quantitative basis for triggering model adjustments.

[0023] Furthermore, this step also includes: calculating the distribution difference index between the newly acquired multidimensional equivalent physical features and historical baseline features in the multidimensional operating condition space; when the distribution difference index exceeds a preset threshold, it is determined that the feature distribution variability meets the preset reconstruction conditions. Since the multidimensional operating condition space is composed of multiple dimensions with different physical dimensions, such as speed, acceleration, and load, simple Euclidean distance often cannot accurately reflect the actual operating condition changes. Therefore, a statistical index that can characterize the distribution characteristics of multidimensional data is used as the distribution difference index to quantify the degree of deviation of the current operating state from the known normal mode. Mahalanobis distance can be used as the distribution difference index. By introducing the covariance matrix of historical baseline features, Mahalanobis distance can effectively eliminate the correlation interference between the dimensions of each physical quantity and automatically normalize the influence of different dimensions, thereby accurately measuring the statistical distance of the newly acquired data points relative to the center of the historical baseline data cluster. For example, when the motor is running under a specific speed and load combination, if the Mahalanobis distance of the newly acquired feature points in the multidimensional space increases significantly, even if the value of its single dimension does not exceed the limit, it may mean that the equipment has entered a certain previously unrecorded coupled operating condition mode. The preset threshold can be dynamically determined based on the statistical distribution characteristics of historical normal operation data (such as the 3σ principle or percentile method), or it can be manually set based on actual operation and maintenance experience. When the calculated distribution difference index continuously or cumulatively exceeds the preset threshold, the system determines that the current characteristic distribution variability has exceeded the normal fluctuation range, meeting the preset reconstruction conditions, thereby triggering the subsequent framework governance process. This monitoring mechanism based on distribution differences enables the system to distinguish between normal operating condition expansion and abnormal equipment failure, avoiding the problem of frequent false alarms when operating conditions change, which is common with traditional fixed threshold methods.

[0024] Step 400: In response to the feature distribution variability meeting the preset reconstruction conditions, perform reconstruction governance on the current equipment state assessment framework to obtain the updated assessment framework.

[0025] Specifically, when the detected variability exceeds a set threshold, it indicates that the original assessment framework is no longer suitable for the current equipment status. At this point, the system will automatically trigger a reconfiguration and governance operation. This operation may involve expanding the boundary of the operating condition space, migrating cluster centers, or correcting the parameters of the decision model, so that the assessment framework can dynamically accommodate new operating condition changes and avoid the problem of frequent false alarms or failures of traditional fixed models after operating condition changes.

[0026] Furthermore, it also includes at least one of the following operations: updating the spatial boundary parameters of the multidimensional working condition space; adjusting the feature cluster centers within the multidimensional working condition space; and reconstructing the state classification decision model. To achieve unsupervised adaptive evaluation, this application automatically incorporates newly emerging working condition patterns into the system's cognitive framework, rather than simply marking them as anomalies.

[0027] In updating the spatial boundary parameters of the multidimensional operating condition space, the minimum bounding box or confidence ellipsoid of the multidimensional space can be dynamically adjusted. For example, when newly acquired feature points are detected to be continuously distributed outside the original spatial boundary and are determined to be valid new operating conditions (rather than transient noise), the system automatically expands the upper and lower bound parameters of the spatial boundary to accommodate the new data distribution range. This expansion can be linear or non-linear based on the new data density, ensuring that the updated evaluation framework covers the entire operating condition range of the actual equipment operation.

[0028] In adjusting feature cluster centers within a multi-dimensional work condition space, incremental clustering algorithms or sliding window mean shift algorithms can be employed. Unlike traditional offline full retraining, cluster center adjustment is performed online in real time. For example, when new work condition data falls on the edge of existing clusters or forms new dense regions, the system only uses the newly added samples to iteratively update the affected local cluster centers, or automatically generates new cluster centers in blank areas. This ensures the model's rapid adaptability to new work conditions while avoiding the huge computational overhead of full retraining, making it particularly suitable for resource-constrained environments of edge computing devices.

[0029] In reconstructing the state classification decision model, the system can learn online how to split nodes in the decision tree, fine-tune neural network weights, or dynamically supplement the rule base. For example, when the original decision boundary cannot effectively distinguish between health states under new and old operating conditions, the system will correct the local structure of the decision model based on newly accumulated labeled data (or semi-supervised data generated through pseudo-labels), adding new discriminative branches or adjusting the splitting thresholds of existing nodes.

