Digital intelligent fault diagnosis method for cable-driven key component of large radio telescope

Abstract

The invention discloses a digital intelligent fault diagnosis method for a cable-driven key component of a large radio telescope. The method is composed of a digital state monitoring part and an intelligent fault diagnosis part, the digital state monitoring part is used for monitoring the states of key components such as a cable-driven high-voltage cable connector, a through-wall shaft and a speed reducer, and then operation state digitization is achieved; according to the intelligent fault diagnosis, a fault threshold value is extracted through historical operation data, and intelligent fault diagnosis is carried out on the equipment state based on the fault threshold value. According to the invention, state digitization is realized through real-time monitoring of the operation state of the cable-driven key component, and intelligent fault diagnosis is carried out based on historical operation data, so that digital intelligent condition-based maintenance of the cable-driven key component is realized, and powerful guarantee is provided for high-quality operation of a telescope.

Description

Technical Field

[0002] This invention relates to the field of digital maintenance technology, and in particular to a digital fault diagnosis method for a key component of the cable drive of a large radio telescope. Background Technology

[0004] The cable drives of the Five-hundred-meter Aperture Spherical Radio Telescope (FAST) are the driving devices that enable high-precision operation of the telescope's feed cabin, totaling six sets. The cable drive is a large electromechanical system. Its mechanical components mainly include the through-wall shaft, reducer, drum, brake, and drum coupling; its electrical components mainly include servo motors, frequency converters, resistor cabinets, control cabinets, filters, and encoders. Structurally and spatially, the six cable drives are independent systems; a failure in one set will not affect the reliability of the others. However, from the perspective of practical use and safety design requirements, the six cable drives constitute a large, flexible, parallel robotic system. Only when all six cable drives are operational can the feed support system operate normally; otherwise, the telescope will be unable to conduct observations. Therefore, reliable operation of the cable drives is fundamental to the telescope's normal observation, and their maintenance is one of the most important aspects of telescope maintenance.

[0005] In the early design, the cable drive monitoring included the voltage, current, output torque, and speed of the servo motor, while other critical components in the drive system were not monitored. However, as the equipment aged, some critical components experienced unexpected early failures. For example, during early maintenance, several sudden failures with significant safety hazards occurred, including loose high-voltage cable connectors causing discharge, abnormal temperature rise in the high-voltage filter, wall-mounted shaft failure, and reducer failure, posing a major threat to the telescope's safety. Therefore, early fault diagnosis research on these critical components is urgently needed to detect and resolve these failures in their early stages, preventing them from escalating into major safety hazards and impacting the telescope's high-quality operation.

[0006] The information disclosed in this background section is intended only to enhance the understanding of the general background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0008] The purpose of this invention is to provide a digital and intelligent fault diagnosis method for key components of the cable drive of a large radio telescope, so as to solve the technical problems existing in the prior art.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A digital fault diagnosis method for key components of cable-driven systems in large radio telescopes includes the following steps:

[0011] S1. Data Acquisition: Real-time status data of high-voltage cable connectors, high-voltage filters, wall-mounted shafts, and reducers in the FAST cable drive room are acquired through sensors.

[0012] S2. Data transmission: The PLC reads the real-time data from the sensor and designs a photoelectric conversion module to convert the data signal into an optical signal. Using the OPC communication protocol, the real-time status data is transmitted back to the host computer via optical fiber.

[0013] S3, Data Storage: Includes real-time monitoring data preprocessing, real-time operation status data visualization, historical status data query and visualization, and historical status database module;

[0014] S4. Historical Data Analysis: First, historical status data is cleaned, summarized, and collected, and then combined with human empowerment to build a historical operating status database that includes a fault maintenance database, a fault status database, and a normal operating status database.

[0015] S5. Real-time status assessment: Based on the existing fault status database and normal operation status database, machine learning methods are used to train fault prediction models and life prediction models, and real-time monitoring data is used to assess the equipment operating status through the trained fault models and life prediction models.

[0016] S6. Intelligent Maintenance Decision: Employing the dynamic bag-of-words model (DBOW) machine learning method, the system inputs the equipment's operational status assessment results into the fault maintenance database and combines this with manual correction to automatically generate the final maintenance decision.

