State detection method and system for industrial equipment

Abstract

The invention relates to the technical field of equipment detection, and provides a state detection method and system for industrial equipment, and the method comprises the steps: collecting a part of time sequence characteristic parameters of the industrial equipment, constructing a change curve of the characteristic parameters of the industrial equipment along with time based on the part of time sequence characteristic parameters, and obtaining a historical degradation curve, obtaining a complete time sequence characteristic parameter of the industrial equipment based on the historical degradation curve; on the basis of the complete time sequence characteristic parameters of the industrial equipment, constructing a health degree mapping model for outputting health indexes of the industrial equipment on the basis of the characteristic parameters on the basis of the complete time sequence characteristic parameters; acquiring characteristic parameters of the industrial equipment in real time by using a plurality of sensors; inputting the characteristic parameter data into a health degree mapping model to output a health index; and if the health index deviates from the preset normal value range, judging that the state of the industrial equipment is abnormal. According to the invention, real-time monitoring and fault early warning of the industrial equipment state can be realized.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of equipment detection, and in particular relates to a state detection method and system for industrial equipment. Background Art

[0002] Currently, industrial equipment status detection mainly relies on two modes: manual inspection mode and single sensor data collection mode; Among them, the manual inspection mode (such as Figure 3 (As shown) technicians regularly (e.g., daily or weekly) visually inspect the equipment, manually record operating parameters (such as temperature and vibration values), and use experience to determine whether the equipment is abnormal. This model requires the use of handheld detection tools (such as infrared thermometers and vibration analyzers) as technical support, and data recording is mainly done on paper or simple spreadsheets. However, due to limited human resources, inspection frequency is typically once a day, which cannot cover sudden equipment failures (such as sudden bearing overheating). Response times can take hours or even days, making it difficult to meet the requirements of high-precision production lines. The single-sensor data collection model has a data upload cycle of minutes (e.g., once per minute) and is limited by the communication protocol bandwidth (e.g., the low transmission rate of RS-485), resulting in significant delays in key status information. Single sensor data acquisition mode (such as Figure 4 (As shown in the figure), a single type of sensor (such as temperature, current, or pressure) is connected to a PLC (Programmable Logic Controller) system, which collects data regularly and uploads it to a local server or cloud storage. Its technical support is based on transmission via traditional industrial protocols such as Modbus and OPC. The analysis module relies on threshold alarms (e.g., temperature exceeding a set value triggers an alarm), lacking multi-dimensional data correlation analysis. However, current industrial equipment status detection technologies generally have problems such as insufficient real-time performance, low prediction accuracy, poor system integration, and high maintenance costs, which seriously restrict production efficiency and equipment reliability. Summary of the Invention

[0003] To solve the above problems, the present disclosure provides a status detection method and system for industrial equipment, which adopts redundant sensor cross-validation technology, edge computing and health mapping model, can realize multi-dimensional data fusion and dynamic prediction, improve the real-time, accuracy and automation level of equipment status detection, and reduce manual operation and maintenance costs.

[0004] The following is the technology disclosed in this disclosure: A method for detecting the status of industrial equipment, comprising: Collect some time series characteristic parameters of industrial equipment, construct a characteristic parameter change curve of the industrial equipment over time based on the partial time series characteristic parameters, obtain a historical degradation curve, and obtain the complete time series characteristic parameters of the industrial equipment based on the historical degradation curve; Based on the complete time series characteristic parameters of the industrial equipment, a health mapping model is constructed based on the complete time series characteristic parameters to output the health index of the industrial equipment; Use multiple sensors to collect characteristic parameters of industrial equipment in real time; The characteristic parameter data is input into the health mapping model to output the health index; if the health index deviates from the preset normal value range, it is judged that the state of the industrial equipment is abnormal.

[0005] Further, The method of constructing a characteristic parameter change curve of industrial equipment over time based on some time series characteristic parameters to obtain a historical degradation curve includes: Obtain the timing characteristic parameter sequence of industrial equipment under normal conditions; Based on the time series characteristic parameter sequence, the least square method is used to construct a curve showing the characteristic parameter changes over time, and a historical degradation curve is obtained.