[0030] The aforementioned refactoring and governance operations are all automatically triggered processes driven by data variability, requiring no manual intervention for model calibration or retraining. This fundamentally solves the problem of model failure caused by process adjustments, load changes, or component replacements in industrial field equipment, enabling the evaluation framework to dynamically evolve and self-improve throughout the equipment's entire lifecycle.

[0031] Step S500: Based on the updated evaluation framework, the current operating condition is mapped to a multi-dimensional operating condition space, and the status evaluation result of the target device is output.

[0032] Specifically, after dynamically adjusting the framework, the system uses the updated evaluation framework to remap and analyze the current data, outputting a status assessment result that accurately reflects the current health level of the equipment. This result not only includes a judgment on whether the equipment is functioning normally, but also quantitatively characterizes the equipment's degradation trend, providing precise decision support for maintenance personnel.

[0033] Furthermore, this includes: under the updated evaluation framework, traversing a multi-dimensional operating condition space and statistically analyzing the probability of occurrence of anomaly characteristic values ​​of the target equipment within a specific operating condition coordinate interval; based on the occurrence probability and combined with a pre-defined degradation quantification model, calculating the degree of mechanical degradation of the target equipment. After completing the reconstruction and governance of the evaluation framework, the system quantifies the health degradation level of the equipment under the current operating condition through spatial traversal and probability statistics. The specific operating condition coordinate interval is jointly defined by the value ranges of at least two dimensions of multi-dimensional physical quantities, and the anomaly characteristic values ​​of the target equipment include characteristic parameters characterizing mechanical wear, electrical faults, or abnormal loads. For example, in a three-dimensional operating condition space, a specific operating condition coordinate interval can be defined as a three-dimensional region with an equivalent speed of 800-1200 rpm and an equivalent load rate of 70%-90%. Within this region, the system statistically analyzes the probability of occurrence of anomaly characteristic values ​​such as bearing fault characteristic frequency amplitude, stator current harmonic distortion rate, or torque fluctuation variance. By using a statistical method based on multi-dimensional coordinate intervals, the problem of poor applicability of a single threshold under different operating conditions is effectively avoided, ensuring a strong correlation between the evaluation results and the actual operating status of the equipment.

[0034] Furthermore, the degradation quantification model is configured to calculate and output continuous degradation degree values ​​through weighted summation based on the difference between the occurrence probability and a preset probability threshold, and the weight coefficients corresponding to specific operating condition coordinate intervals, in order to quantitatively reflect the health status degradation trend of the target equipment. The mechanical degradation degree D can be calculated using the following formula:

[0035] in, This indicates the degree of mechanical deterioration of the target equipment. Its value is usually a continuous value between 0 and 1, with a larger value indicating a more severe degree of deterioration. Indicates the number of coordinate intervals for a specific operating condition involved in the evaluation; Indicates the first The actual probability of occurrence of anomaly characteristic values ​​of target equipment within a specific working condition coordinate interval; Indicates the first A preset probability threshold corresponding to a specific operating condition coordinate interval is usually derived based on historical benchmark data of the equipment's health status. Indicates the first Weighting coefficients corresponding to specific working condition coordinate intervals. This reflects the differences in the sensitivity of equipment to degradation under different operating conditions. For example, in the high-load operating range, mechanical components experience greater stress, and even minor wear characteristics often indicate a faster rate of degradation. Therefore, the weighting coefficient for this range should be significantly higher than that for the low-load idle operating range. Similarly, for the low-speed steady-state range in precision machining scenarios, the sensitivity to vibration anomalies may be higher than that in the high-speed range, and the corresponding weighting should also be increased. By introducing weighting coefficients with clear physical meaning, the degradation quantification model can aggregate the dispersed anomaly probabilities in multi-dimensional space into a comprehensive indicator that truly reflects the overall health trend of the equipment. By outputting continuous degradation characterization values, rather than the traditional "normal / abnormal" binary classification results, maintenance personnel can intuitively observe the gradual change process of equipment performance. Furthermore, since this evaluation process is performed within the reconstructed new framework, even if the equipment enters an unprecedented new operating condition region, the system can still accurately count the anomaly probability within that region and provide a reasonable degradation score by dynamically adjusting the spatial boundaries and cluster centers, completely solving the problem of fixed models failing after changes in operating conditions.

