A system and method for assessing equipment condition

By introducing an asynchronous decoupling architecture using message queues and adaptive preprocessing based on operating conditions, the problems of link coupling and data inconsistency in industrial equipment condition monitoring were solved. Stable assessment and maintenance support were achieved in scenarios such as high-frequency vibration waveforms, improving the availability and accuracy of the system.

CN123114906APending Publication Date: 2026-11-17ANHUI ZHIYUAN INTELLIGENT TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611118349.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-27
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies for industrial equipment condition monitoring suffer from problems such as severe coupling between data acquisition, inference, and display links, inconsistent raw data processing, insufficient stability of evaluation results, lack of reliable feedback mechanisms and operation and maintenance support, making it difficult to operate stably for a long time in scenarios such as high-frequency vibration waveforms and sound data.

Method used

An equipment status assessment system is constructed by adopting an asynchronous decoupled architecture with message queues, a tracking and identification mechanism, adaptive preprocessing based on operating conditions, and visualized linkage handling. The system includes a main software system module, a message queue scheduling module, an independent status assessment module, and a result fusion and feedback module, which realizes data verification, fixed-length sequence construction, and adaptive processing based on operating conditions.

Benefits of technology

It achieves highly available, highly accurate, scalable, and traceable device status assessment, improving the system's robustness and operational decision-making efficiency. It can smooth out peaks and valleys, isolate anomalies, and recover from faults in high-concurrency scenarios, providing comprehensive assessment results such as health status and fault level.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114906A_ABST
    Figure CN123114906A_ABST
Patent Text Reader

Abstract

This invention discloses an equipment condition assessment system and method, belonging to the field of equipment condition monitoring. The system includes a main software system module for collecting raw waveform data from the equipment to construct a message to be assessed; a message queue scheduling module for temporarily storing and routing the message to be assessed; an independent condition assessment module for acquiring the message to be assessed, performing data verification, fixed-length sequence construction, adaptive preprocessing based on operating conditions, and condition assessment calculation; a result fusion and feedback module for encapsulating the condition assessment results and feeding them back to the result feedback queue; and a visualization and processing module for consuming the condition assessment results from the result feedback queue and performing visualization and processing. This invention, by introducing an asynchronous decoupling architecture for message queues, a tracking and identification mechanism, adaptive preprocessing based on operating conditions, and visualization-linked processing, constructs a highly available, highly accurate, scalable, and traceable equipment condition assessment system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of equipment condition monitoring. Specifically, this invention relates to an equipment condition assessment system and method. Background Technology

[0002] In the field of industrial equipment condition monitoring and fault diagnosis, operating parameters such as high-frequency vibration waveforms, sound data, temperature data, and rotational speed typically need to be continuously collected and evaluated online. Although existing technologies have been able to apply message middleware and machine learning models to relevant scenarios, they still have significant shortcomings in application environments with multiple measurement points, continuous flow, and strong real-time requirements in industrial equipment.

[0003] 1. The data acquisition, inference, and display processes are heavily coupled. Traditional solutions often deploy device data acquisition, model calculation, and front-end display in the same business process. Once the inference computation load increases, it can easily consume the main system resources, leading to acquisition blockage, interface lag, or even data loss.

[0004] 2. The original time series data has characteristics such as inconsistent length, inconsistent sampling rate, and large differences in rotation speed under different measurement points and different operating conditions. Existing solutions often directly apply the input method of general models, lacking the fixed length, alignment and operating condition adaptive construction process for industrial scenarios, resulting in insufficient stability of evaluation results.

[0005] 3. Existing solutions typically only focus on model classification or scoring itself, lacking mechanisms for message tracking, deduplication verification, failure retry, anomaly isolation, result binding, and reliable feedback for industrial scenarios, making it difficult to operate stably for a long time in continuous online scenarios.

[0006] 4. Existing assessment outputs are mostly single threshold alarms or simple category labels, which cannot simultaneously provide health status, fault level, fault type, confidence information and maintenance suggestions, making it difficult to directly support closed-loop operation and maintenance.