[0006] Further, The method of constructing a curve of characteristic parameters changing with time using the least squares method includes: The equation defining the curve is T=at+b, where T is the characteristic parameter and t is the time. n data points , i-1,2,3...n, Represents the characteristic parameters of the i-th data point at time t; Constructing the objective function ; Calculate the partial derivatives of S with respect to a and b respectively and set them to 0, and we get the following system of equations:

[0007]

[0008] Solving the system of equations yields:

[0009] Substituting the values of a and b into T=at+b, we can obtain the curve of characteristic parameters changing with time.

[0010] Further, The health mapping model is an LSTM network.

[0011] Further, The method of using multiple sensors to collect characteristic parameters of industrial equipment in real time includes: Use multiple sensors to collect characteristic parameters of industrial equipment and transmit them to edge computing devices; The edge computing device uses redundant sensor cross-validation technology to correct the multiple characteristic parameters to obtain corrected data; after filtering the corrected data, the feature information is extracted using a feature extraction algorithm; the feature information is compressed to obtain final characteristic parameter data.

[0012] A status detection system for industrial equipment, comprising: A data processing module is used to collect partial time series characteristic parameters of industrial equipment, construct a characteristic parameter change curve of the industrial equipment over time based on the partial time series characteristic parameters, obtain a historical degradation curve, and obtain the complete time series characteristic parameters of the industrial equipment based on the historical degradation curve; Based on the complete time series characteristic parameters of the industrial equipment, a health mapping model is constructed based on the complete time series characteristic parameters to output the health index of the industrial equipment; The device perception module is used to collect characteristic parameters of industrial equipment in real time using multiple sensors; The data processing module is further configured to input characteristic parameter data into a health mapping model to output a health index; if the health index deviates from a preset normal value range, it is determined that the state of the industrial equipment is abnormal.

[0013] Further, The device perception module includes: A data acquisition unit, used to collect multiple characteristic parameters of industrial equipment using a programmable logic controller system integrating multiple sensors and then output them; The edge computing unit is used to correct the characteristic parameters using redundant sensor cross-validation technology to obtain correction data; after filtering the correction data, feature information is extracted using a feature extraction algorithm; and the feature information is compressed to obtain final characteristic parameter data.

[0014] Further, The EtherCAT protocol is used to transmit characteristic parameters between the edge computing unit and the data processing module.

[0015] Further, Also includes: The early warning module is used to calculate the health deviation of the equipment and issue graded early warnings based on the health deviation; The calculation of the health deviation is:

[0016] in, is the health index, The middle value of the normal range of health index, This is the minimum value of the normal range of health index.

[0017] Further, The characteristic parameters include: vibration, temperature, current, and pressure.

[0018] Compared with the prior art, the present disclosure has the following advantages: This paper collects some time series characteristic parameters of industrial equipment and uses the least squares method to construct a historical degradation curve to complete the complete time series characteristic parameters, providing full-cycle data support for the health mapping model. The health mapping model is constructed based on the complete time series characteristic parameters to achieve dynamic mapping between multi-dimensional characteristic parameters and health index. Then, multiple sensors are used to collect characteristic parameters in real time and input them into the model to output a health index after edge computing pre-processing. The abnormal equipment status is judged by the deviation of the health index from the preset normal range; historical data gaps are filled through data fitting, multi-source data fusion analysis is realized with the help of dynamic models, and the real-time and accuracy of detection are improved by relying on real-time collection and intelligent algorithms, thereby solving the problems of insufficient real-time performance, low prediction accuracy, poor system integration and high maintenance cost in the background technology.