[0017] Furthermore, in step S1, the high-voltage cable joint monitoring uses an infrared thermal imager from Udeli Market for real-time monitoring; the high-voltage filter monitoring uses a Redsmay patch-type temperature sensor for real-time monitoring; and the wall-mounted shaft and reducer monitoring uses a triaxial temperature and vibration sensor from Kenda Renke for real-time monitoring.

[0018] Furthermore, in step S5, the fault prediction model adopts the K-Nearest Neighbors (KNN) algorithm. The KNN algorithm selects K training samples that are closest to the test sample in Euclidean space based on the similarity of the Euclidean distance between samples, and its regression value is the mean or weighted value of the K nearest neighbor samples. The specific steps are as follows:

[0019] (1) Yes Let the training samples be represented as... Each training sample can be represented as , Then the training samples , and the sample to be tested Euclidean distance between It can be represented as:

[0020] (1)

[0021] In the formula and As an auxiliary variable, and For monitoring variables;

[0022] (2) Based on the sample to be tested With all training samples Find the Euclidean distance between them. The K nearest neighbor training samples , ;

[0023] (3) Calculate the test sample estimated values ​​of monitoring variables The calculation formula is:

[0024] (2).

[0025] Furthermore, in step S5, the lifetime prediction model employs an improved recurrent neural network (RNN) algorithm; it combines the Long Short-Term Memory (LSTM) network model within the RNN with the Autoregressive Conditional Heteroskedasticity (ARCH) model; the state update equation for the LSTM model is as follows:

[0026] (3)

[0027] (4)

[0028] (5)

[0029] (6)

[0030] (7)

[0031] The ARCH model is represented as follows:

[0032] (8).

[0033] Furthermore, the calculation formula for the DBOW model in step S6 is as follows:

[0034] (9).

[0035] By adopting the above technical solution, the present invention has the following beneficial effects:

[0036] This invention achieves digital status monitoring of the operating status of key cable-driven components and performs intelligent fault diagnosis based on historical operating data, thereby realizing intelligent condition-based maintenance of key cable-driven components and providing strong support for the high-quality operation of the telescope. Attached Figure Description

[0038] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0039] Figure 1 This is a system architecture diagram of the intelligent fault diagnosis method for key components of the cable drive of a large radio telescope provided in an embodiment of the present invention;

[0040] Figure 2 The flowchart illustrates the operation of the intelligent fault diagnosis method for key components of the cable drive of a large radio telescope provided in this embodiment of the invention. Detailed Implementation

[0042] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0044] like Figure 1-2 As shown, this invention proposes a digital and intelligent fault diagnosis method for key components of the cable drive system in large radio telescopes. This method includes: digital condition monitoring and intelligent fault diagnosis. Digital condition monitoring comprises three parts: data acquisition, data transmission, and data storage; intelligent fault diagnosis comprises three parts: historical data analysis, real-time condition assessment, and intelligent maintenance decision-making. The specific implementation steps are as follows:

[0045] The first step is data acquisition. Data acquisition mainly involves using sensors to collect real-time status data of high-voltage cable connectors, high-voltage filters, wall-mounted shafts, and reducers in the FAST cable drive room. Specifically, high-voltage cable monitoring utilizes an infrared thermal imager from UDELI Market; high-voltage filter monitoring uses a Redsmite patch-type temperature sensor; and wall-mounted shaft and reducer monitoring uses a JianDa RenKe triaxial temperature and vibration sensor.

[0046] The second step is data transmission. Data transmission mainly involves the PLC reading real-time data from the sensors, and a photoelectric conversion module is designed to convert the data signals into optical signals. Using the OPC communication protocol, the real-time status data is transmitted back to the host computer via optical fiber.

[0047] The third step is data storage. Data storage includes modules such as real-time monitoring data preprocessing, real-time operational status data visualization, historical status data query and visualization, and a historical status database.

[0048] The fourth step is historical data analysis. First, the historical status data is cleaned, summarized, and collected. Then, with the help of human intervention, a historical operational status database is constructed, which includes a fault maintenance database, a fault status database, and a normal operation status database.