[0007] To address the above problems, this invention proposes a system and method for assessing equipment condition. Summary of the Invention

[0008] This invention aims to overcome the shortcomings of existing technologies and proposes an equipment status assessment system and method to achieve the following objectives: by introducing a message queue asynchronous decoupling architecture, a tracking and identification mechanism, adaptive preprocessing of operating conditions, and visualized linkage handling, a highly available, highly accurate, scalable, and traceable intelligent equipment status assessment system is constructed.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows: This invention provides a device condition assessment system, the system comprising: The main software system module is used to acquire raw waveform data of the equipment in real time and its corresponding sampling parameters, operating parameters and measurement point identification information, and to construct a message to be evaluated based on this; the message to be evaluated includes a trace identifier (trace_id) generated by the measurement point identifier, sampling time and data type; The message queue scheduling module includes a raw data queue, a retry queue, a dead letter queue, and a result return queue, which are used to temporarily store and route messages to be evaluated. The independent state assessment module is used to obtain the message to be assessed from the message queue scheduling module, and perform data verification, fixed-length sequence construction, working condition adaptive preprocessing, and state assessment calculation. The result fusion and feedback module is used to encapsulate the status evaluation results and send them back to the result feedback queue.

[0010] The visualization and processing module is used to process the status evaluation results returned to the queue of consumption results and perform visualization and processing.

[0011] Furthermore, the message to be evaluated also includes a message priority for message scheduling and a number of retries for failure retry control.

[0012] Furthermore, in the independent state evaluation module, the data verification includes first performing deduplication verification based on trace_id, and then performing consistency checks on preset message fields to be evaluated. The preset message fields to be evaluated include the original waveform data length, sampling frequency, and sampling point interval.

[0013] Furthermore, for messages awaiting evaluation that fail data verification, they are first transferred to the retry queue to re-execute data verification; if data verification still fails after a preset number of retries, the corresponding message awaiting evaluation is transferred to the dead letter queue.

[0014] Furthermore, in the independent state assessment module, the fixed-length sequence construction includes: performing fixed-length processing based on the length of the original waveform data; when the length of the original waveform data is greater than the target length, performing truncation or segment extraction; when the length of the original waveform data is less than the target length, performing padding processing.

[0015] Furthermore, during the completion process, a mask sequence is generated synchronously for the completion position to distinguish between the actual sampling point and the completion sampling point.

[0016] Furthermore, in the independent state assessment module, the adaptive preprocessing of operating conditions includes: normalizing each fixed-length sequence in combination with the equipment operating condition parameters to unify the equipment operating condition parameters of each fixed-length sequence; and organizing each fixed-length sequence into a unified input structure.

[0017] Furthermore, in the independent state assessment module, the state assessment calculation includes: inputting the preprocessed fixed-length sequence into a preset state assessment model for forward calculation, and outputting the state assessment result, which includes health status, fault level, and fault type.

[0018] Furthermore, in the result fusion and feedback module, when encapsulating the state evaluation results, in addition to adding the state evaluation results to the encapsulation, the trace identifier trace_id is also added to the encapsulation.

[0019] The present invention also provides a method for assessing equipment condition, using the above-described equipment condition assessment system, the method comprising: Step S1: Real-time acquisition of raw waveform data from the device and its corresponding sampling parameters, operating parameters, and measurement point identification information, and construction of a message to be evaluated based on this; the message to be evaluated includes a trace identifier (trace_id) generated from the measurement point identifier, sampling time, and data type; Step S2: Temporarily store and route messages to be evaluated through a pre-built message queue, which includes a raw data queue, a retry queue, a dead letter queue, and a result return queue. Step S3: Obtain the message to be evaluated from the original data queue or retry queue, and perform data verification, fixed-length sequence construction, working condition adaptive preprocessing, and state evaluation calculation. Step S4: After encapsulating the state evaluation results, send them back to the result return queue.

[0020] Step S5: The consumption result is returned to the status evaluation result in the queue and then visualized and processed.