[0019] Other features and advantages of the present disclosure will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present disclosure. The purposes and other advantages of the present disclosure can be realized and obtained by the structures indicated in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 A schematic diagram of the disclosed method is shown; Figure 2 A schematic diagram of data preprocessing logic disclosed in the present invention is shown; Figure 3 Shows a manual inspection flow chart; Figure 4 Shown is a single sensor data acquisition flow chart. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0023] Technical Overview: This paper integrates multiple types of sensors such as vibration, temperature, current, and pressure through a Programmable Logic Controller (PLC) system, adopts a high-speed industrial communication protocol (EtherCAT protocol, with a transmission rate ≥100Mbps and an extended protocol of up to 10Gbps), shortens the data acquisition cycle to milliseconds (≤10ms), and uses an edge computing module to perform local pre-processing (noise filtering, feature extraction, and data compression) on the raw data, significantly reducing cloud load and transmission delay, and ensuring the real-time and integrity of device status information.

[0024] The core contents of this disclosure are mainly three points: redundant sensor cross-validation technology and edge computing technology, dynamic fusion characteristic curve of LSTM-based neural network to establish an aging / health mapping model, and an integrated platform to realize full-process automated management.

[0025] Figure 1 A schematic diagram of the method according to the present disclosure is shown, and the specific implementation details of the present disclosure include: Step 1: Deploy multiple sensors to collect sensor data; Step 2: Use the edge computing module (one edge computing module is connected to multiple identical sensors) and adopt the redundant sensor cross-validation technology to calibrate and fit the data collected by multiple sensors; and perform local preprocessing (filtering, feature extraction, data compression) on the calibrated and fitted sensor data; (e.g. Figure 2 shown) Among them, the redundant sensor cross-validation technology includes: The confidence algorithm is used to determine whether there is any deviation in the multiple characteristic parameters collected by the data acquisition unit, and the backup data acquisition unit is used to replace the deviated data acquisition unit, and the deviated data acquisition unit is automatically corrected based on the normal data acquisition unit.

[0026] Step 3: Transmit the pre-processed data to the subsequent processing module through a high-speed industrial communication protocol; Step 4: Based on the preprocessed data, the following processing can be performed: 1) The performance degradation law of computing equipment. For example: Assuming that the collected device characteristic parameters (such as temperature T) change with time t, define the linear regression equation as T=at+b, and use the least squares method to determine the coefficients a and b; Assume there are n data points, where , i-1,2,3...n; Constructing the objective function ; Calculate the partial derivatives of a and b respectively and set them to 0. The calculation formula is:

[0027]

[0028] Solving the above equations yields:

[0029] Based on the calculated a and b, the linear change relationship of the equipment characteristic parameters over time can be obtained, thereby quantifying the performance degradation law.

[0030] Purpose: The collected equipment characteristic parameters are not the original parameters of the industrial equipment from the beginning of use, but only partial time series characteristic parameters. The training data of the following time series prediction network model requires complete time series data. Therefore, by constructing a historical degradation curve, the complete time series characteristic parameters of the industrial equipment can be obtained for LSTM network training.

[0031] 2) Mapping of aging model and monitoring status. For example: An aging / health mapping model is established based on the LSTM neural network; the core of LSTM is the memory unit and gating mechanisms.

[0032] The memory unit update formula is:

[0033] in: It's the Gate of Forgetfulness. ; is the input gate, ; is the output gate, ; is the sigmoid function; Represents element-wise multiplication; is the weight is the bias, is the hidden layer output at the previous moment, is the current moment input.

[0034] The hidden layer output is: ; When establishing the aging / health mapping model, the device characteristic parameters (such as current harmonic distortion THD) are used as input , the LSTM network continuously learns and updates parameters by obtaining input parameters and converts them into device monitoring index status values , By mapping the LSTM output to a certain level (such as linear mapping , T is the length of the time series, and is the mapping value and bias).

[0035] 3) Online monitoring, detection and real-time warning; for example: Collect online equipment data in real time, combine historical degradation curves with LSTM prediction results to trigger graded warnings; Assume that the normal range of the equipment health index obtained based on historical data is [ , ], the current health index predicted by LSTM is ; Calculate the health index deviation:

[0036] When D exceeds a certain threshold (such as the yellow warning threshold Dy, the red warning threshold Dr, and when Dr is greater than Dy), the corresponding warning is triggered.