[0036] Step S600: Monitor the cumulative status of the target equipment's operating data in real time; in response to the cumulative status of the operating data reaching a preset time length threshold, or in response to the target equipment completing a preset production cycle, automatically trigger the rolling update mechanism for historical status data.

[0037] Specifically, to supplement the adaptive capabilities of the time dimension and address the performance degradation of the model caused by natural component aging, gradual environmental changes, or process fine-tuning during long-term equipment operation, a dual-trigger mechanism is employed for monitoring. This ensures that model maintenance needs are met across both the time and event dimensions. The preset time length threshold can be set to a fixed natural time span, such as 30 or 90 days, or to the actual cumulative operating time of the equipment, such as 2000 hours, to address the impact of natural aging or seasonal environmental changes on the baseline data over time. The preset production cycle focuses more on the completeness of operating events, such as completing a batch of product processing, accumulating a certain number of workpieces, or experiencing a complete start-stop cycle. When any of the above conditions are met, the system determines that the current historical baseline features may no longer accurately represent the latest health baseline of the equipment, thus automatically initiating a rolling update process. This avoids the update lag or excessively frequent updates that may result from relying solely on a single time trigger, ensuring that the timing of model iteration matches the actual evolution rhythm of the equipment's condition.

[0038] Furthermore, under the rolling update mechanism, a sliding time window is used to capture the latest cycle's operational data and replace outdated data in the historical state data to update the historical baseline features. The length of the sliding time window can be flexibly configured according to the actual application scenario; it can be set by the number of sample points (e.g., retaining the most recent 100,000 sampling points) or by the time span (e.g., retaining data from the most recent 7 days). During the data replacement process, the system follows the first-in, first-out (FIFO) principle, removing the oldest historical data outside the window and incorporating the most recently collected operational data within the window into the baseline set. This endows the model with the ability to forget outdated information and remember emerging trends. In industrial settings, after equipment undergoes maintenance, component replacement, or process adjustments, its normal operating mode often undergoes permanent changes. If old baseline data is not removed in a timely manner, this outdated information will continue to interfere with the calculation of distribution difference indicators, leading to misjudgments of the new normal by the system. Through the dynamic replacement of the sliding window, the historical baseline features are always maintained within an interval that reflects the recent true state of the equipment, effectively solving the model aging and concept drift problems that fixed baseline models inevitably face during long-term operation.

[0039] Furthermore, based on the updated historical benchmark features, the probability of occurrence of abnormal feature values ​​of the target equipment within a specific operating condition coordinate interval is recalculated, and the calculation benchmark for the degree of mechanical degradation is updated simultaneously to achieve dynamic calibration of the evaluation standard. When the historical benchmark features are updated, the normal data distribution density of each coordinate interval in the multi-dimensional operating condition space changes. Based on this, the system recalculates the probability of occurrence of abnormal feature values ​​(such as bearing failure frequency amplitude, current harmonic distortion rate, etc.) under the new benchmark, and then adjusts the preset probability threshold on which the degradation quantification model relies. and weighting coefficients For example, if the overall vibration level of the equipment increases slightly after a maintenance but remains within the normal range, the updated benchmark will incorporate this new vibration level as normal background and accordingly increase the probability threshold for that operating condition range, avoiding continuous false degradation alarms in subsequent assessments. Because the calculation benchmark is updated synchronously, the numerical value of the mechanical degradation characterization is always measured relative to the equipment's current health zero point, rather than relative to a rigid factory standard, thus significantly improving the reliability and engineering practical value of long-term monitoring results.

[0040] The rolling update mechanism executes asynchronously in the background without manual intervention, enabling unsupervised adaptive iteration for target device status assessment and eliminating model aging errors caused by long-term operation of the target device. This mechanism typically runs as a separate low-priority thread or scheduled task, completely decoupled from the main process responsible for real-time data acquisition, feature parsing, and status assessment. The system only calls upon computing resources to perform data replacement and baseline recalculation operations when processor load is low or the main process is idle, ensuring that background model iteration tasks do not block the real-time monitoring response in the foreground. This achieves unsupervised operation, enabling edge computing devices to rely entirely on their own data to drive model self-evolution and maintenance throughout deployment cycles lasting months or even years. This completely eliminates the high manpower costs associated with periodically exporting data, retraining models offline, and redeploying by maintenance personnel, truly achieving full lifecycle adaptive operation of the device status assessment system.