[0021] The technical effects of this invention are as follows: 1. This invention is designed for high-frequency online scenarios of industrial equipment, and integrates message organization, asynchronous scheduling, adaptive sequence construction, and result feedback methods, with clear engineering technical solution features; 2. Through the collaborative design of raw data queue, retry queue, dead letter queue and result feedback queue, the acquisition link and the computing link can be decoupled. In high-concurrency scenarios, it has the ability to smooth out peaks and valleys, isolate anomalies and recover from faults, thereby improving the robustness of the system. 3. By performing fixed-length normalization, masking and adaptive normalization on the original time-domain data, the input differences caused by different measurement points, different working conditions and different sampling strategies can be reduced, so that the subsequent evaluation model can obtain a more stable input structure. 4. By converting the model output into hierarchical results such as health status, fault level, fault type, maintenance recommendation time, fault probability, and confidence information, the traditional single alarm method can be upgraded to a comprehensive status assessment method oriented towards operation and maintenance decision-making. 5. By binding trace_id and confirming the return of results, it can be ensured that each evaluation result can be traced back to the corresponding original sampling batch, which facilitates subsequent review, accountability and continuous model optimization. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall architecture of the equipment condition assessment system of the present invention; Detailed Implementation

[0023] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. This is to help those skilled in the art to have a more complete, accurate, and in-depth understanding of the inventive concept and technical solutions of the present invention, and to facilitate its implementation. It should be noted that the terms "first," "second," etc., used in this application are only for the convenience of describing the technical solutions and to distinguish components; the corresponding component configurations may be the same or different, and are not intended to limit the scope of this application. To make the technical solutions of the present invention clearer, the present invention will be explained and illustrated through the following embodiments.

[0024] This invention provides a device status assessment system, such as... Figure 1 As shown, the system includes: The main software system module is used to acquire raw waveform data of the equipment in real time and its corresponding sampling parameters, operating parameters and measurement point identification information, and to construct a message to be evaluated based on this; the message to be evaluated includes a trace identifier (trace_id) generated by the measurement point identifier, sampling time and data type; The message queue scheduling module includes a raw data queue, a retry queue, a dead letter queue, and a result return queue, which are used to temporarily store and route messages to be evaluated. The independent state assessment module is used to obtain the message to be assessed from the message queue scheduling module, and perform data verification, fixed-length sequence construction, working condition adaptive preprocessing, and state assessment calculation. The result fusion and feedback module is used to encapsulate the status evaluation results and send them back to the result feedback queue.

[0025] The visualization and processing module is used to process the status evaluation results returned to the queue of consumption results and perform visualization and processing.

[0026] Specifically, the main software system module is deployed on edge computing nodes or a central server at the equipment site. It receives raw waveform data in real time from monitoring devices such as accelerometers, velocity sensors, and displacement sensors via multi-channel data acquisition cards or network interfaces. During acquisition, it simultaneously reads sampling parameters such as sampling frequency, sampling interval, and sampling time for each measurement point, as well as operating parameters such as equipment speed, load, and temperature. It also acquires measurement point identification information (such as unique measurement point identifier and measurement point code). The system encapsulates this information into a unified format message to be evaluated. This message includes at least a trace identifier (trace_id) generated from the measurement point identifier, sampling time, and data type, used for link tracing. Preferably, the message also includes message priority for message scheduling, prioritizing messages with higher priority during subsequent message routing; and the number of retries for failure retry control.

[0027] The above methods enable data identity and operational condition binding to be completed at the very beginning of the raw data entering the evaluation process, providing a foundation for subsequent reliable scheduling, anomaly isolation, and result tracking. The main software system module decouples the acquisition and evaluation ends through message-based encapsulation, facilitating the independent expansion of acquisition or evaluation nodes and improving the system's maintainability and scalability.

[0028] In high-concurrency, high-throughput data flow scenarios in industrial settings, if synchronous calls are used to process messages to be evaluated, a momentary overload or failure in the status evaluation module will directly cause the main software system module to block, leading to data backlog or even data acquisition interruption. Therefore, this embodiment introduces a message queue scheduling module as an asynchronous buffer layer. The message queue scheduling module constructs multiple queues, including a raw data queue, a retry queue, a dead-letter queue, and a result feedback queue.