[0037] The following is an all-in-one platform that can automatically implement the above methods: Furthermore, to achieve full-process automated management of intelligent operation and maintenance of industrial equipment, this paper designs an integrated platform based on the above method to achieve full-process automated management. Its core lies in integrating all aspects of equipment operation, from data collection, processing, analysis to decision execution, all of which are automated to improve operation and maintenance efficiency and reduce the risk of human intervention. The following is an introduction to the platform: (1) Platform architecture construction: Architecture: The integrated platform adopts a layered distributed architecture, which is divided into device perception layer, data transmission layer, data processing layer, application management layer and user interaction layer from bottom to top.

[0038] Among them, the device perception layer integrates multiple types of redundant sensor groups to comprehensively collect equipment operation status data; the data transmission layer relies on the EtherCAT high-speed communication protocol to ensure fast and stable data transmission; the data processing layer integrates redundant sensor cross-validation technology and edge computing technology to clean, calibrate and pre-process data; the application management layer carries various business applications, such as equipment status monitoring, fault prediction, maintenance decision-making, etc.; the user interaction layer provides an intuitive and convenient operation interface for operation and maintenance personnel.

[0039] (2) Data-driven automated processes: 2.1) Automated Data Collection: Multiple redundant sensor groups automatically collect equipment operating data at a set interval (millisecond level) based on equipment operating characteristics and key monitoring points. Equipment operation data includes parameters such as temperature, pressure, vibration, and current; The sensor group has a self-diagnosis function and monitors its own working status in real time. Once a fault or abnormality is detected, it immediately switches to the redundant backup sensor to ensure the continuity of data collection.

[0040] 2.2) Data processing automation: The collected data is transmitted to the edge computing node via the EtherCAT protocol and first enters the redundant sensor cross-validation module; As mentioned above, through cross-comparison of dual backup temperature sensors, etc., it is automatically determined whether the data has anomalies such as zero drift and temperature drift. If anomalies are found, fitting and calibration are automatically performed; the calibrated data is then processed by edge computing such as noise filtering, feature extraction, and data compression in sequence to provide high-quality data for subsequent analysis.

[0041] 2.2) Automated Equipment Status Analysis and Prediction: Utilizing an aging / health mapping model based on an LSTM neural network, the platform accesses preprocessed data in real time. The model automatically runs, converting equipment operating parameters into equipment monitoring index status values to quantify equipment health. Simultaneously, it combines historical and real-time data to predict the future operating status of the equipment and anticipate potential failure risks.

[0042] 2.3) Maintenance decision-making and execution automation: Based on equipment status analysis and prediction results, the platform's built-in decision engine automatically generates maintenance decisions according to preset rules and strategies; For example, when the health of a device falls below a set threshold or a failure is predicted, a maintenance work order is automatically triggered and corresponding maintenance resources, such as maintenance personnel and spare parts, are arranged. The maintenance work order is sent to relevant personnel through the platform, and maintenance personnel can receive the work order on their mobile devices and view detailed device information and fault diagnosis reports, and perform maintenance operations according to the instructions. After the maintenance is completed, the maintenance personnel will provide feedback on the maintenance results on the platform, and the platform will automatically update the device status information to complete the maintenance closed-loop management.

[0043] (3) Intelligent function realization: 3.1) Intelligent Early Warning: The platform sets up a multi-level early warning mechanism, automatically issuing different levels of early warning information according to the degree to which the equipment status deviates from the normal range; In addition to traditional threshold warnings, a machine learning-based anomaly detection algorithm has been introduced, which can detect abnormal patterns in the data and issue early warnings, thus buying more processing time for operation and maintenance personnel.

[0044] 3.2) Intelligent report generation: The platform automatically generates various equipment operation reports based on user needs, including daily, weekly, and monthly reports; The report content covers equipment operating status statistics, fault analysis, maintenance records and other information, and is presented in the form of a combination of charts and data, allowing users to intuitively understand the overall operation of the equipment and provide data support for management decisions.