[0041] In the above embodiments, such as Figure 2 As shown, step S200 above, which involves resolving multi-dimensional equivalent physical features characterizing the operating conditions of the target equipment based on current signals and constructing a multi-dimensional operating condition space based on these features, further includes: Step S201: Decompose the current signal in the time and frequency domain to extract signal component features.

[0042] Specifically, since the motor power source current signal is a non-stationary and complex waveform, directly using the time-domain amplitude is insufficient to characterize the actual operating state of the equipment. Therefore, signal processing algorithms are needed to convert it to the frequency domain or time-frequency domain for analysis. Methods such as Fast Fourier Transform (FFT), Short-Time Fourier Transform (STFT), or wavelet transform can be used to decompose the acquired current signal. Through time-frequency domain decomposition, the system can extract various signal component features, including the fundamental frequency, harmonics, envelope spectrum characteristics, and specific frequency band sidebands. The fundamental frequency typically corresponds to the motor's electrical rotation frequency, harmonic components reflect electromagnetic field distortion, while envelope and sideband components often carry modulation information from mechanical transmission components (such as bearings and gears). These signal component features serve as a bridge connecting electrical signals and mechanical states, providing a rich source of information for subsequent physical quantity mapping.

[0043] Step S202: The signal component features are mapped into multidimensional physical quantities that characterize the action state of the target device. The multidimensional physical quantities constitute multidimensional equivalent physical features.

[0044] Specifically, to establish a deterministic correlation between electrical signal characteristics and mechanical physical quantities, the system has a pre-built mapping model or calibration parameter table based on the electromagnetic-mechanical coupling mechanism of the motor. For AC motors, there is a strict synchronization or slip relationship between the power supply frequency and the rotor speed, mapping the extracted fundamental frequency to the equivalent speed (or velocity) of the target equipment; or changes in load directly cause changes in the current spectrum structure, mapping the amplitude of specific harmonics or total harmonic distortion rate to the equivalent load or torque; or high-frequency components are usually caused by mechanical impact or friction, and the envelope energy or specific sideband amplitude of the high-frequency band can be mapped to the equivalent acceleration or vibration intensity. Through the mapping relationship, the originally abstract current spectrum data is transformed into multi-dimensional physical quantities such as speed, acceleration, and load with clear engineering meaning. These multi-dimensional physical quantities together constitute multi-dimensional equivalent physical features, which not only retain the dynamic change information of the original signal, but also endow the data with intuitive physical semantics.

[0045] Step S203: After completing the analysis of the multidimensional equivalent physical features, the system constructs a multidimensional working condition space based on the multidimensional equivalent physical features.

[0046] Specifically, the multidimensional operating condition space is a vector space spanned by the various multidimensional physical quantities resolved as coordinate axes. For example, a three-dimensional operating condition space can be constructed, with its three coordinate axes representing equivalent velocity, equivalent load, and equivalent acceleration, respectively. In this space, the equipment state at each sampling moment can be represented as a unique operating condition coordinate point. The dimensions of the multidimensional operating condition space are not limited to three dimensions; depending on the actual monitoring requirements and the types of physical quantities resolved, it can also be a two-dimensional, four-dimensional, or even higher-dimensional space. Furthermore, the selection of coordinate axes can be dynamically adjusted; for example, in certain precision machining scenarios, torque ripple rate can be added as a new dimension to improve resolution.

[0047] Through the methods described above, this application constructs an adaptive evaluation system driven by data variation. In particular, the acquisition of high-precision current signals lays a solid data foundation for the entire method, ensuring that subsequent feature analysis and framework reconstruction are based on reliable physical information. This effectively solves the technical challenges of pure data-driven methods being susceptible to noise interference and lacking physical interpretability in industrial settings. The time-frequency domain decomposition and physical quantity mapping mechanism enables cross-domain feature extraction from the electrical signal domain to the mechanical physical domain. This means that the monitoring of feature distribution variation is no longer merely a mathematical calculation of Euclidean distance or statistical distribution differences, but has a clear physical meaning. When the operating condition coordinate points shift significantly in multidimensional space, maintenance personnel can directly interpret this as the equipment entering a specific abnormal operating mode (such as a low-speed, heavy-load, high-vibration region), thus avoiding the problems of poor interpretability and difficulty in troubleshooting false alarms common in pure black-box data-driven models in industrial applications. Simultaneously, the spatial construction based on physical quantities provides a unified standardized benchmark for the condition assessment of different models and specifications of equipment, enhancing the generalization ability and robustness of the evaluation framework.