[0029] The message queue scheduling module is implemented based on distributed message middleware such as RabbitMQ. The raw data queue receives messages to be evaluated from the main software system modules, storing them in order of priority. The independent state evaluation module, acting as a consumer, pulls messages from the raw data queue for processing. When a message fails to process due to an exception in the independent state evaluation module, it is transferred to the retry queue for further processing and retrying. The number of retries is recorded. If the message still fails after reaching a preset maximum number of retries, it is transferred to the dead-letter queue for manual intervention or logging, thus preventing abnormal messages from continuously blocking the normal evaluation process. The result feedback queue receives the state evaluation results encapsulated by the result fusion feedback module for consumption by the visualization and processing module.

[0030] The message queue scheduling module can simultaneously achieve asynchronous decoupling, peak shaving and valley filling, anomaly isolation and recoverable processing in high-concurrency scenarios, significantly improving the engineering availability of the online evaluation system.

[0031] Industrial equipment condition assessment involves multiple complex computational stages, including data verification, sequence normalization, operating condition adaptation, and model inference. If coupled with the main software system, this would severely restrict the iterative upgrades of the assessment algorithm and the independent scheduling of computing resources. Therefore, this embodiment separates the condition assessment logic into an independent condition assessment module, using a message queue for asynchronous consumption. This enables on-demand allocation of computing resources and independent evolution of the assessment algorithm. Furthermore, through steps such as data verification, fixed-length sequence construction, and operating condition adaptive preprocessing, the consistency of data quality and operating condition independence input to the condition assessment model are ensured, thereby improving the accuracy and generalization ability of the model inference.

[0032] Specifically, the independent state assessment module is used to obtain the message to be assessed from the message queue scheduling module, and perform data verification, fixed-length sequence construction, working condition adaptive preprocessing, and state assessment calculation.

[0033] The data verification process includes first performing deduplication based on the trace_id. This involves querying the local or distributed cache to determine if the trace_id has already been processed. If it already exists, the message is discarded to prevent duplicate evaluation due to repeated message delivery. Next, a consistency check is performed on preset fields of the message to be evaluated, including the original waveform data length, sampling frequency, and sampling interval. For messages that fail data verification, they are first transferred to a retry queue for re-verification. If data verification still fails after a preset number of retries, the corresponding message is transferred to a dead-letter queue.

[0034] After data validation, fixed-length sequence construction is performed because state evaluation models typically require input sequences of fixed length. Fixed-length sequence construction specifically includes: lengthening the original waveform data; truncating or segmenting the original waveform data if its length exceeds the target length; and padding the original waveform data if its length is less than the target length. During padding, a mask sequence is simultaneously generated for the padding positions to distinguish between actual sampling points and padding sampling points.

[0035] For each obtained fixed-length sequence, adaptive preprocessing based on operating conditions is performed. This adaptive preprocessing includes normalizing each fixed-length sequence based on equipment operating parameters to unify these parameters. This is because the normal or faulty state of the equipment may differ under different operating conditions. For example, in operating condition A, detecting feature C is considered a fault; however, in operating condition B, the same detected feature C is not considered a fault. After normalization, the fixed-length sequences are organized into a unified input structure. Compared to directly applying a general model input method, adaptive preprocessing based on operating conditions significantly reduces the impact of dimensional deviations and structural differences between different measurement points and operating conditions on the stability of the evaluation results.

[0036] Finally, a state assessment calculation is performed, including: inputting the preprocessed fixed-length sequence into a preset state assessment model for forward calculation, and outputting the state assessment result. The state assessment result includes health status, fault level, and fault type. Preferably, it also includes maintenance recommendation time, fault probability, and confidence information to form a hierarchical and interpretable output result suitable for industrial sites. In this embodiment, the state assessment model adopts a Transformer temporal model with a self-attention mechanism to model long-distance temporal dependencies and global feature associations. In specific implementations, the state assessment model is not limited to a specific implementation of Transformer; other attention-based temporal models with global dependency modeling capabilities can also be used, provided that they are used in conjunction with the message organization and adaptive input structure of this invention.