[0045] 3.3) Knowledge Base and Experience Accumulation: The platform builds a knowledge base to store equipment failure cases, repair experience, maintenance strategies, and other knowledge; During the equipment operation and maintenance process, new fault cases and solutions are continuously added to the knowledge base to achieve knowledge accumulation and inheritance; at the same time, through knowledge graph technology, different knowledge nodes are linked to provide operation and maintenance personnel with intelligent search and knowledge recommendation services to assist them in quickly solving problems.

[0046] The above full process integrates data collection, processing, analysis and decision-making to form a "detection-analysis-control" closed loop, avoiding system fragmentation. Combined with self-calibration and automated model parameter adjustment, it reduces the frequency and cost of manual maintenance. Ultimately, through the collaborative technology chain, it improves real-time performance, prediction accuracy, integration, and reduces operation and maintenance costs, solving the core problems in the background technology.

[0047] The following are the significant technological advancements brought about by this disclosure: This invention solves the core defects of the existing technology one by one through innovative technical solutions, and achieves significant breakthroughs in real-time performance, prediction accuracy, integration and operation and maintenance costs. Specifically: 1) Improved real-time performance: millisecond-level response and high-frequency data collection This invention integrates multiple sensors for high-speed, synchronous collection of industrial equipment status data and field environmental data. It uses edge computing and localized data processing (data filtering, linear fitting, range conversion, and alarm forecasting) to reduce the computing load on the main station and data transmission bandwidth. Transmitting edge data via a high-speed communication protocol (EtherCat, with a transmission rate of ≥100Mbps, and up to 10Gbps using the EtherCat extended protocol) significantly reduces communication latency and improves real-time response rates.

[0048] By improving the real-time performance of overall industrial equipment detection through the above model, the data acquisition cycle can be shortened to milliseconds (≤10ms), and the response time to sudden faults (such as motor instantaneous overload, excitation system shutdown, equipment short circuit / open circuit) can be shortened to ≤200ms.

[0049] 2) Prediction Accuracy Optimization: Multi-Source Data Fusion and Dynamic Modeling This invention uses an LSTM (Long Short-Term Memory) neural network to dynamically integrate multiple sensors to simultaneously collect industrial equipment status data and field environmental data (such as temperature, voltage, current, pressure, speed, and vibration). The neural network's online learning mechanism adaptively optimizes model parameters based on equipment aging and changes in sensor-collected equipment operating conditions, enabling early prediction of faults and automatic identification of fault types and response measures (such as gate disconnection, equipment shutdown, and speed reduction).

[0050] The adaptive learning and adjustment of model parameter modes by the multi-source data fusion neural network dynamic model can significantly reduce the cost of manual parameter adjustment and calibration. Compared with manual detection, it significantly improves the speed of accident response and processing, improves the fault identification accuracy to more than 95%, and the error rate of working condition aging life prediction is less than 8%.

[0051] 3) Enhanced integration: Full-process automated closed-loop control This invention aims to build an integrated platform that integrates data collection, analysis, and control functions, supports OPC UA protocol data transmission within seconds, and displays real-time industrial equipment operating status information, sub-module operating status information, overall system health status index, and maintenance recommendations through a visual interface.

[0052] By forming a closed-loop "detection-analysis-control" link, the disadvantages of independent operation of analysis software and PLC system are avoided, and the decision response time is improved from less than 10 minutes to less than 1 second.

[0053] 4) Reduced operation and maintenance costs: Intelligent self-calibration and parameter optimization Traditional analog signal acquisition devices for industrial equipment, such as voltage sensors, current sensors, and temperature sensors, are subject to temperature drift, zero drift, and aging. If these analog signal acquisition devices use fixed fitting values, they can lose accuracy over time. Regular monthly calibration of industrial equipment is required, and overall operating parameters require expert adjustment and maintenance, increasing operational and maintenance costs.