[0048] Embodiment 2 of this application provides an edge computing device, which serves as the physical execution carrier for the aforementioned device state adaptive evaluation method and integrates data acquisition, storage, and intelligent analysis functions. It includes a data acquisition interface, a memory, and a processor.

[0049] The data acquisition interface is used to connect to the current sensor 1 to obtain the power source current signal of the target device. The data acquisition interface can be integrated inside the edge computing device 5 or connected to the external data acquisition module 6, and it is connected to the external current sensor 1 via the signal cable 2. It should be noted that the data acquisition interface is not merely a simple physical connector, but a functional unit that includes analog-to-digital conversion (ADC), signal conditioning, and timing control circuitry. To support the requirement for high-precision signal acquisition, the data acquisition module 6 has multi-channel time-synchronous acquisition capabilities and a preset high-frequency sampling rate (e.g., not less than 10kHz), which can losslessly convert the analog electromagnetic signal output by the current sensor 1 into a digital signal and transmit it to the back-end analysis unit via the internal transmission cable 7. This hardware-level high-frequency synchronization mechanism is the physical basis for ensuring that subsequent algorithms can analyze the minute mechanical fluctuation characteristics from the current waveform. Without this hardware support, relying solely on software interpolation or low-frequency sampling will not be able to recreate the true transient operating conditions of the device.

[0050] The memory stores the computer program, and the processor executes the computer program in the memory to implement the device state adaptive evaluation method described above. In specific implementation, the processor corresponds to... Figure 3The analysis and processing module 8 can employ one or more combinations of multi-core CPUs, DSPs, FPGAs, or AI acceleration chips to meet the parallel computing power requirements of complex operations such as multi-dimensional working condition space construction, real-time monitoring of feature distribution variability, and dynamic reconstruction of the evaluation framework. For example, when performing Mahalanobis distance calculations or incremental clustering updates, the processor's parallel computing capability can ensure the completion of spatial mapping and statistics of a large number of sample points within milliseconds, avoiding state evaluation lag caused by computational delays. The memory needs to have sufficient capacity and read / write bandwidth to store the program code used to implement the adaptive evaluation logic, as well as cache historical baseline data and running data within the sliding window for the rolling update mechanism. In addition, to ensure the long-term stable operation of the edge computing device 5 in the harsh electromagnetic environment of the industrial site, the device casing is grounded to the internal circuitry via a grounding cable 3 and connected to the earth via an external grounding 4. The chassis is also designed with heat dissipation ducts or heat dissipation fins to maintain the processor and acquisition module within a suitable operating temperature range.

[0051] Embodiment 3 of this application provides a device status monitoring system, which constructs an overall architecture for collaborative operation of edge computing and cloud management. It includes an edge computing device 5 and an external system platform 10 communicatively connected to the edge computing device 5. At the physical connection level, the edge computing device integrates a network communication module 9, which establishes a data link with the external system platform 10 via wired Ethernet, industrial fieldbus, or wireless communication protocols (such as 4G / 5G, Wi-Fi, LoRa, etc.).

[0052] In this system architecture, edge computing device 5 executes the aforementioned adaptive equipment status assessment method locally and transmits the quantified mechanical degradation characterization degree and operation logs to the external system platform 10. Utilizing its built-in processor and analysis processing module 8, the edge computing device independently completes the entire process from high-frequency current signal acquisition, multi-dimensional operating condition space construction, feature distribution variability monitoring to assessment framework reconstruction and governance. This allows the tens of megabytes of high-frequency raw current waveform data per second, which would normally need to be uploaded to the cloud, to be processed at the edge, ultimately sending only the highly compressed and abstracted mechanical degradation characterization degree values ​​and crucial operation logs to the external system platform 10. This data processing mode reduces the uplink data transmission bandwidth requirement by several orders of magnitude, enabling stable system operation even in large-scale equipment monitoring scenarios using low-bandwidth industrial IoT communication solutions, effectively avoiding network congestion and soaring cloud storage costs caused by concurrent uploading of massive amounts of raw data.