[0037] The independent state assessment module decouples assessment calculation from data acquisition through independent deployment and asynchronous consumption mechanisms, enabling the assessment algorithm to be iteratively upgraded independently without affecting the operation of the acquisition end. The data verification process effectively filters out abnormal and duplicate data, ensuring the reliability of the assessment input. Fixed-length sequence construction and masking mechanisms ensure the standardization of model inputs while avoiding the spurious impact of data completion on the assessment results. The adaptive preprocessing of operating conditions eliminates the interference of operating condition changes on state assessment, allowing the model to focus on the true state degradation characteristics of the equipment, significantly improving the accuracy of state assessment and the ability to generalize operating conditions.

[0038] The result fusion and feedback module is used to encapsulate the status assessment results and send them back to the result feedback queue. In this embodiment, when encapsulating the status assessment results, in addition to adding the status assessment results to the encapsulation, the trace identifier (trace_id) is also added. Adding the trace identifier (trace_id) to the result encapsulation enables full-link closed-loop traceability between the assessment results and the original data, ensuring that each assessment result can be accurately traced back to its data source. In addition, other information such as measurement point identifiers, sampling time, and processing status can also be added to the encapsulation as needed. Preferably, the subsequent visualization and processing module only consumes and confirms the original message to be evaluated after the result message is successfully fed back, thereby ensuring the consistency and traceability between the original data and the fed-back results.

[0039] If status assessment results are stored solely as data, they are difficult for maintenance personnel to understand intuitively and make quick decisions, and the lack of automated handling and linkage mechanisms will lead to delayed fault response. Therefore, this embodiment introduces a visualization and handling module. By consuming the assessment results and sending them back to the queue for visualization, abstract information such as health level and fault level can be transformed into intuitive charts and alarm prompts. At the same time, combined with preset handling rules, automated handling can be performed to achieve timely fault detection and rapid response, reducing the risk of unplanned equipment downtime. Specifically, for cases where the fault level reaches a preset threshold or the result credibility is lower than a preset threshold, an alarm, work order, manual review, or downgrade display process is preferably triggered, thus forming a closed loop from data collection to handling.

[0040] The present invention also provides a method for assessing equipment condition, using the above-described equipment condition assessment system, the method comprising: Step S1: Real-time acquisition of raw waveform data from the device and its corresponding sampling parameters, operating parameters, and measurement point identification information, and construction of a message to be evaluated based on this; the message to be evaluated includes a trace identifier (trace_id) generated from the measurement point identifier, sampling time, and data type; Step S2: Temporarily store and route messages to be evaluated through a pre-built message queue, which includes a raw data queue, a retry queue, a dead letter queue, and a result return queue. Step S3: Obtain the message to be evaluated from the original data queue or retry queue, and perform data verification, fixed-length sequence construction, working condition adaptive preprocessing, and state evaluation calculation. Step S4: After encapsulating the state evaluation results, send them back to the result return queue.

[0041] Step S5: The consumption result is returned to the status evaluation result in the queue and then visualized and processed.

[0042] In summary, this embodiment achieves independent expansion and fault isolation between the acquisition and assessment ends by asynchronously decoupling the main software system module and the independent state assessment module through message queues. A multi-level scheduling mechanism using raw data queues, retry queues, and dead-letter queues effectively addresses peak shaving and valley filling and anomaly recovery in high-concurrency scenarios. Trace_id enables closed-loop traceability from data acquisition to result feedback, while data verification, fixed-length sequence construction, and masking mechanisms ensure the standardization and reliability of model inputs. Adaptive preprocessing under different operating conditions eliminates dimensional biases and state misjudgments, enabling state assessment based on the Transformer time-series model to focus on the true degradation characteristics of the equipment. Finally, through visualization and automated handling linkage, abstract assessment results are transformed into intuitive alarms and rapid response actions, significantly improving the engineering usability, assessment accuracy, operating condition generalization ability, and maintenance decision-making efficiency of the industrial equipment state assessment system.