[0054] This invention utilizes redundant sensor cross-validation technology, such as dual temperature probes, with interaction serving as a fitting benchmark. This interactive operation enables automatic calibration without offline operation. The entire system automatically adjusts model thresholds through a rules engine and a reinforcement neural network learning algorithm, minimizing manual intervention. This significantly reduces manual operation costs and the number of manual interventions, reducing manual calibration frequency from monthly or semi-annual to annual, resulting in a 35% reduction in annual maintenance costs.

[0055] Alternatives: This invention is based on a platform for automatically analyzing and predicting the status of automated detection equipment in a PLC system. It aims to uncover potential correlations between equipment failures by simultaneously analyzing the spatiotemporal characteristics of multi-source data (such as vibration spectra, temperature trends, and current harmonics) through a dynamic fusion model of an LSTM (Long Short-Term Memory) neural network. Redundant sensor cross-validation technology enables non-offline calibration of sensor modules, and cross-validation improves calibration accuracy during operation. High-speed bus communication protocols and edge computing modules reduce the data processing load on the central processing module and improve the overall system fault response rate. Through an integrated automated control platform, the entire "detection-analysis-control" chain is connected, reducing manual intervention, improving the operating efficiency of the entire automated detection platform, and reducing decision-making delays.

[0056] Currently, there are several technical routes to realize the automatic analysis and prediction platform of automated detection equipment status: 1) SCADA system and cloud-based AI analysis solution Traditional SCADA (Supervisory Control and Data Acquisition) is used to collect multi-sensor data, and the raw data is uploaded to the cloud server via the 4G / 5G network. Cloud-based AI models (such as convolutional neural networks (CNN)) are used for fault analysis, and the PLC is then controlled to execute actions through remote commands.

[0057] This system enables multi-source data collection and cloud-based analysis, supporting remote control and a visual interface. However, it requires a 4G / 5G network to upload raw data to the cloud for analysis. This results in network transmission delays, making it difficult to meet millisecond-level response requirements and making it easy to miss sudden failures. Furthermore, network transmission fluctuations can lead to data loss or command delays. This system also requires additional cloud-based computing resources and bandwidth costs, increasing both system setup and operational maintenance costs.

[0058] 2) Localized processing solution based on FPGA (field programmable gate array) FPGA replaces the edge computing unit of PLC, directly connects to sensors and realizes high-speed data acquisition (≤1ms cycle), has built-in lightweight machine learning models (such as random forest) for real-time analysis, and triggers control instructions through hardware logic circuits.

[0059] Due to the characteristics of FPGAs, they can process multiple data requests in parallel with very low response latency. This enables ultra-low-latency data acquisition and localized processing of logic and data calculations. FPGAs also support hardware acceleration, which can improve the computational efficiency of automatic prediction and early warning models. However, FPGA firmware requires custom development and struggles to support dynamic model updates—that is, dynamic learning and adjustment of model parameters based on field conditions and operational data. FPGAs are essentially digital circuits. The complex multi-sensor interface types and diverse data link communication protocols of industrial equipment necessitate high hardware resource requirements for FPGAs. FPGA development and debugging costs are also significantly higher than those of PLCs.

[0060] 3) Based on traditional DCS (Distributed Control System) integration solution The DCS system is used to replace the PLC, and the built-in control module is used to realize the integration of data acquisition, analysis and execution, and the traditional PID algorithm is combined to optimize the equipment operating parameters.

[0061] The DCS's built-in control modules enable integrated control and data management. The DCS distributed control system supports collaborative control of multiple devices, offering excellent scalability for expanding automated inspection points for industrial equipment and modularly configuring automated inspection equipment. However, the DCS system scan cycle is typically 100ms to 1s, which cannot meet real-time requirements. The DCS system's control system relies on PID (Proportional-Integral-Derivative), which is primarily used for motion control rather than fault prediction. PID can only provide responsive motion control after a fault occurs, but cannot provide early warning. The cost of setting up a DCS system is also higher than that of a PLC solution.