[0053] Correspondingly, the external system platform 10 is used to receive and display the degree of mechanical degradation and operation logs. As the centralized management and visualization center of the system, the external system platform no longer undertakes heavy real-time signal processing tasks, but focuses on long-term trend analysis, cross-device comparison, and operation and maintenance decision support. For example, the external system platform can plot the equipment health degradation curve based on the received continuous degradation degree values, and combine it with the framework reconstruction time points in the operation logs to help operation and maintenance personnel determine whether the equipment performance decline is due to natural wear and tear of components or model adaptive adjustment caused by changes in operating conditions. In addition, when the communication link between the edge computing device 5 and the external system platform 10 is temporarily interrupted, the edge computing device also has local caching capabilities, which can temporarily store the evaluation results and logs generated during the interruption in the internal memory, and automatically resume transmission after the communication is restored, thereby ensuring the integrity of monitoring data and the robustness of the system.

[0054] By employing a collaborative architecture that combines edge computing and cloud-based display, the equipment status monitoring system achieves an optimal balance between real-time performance, cost-effectiveness, and scalability. On one hand, by offloading the core adaptive evaluation logic to the edge for execution, millisecond-level status awareness response speeds are ensured, meeting the real-time requirements of rapid early warning of equipment anomalies in industrial settings. On the other hand, by distributing computational load across various edge nodes, the reliance on external system platform computing power and network bandwidth is significantly reduced, enabling the system to scale to monitor hundreds or thousands of devices at extremely low marginal costs. This not only solves the problems of difficult and costly deployment of traditional centralized monitoring solutions in complex industrial environments but also reserves ample system flexibility for the subsequent introduction of more intelligent algorithms and value-added services.

[0055] Embodiment 4 of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned device state adaptive evaluation method.

[0056] Specifically, computer-readable storage media are intended to cover all non-transient media forms capable of persistently or temporarily storing data and being read by computing devices. For example, such storage media may include semiconductor storage devices such as read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), and flash memory; magnetic or optical storage media such as disks (e.g., floppy disks, hard disks), optical discs (e.g., CD-ROMs, DVDs), and magneto-optical discs; and portable storage devices such as USB flash drives, external hard drives, and memory cards.

[0057] The computer program stored on this medium contains a series of instruction codes for controlling the processor to perform specific operations. When these instruction codes are loaded into the edge computing device 5 or other terminals with computing capabilities, the processor will sequentially execute steps such as acquiring the power source current signal, analyzing multi-dimensional equivalent physical characteristics, constructing a multi-dimensional operating condition space, monitoring the variability of feature distribution, triggering the reconstruction and governance of the evaluation framework, and outputting the degree of mechanical degradation characterization. This means that the complex data processing logic and adaptive learning mechanism involved in the aforementioned method can be packaged into independent software products for distribution and deployment.

[0058] By providing a computer-readable storage medium, the technical solution of this application is no longer limited to a specific hardware entity, but can be flexibly delivered in various forms such as software licensing, firmware upgrade packages, or cloud images. In practical applications, maintenance personnel can update the program version in the storage medium to enable existing general-purpose industrial control computers or edge boxes to quickly obtain the latest adaptive evaluation capabilities without replacing hardware devices. This hardware-software decoupling implementation not only significantly reduces the marginal cost of intelligent transformation of large-scale equipment groups, but also facilitates cross-platform migration and continuous iteration of technical solutions, further enhancing the commercial applicability and protection integrity of this application in the field of Industrial Internet of Things.

[0059] It should be understood that this application has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of this application. Furthermore, based on the teachings of this application, these features and embodiments can be modified to suit specific circumstances and materials without departing from the spirit and scope of this application. Therefore, this application is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of this application.

Claims

1. An adaptive equipment status evaluation method, characterized in that, include: Acquire the power source current signal of the target device; Based on the current signal, multidimensional equivalent physical features characterizing the operating conditions of the target equipment are analyzed, and a multidimensional operating condition space is constructed based on the multidimensional equivalent physical features. Monitor the variability of the feature distribution of the multidimensional equivalent physical features in the multidimensional working condition space; In response to the feature distribution variability meeting the preset reconstruction conditions, the current device state assessment framework is reconstructed and reorganized to obtain the updated assessment framework. Based on the updated evaluation framework, the current operating conditions are mapped to a multi-dimensional operating condition space, and the status evaluation results of the target equipment are output.