[0043] The present invention has been described above by way of example with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-described manner. Any non-substantial improvements made using the inventive concept and technical solution; or the direct application of the inventive concept and technical solution to other situations without modification, are all within the protection scope of the present invention.

Claims

1. A device condition assessment system, characterized in that: The system includes: The main software system module is used to acquire raw waveform data of the equipment in real time and its corresponding sampling parameters, operating parameters and measurement point identification information, and to construct a message to be evaluated based on this; the message to be evaluated includes a trace identifier (trace_id) generated by the measurement point identifier, sampling time and data type; The message queue scheduling module includes a raw data queue, a retry queue, a dead letter queue, and a result return queue, which are used to temporarily store and route messages to be evaluated. The independent state assessment module is used to obtain the message to be assessed from the message queue scheduling module, and perform data verification, fixed-length sequence construction, working condition adaptive preprocessing, and state assessment calculation. The result fusion and feedback module is used to encapsulate the status evaluation results and send them back to the result feedback queue. The visualization and processing module is used to process the status evaluation results returned to the queue of consumption results and perform visualization and processing.

2. The equipment condition assessment system according to claim 1, characterized in that: The message to be evaluated also includes message priority for message scheduling and the number of retries for failure retry control.

3. The equipment condition assessment system according to claim 1, characterized in that: In the independent state evaluation module, the data verification includes first performing deduplication verification based on trace_id, and then performing consistency checks on preset message fields to be evaluated. The preset message fields to be evaluated include the original waveform data length, sampling frequency, and sampling point interval.

4. A device condition assessment system according to any one of claims 1 or 3, characterized in that: For messages awaiting evaluation that fail data verification, they are first transferred to the retry queue to re-execute data verification; if data verification still fails after a preset number of retries, the corresponding message awaiting evaluation is transferred to the dead letter queue.

5. The equipment condition assessment system according to claim 1, characterized in that: In the independent state assessment module, the fixed-length sequence construction includes: performing fixed-length processing based on the length of the original waveform data; when the length of the original waveform data is greater than the target length, performing truncation or segment extraction; when the length of the original waveform data is less than the target length, performing padding processing.

6. The equipment condition assessment system according to claim 5, characterized in that: During the padding process, a mask sequence is generated synchronously for the padding position to distinguish between the actual sampling point and the padding sampling point.

7. The equipment condition assessment system according to claim 1, characterized in that: In the independent state assessment module, the adaptive preprocessing of operating conditions includes: normalizing each fixed-length sequence in combination with the equipment operating condition parameters to unify the equipment operating condition parameters of each fixed-length sequence; and organizing each fixed-length sequence into a unified input structure.

8. The equipment condition assessment system according to claim 1, characterized in that: In the independent state assessment module, the state assessment calculation includes: inputting the preprocessed fixed-length sequence into a preset state assessment model for forward calculation, and outputting the state assessment result, which includes health status, fault level and fault type.

9. The equipment condition assessment system according to claim 1, characterized in that: In the result fusion and feedback module, when encapsulating the state evaluation results, in addition to adding the state evaluation results to the encapsulation, the trace identifier trace_id is also added to the encapsulation.

10. A method for assessing equipment condition, using an equipment condition assessment system according to any one of claims 1-9, characterized in that: The method includes: Step S1: Real-time acquisition of raw waveform data from the device and its corresponding sampling parameters, operating parameters, and measurement point identification information, and construction of a message to be evaluated based on this; the message to be evaluated includes a trace identifier (trace_id) generated from the measurement point identifier, sampling time, and data type; Step S2: Temporarily store and route messages to be evaluated through a pre-built message queue, which includes a raw data queue, a retry queue, a dead letter queue, and a result return queue. Step S3: Obtain the message to be evaluated from the original data queue or retry queue, and perform data verification, fixed-length sequence construction, working condition adaptive preprocessing, and state evaluation calculation. Step S4: Encapsulate the state evaluation results and send them back to the result feedback queue; Step S5: The consumption result is returned to the status evaluation result in the queue and then visualized and processed.