[0062] Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A method for detecting the status of industrial equipment, characterized in that: include: Collect some time series characteristic parameters of industrial equipment, construct a characteristic parameter change curve of the industrial equipment over time based on the partial time series characteristic parameters, obtain a historical degradation curve, and obtain the complete time series characteristic parameters of the industrial equipment based on the historical degradation curve; Based on the complete time series characteristic parameters of the industrial equipment, a health mapping model is constructed based on the complete time series characteristic parameters to output the health index of the industrial equipment; Use multiple sensors to collect characteristic parameters of industrial equipment in real time; Input characteristic parameter data into the health mapping model to output the health index; If the health index deviates from the preset normal value range, it is judged that the status of the industrial equipment is abnormal.

2. The method for detecting the state of industrial equipment according to claim 1, wherein: The method of constructing a characteristic parameter change curve of industrial equipment over time based on some time series characteristic parameters to obtain a historical degradation curve includes: Obtain the timing characteristic parameter sequence of industrial equipment under normal conditions; Based on the time series characteristic parameter sequence, the least square method is used to construct a curve showing the characteristic parameter changes over time, and a historical degradation curve is obtained.

3. The method for detecting the state of industrial equipment according to claim 2, wherein: The method of constructing a curve of characteristic parameters changing with time using the least squares method includes: The equation defining the curve is T=at+b, where T is the characteristic parameter and t is the time. n data points , i-1,2,3...n, Represents the characteristic parameters of the i-th data point at time t; Constructing the objective function ; Calculate the partial derivatives of S with respect to a and b respectively and set them to 0, and we get the following system of equations: Solving the system of equations yields: Substituting the values of a and b into T=at+b, we can obtain the curve of characteristic parameters changing with time.

4. The method for detecting the state of industrial equipment according to claim 1, wherein: The health mapping model is an LSTM network.

5. The method for detecting the state of industrial equipment according to claim 1, characterized in that: The method of using multiple sensors to collect characteristic parameters of industrial equipment in real time includes: Use multiple sensors to collect characteristic parameters of industrial equipment and transmit them to edge computing devices; The edge computing device uses redundant sensor cross-validation technology to correct the multiple characteristic parameters to obtain corrected data; after filtering the corrected data, the feature information is extracted using a feature extraction algorithm; the feature information is compressed to obtain final characteristic parameter data.

6. A status detection system for industrial equipment, characterized in that: include: A data processing module is used to collect partial time series characteristic parameters of industrial equipment, construct a characteristic parameter change curve of the industrial equipment over time based on the partial time series characteristic parameters, obtain a historical degradation curve, and obtain the complete time series characteristic parameters of the industrial equipment based on the historical degradation curve; Based on the complete time series characteristic parameters of the industrial equipment, a health mapping model is constructed based on the complete time series characteristic parameters to output the health index of the industrial equipment; The device perception module is used to collect characteristic parameters of industrial equipment in real time using multiple sensors; The data processing module is further configured to input characteristic parameter data into a health mapping model to output a health index; if the health index deviates from a preset normal value range, it is determined that the state of the industrial equipment is abnormal.

7. The system according to claim 6, characterized in that The device perception module includes: A data acquisition unit, used to collect multiple characteristic parameters of industrial equipment using a programmable logic controller system integrating multiple sensors and then output them; The edge computing unit is used to correct the characteristic parameters using redundant sensor cross-validation technology to obtain correction data; after filtering the correction data, feature information is extracted using a feature extraction algorithm; and the feature information is compressed to obtain final characteristic parameter data.

8. The system according to claim 7, characterized in that The EtherCAT protocol is used to transmit characteristic parameters between the edge computing unit and the data processing module.

9. The system according to claim 6, wherein: Also includes: The early warning module is used to calculate the health deviation of the equipment and issue graded early warnings based on the health deviation; The calculation of the health deviation is: in, is the health index, The middle value of the normal range of health index, This is the minimum value of the normal range of health index.

10. The system according to claim 6, wherein: The characteristic parameters include: vibration, temperature, current, and pressure.

Images