2. The evaluation method according to claim 1, characterized in that, The process of resolving the multidimensional equivalent physical features characterizing the operating conditions of the target equipment based on the current signal includes: The current signal is decomposed in the time-frequency domain to extract signal component features; The signal component features are mapped to multidimensional physical quantities that characterize the action state of the target device, and these multidimensional physical quantities constitute multidimensional equivalent physical features.

3. The evaluation method according to claim 2, characterized in that, The monitoring of the variation degree of the feature distribution of the multidimensional equivalent physical features in the multidimensional working condition space includes: Calculate the distribution difference index between the newly acquired multidimensional equivalent physical features and historical benchmark features in the multidimensional working condition space; When the distribution difference index exceeds the preset threshold, the feature distribution variability is determined to meet the preset reconstruction conditions.

4. The evaluation method according to claim 3, characterized in that, The refactoring governance of the current device state assessment framework includes at least one of the following operations: Update the spatial boundary parameters of the multidimensional working space; Adjust the feature cluster centers within the multidimensional working condition space; Reconstruct the state classification decision model.

5. The evaluation method according to claim 1, characterized in that, The acquisition of the power source current signal of the target device includes: A current sensor equipped with an electromagnetic shielding layer is used to acquire multi-channel time-synchronized current signals at a preset high-frequency sampling rate. The current sensor acquires electromagnetic signals around the power cable of the target device based on the principle of electromagnetic induction, and filters out environmental interference signals through an electromagnetic shielding layer. The multi-channel time-synchronized current signal contains transient current waveform data of the target device at different operating phases, in order to preserve signal details that reflect the mechanical fluctuations of the target device.

6. The evaluation method according to claim 5, characterized in that, The updated evaluation framework maps the current operating condition to a multi-dimensional operating condition space and outputs the state evaluation results of the target equipment, including: Under the updated evaluation framework, the multi-dimensional working condition space is traversed to statistically analyze the probability of occurrence of anomaly characteristic values ​​of the target equipment within a specific working condition coordinate interval. Based on the probability of occurrence and combined with a pre-set degradation quantification model, the degree of mechanical degradation of the target equipment is calculated. Among them, the coordinate range of a specific working condition is jointly defined by the value range of at least two dimensions of the multidimensional physical quantity, and the anomaly characteristic value of the target equipment includes characteristic parameters that characterize mechanical wear, electrical faults or load abnormalities. The degradation quantification model is configured to calculate and output a continuous degradation degree value by weighted summation based on the difference between the occurrence probability and the preset probability threshold, as well as the weight coefficients corresponding to the coordinate interval of a specific working condition, so as to quantitatively reflect the degradation trend of the health status of the target equipment.

7. The evaluation method according to claim 6, characterized in that, The method further includes: Real-time monitoring of the cumulative status of target equipment's operating data; In response to the cumulative status of running data reaching a preset time length threshold, or in response to the target equipment completing a preset production cycle, a rolling update mechanism for historical status data is automatically triggered. Under the rolling update mechanism, a sliding time window is used to capture the latest cycle of running data and replace the expired data in the historical status data in order to update the historical baseline features. Based on the updated historical benchmark features, the probability of occurrence of anomaly feature values ​​of target equipment within a specific working condition coordinate range is recalculated, and the calculation benchmark for the degree of mechanical deterioration is updated simultaneously. The rolling update mechanism is executed asynchronously in the background without manual intervention, enabling unsupervised adaptive iteration of target device status assessment and eliminating model aging errors caused by long-term operation of the target device.

8. An edge computing device, characterized in that, include: The data acquisition interface is configured to connect a current sensor to obtain the power source current signal of the target device. Memory, configured to store computer programs; A processor configured to execute a computer program in the memory to implement the device state adaptive evaluation method as described in any one of claims 1 to 7.

9. A device status monitoring system, characterized in that, include: The edge computing device as described in claim 8; And an external system platform that communicates with the edge computing device; The edge computing device is configured to execute an adaptive evaluation method for device status locally and transmit the quantified output of the degree of mechanical degradation and the operation log to an external system platform; The external system platform is configured to receive and display the degree of mechanical degradation and operation logs.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the device state adaptive evaluation method as described in any one of claims 1 to 7.

Images