[0049] The fifth step is real-time status assessment. Based on the existing fault status database and normal operation status database, machine learning methods are used to train fault models and life prediction models, and real-time monitoring data is used to assess the equipment's operating status through the trained fault models and life prediction models.

[0050] Fault prediction employs the K-Nearest Neighbors (KNN) algorithm. KNN is a classic non-parametric method commonly used to solve classification and regression problems. Its basic idea is to select the K nearest training samples in Euclidean space based on the similarity of the samples between them using Euclidean distance. The regression value is the mean or weighted sum of the K nearest neighbor samples. The specific steps are as follows:

[0051] (1) Yes Let the training samples be represented as... Each training sample can be represented as , Then the training samples , and the sample to be tested Euclidean distance between It can be represented as:

[0052] (1)

[0053] In the formula and As an auxiliary variable, and For monitoring variables.

[0054] (2) Based on the sample to be tested With all training samples Find the Euclidean distance between them. The K nearest neighbor training samples , .

[0055] (3) Calculate the test sample estimated values ​​of monitoring variables The calculation formula is:

[0056] (2)

[0057] Lifespan prediction employs an improved recurrent neural network (RNN) algorithm. RNNs offer advantages over traditional feedforward neural networks when processing data that changes over time. The algorithm combines the most commonly used Long Short-Term Memory (LSTM) network model with an Autoregressive Conditional Heteroskedasticity (ARCH) model. The state update equation for the LSTM model is as follows:

[0058] (3)

[0059] (4)

[0060] (5)

[0061] (6)

[0062] (7)

[0063] The ARCH model can be represented as:

[0064] (8).

[0065] Step 6, Intelligent Maintenance Decision. Using machine learning methods such as the Dynamic Bag-of-Words (DBOW) model, the equipment's operational status assessment results are fed into the fault maintenance database, and combined with manual adjustments, the final maintenance decision is automatically generated. The DBOW model can be represented as:

[0066] (9).

[0067] In this application, the monitoring objects include high-voltage cable joints, high-voltage filters, wall-mounted shafts, and speed reducers in the cable drive room. Among them, high-voltage cable monitoring is carried out by a thermal imaging monitor; high-voltage filter monitoring is carried out by a patch-type temperature sensor; wall-mounted shaft monitoring is carried out by a triaxial temperature and vibration sensor; and speed reducer monitoring is carried out by a vibration sensor.

[0068] In this application, the PLC reads sensor data and transmits it to the host computer via optical fiber using OPC communication. The application utilizes machine learning technology to analyze historical data and obtain a fault characteristic database and a normal operating status database for known faults. Early fault thresholds and severe fault thresholds are extracted from the fault database, and real-time operating data is compared with these thresholds to obtain equipment status assessment results. The normal operating life curve is extracted from the normal operating data database, and real-time operating data is compared with the life curve to obtain a predicted remaining lifespan. A manual approach is used to provide maintenance plans for each fault type and different fault status in the fault database, and equipment maintenance decisions are automatically generated based on the real-time status assessment results.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for diagnosing faults in key components of a large radio telescope cable drive system, characterized by, Comprising the following steps: S1, data acquisition: real-time state data acquisition of high-voltage cable joints, high-voltage filters, wall shafts and speed reducers in the FAST cable drive machine room through sensors; S2, data transmission: real-time data of sensors are read by PLC, and optical-electric conversion module is designed to convert data signals into optical signals, OPC communication protocol is adopted, and real-time state data is transmitted back to the host computer through optical fiber; S3, data storage: including real-time monitoring data preprocessing, real-time running state data visualization, historical state data query and visualization, and historical state database module; S4, historical data analysis: first, the historical state data is cleaned, summarized and collected, and combined with artificial empowerment to build a historical running state database including fault maintenance database, fault state database and normal running state database; S5, real-time state evaluation: based on the existing fault state database and normal running state database, machine learning method is used to train fault prediction model and life prediction model, and real-time monitoring data is evaluated by trained fault model and life prediction model; S6, intelligent maintenance decision: using dynamic bag-of-words model-DBOW machine learning method, the running state evaluation result of the equipment is brought into the fault maintenance database, and combined with artificial empowerment correction, the final maintenance decision is automatically generated.

2. The method according to claim 1, wherein the method is characterized by, In step S1, the high-voltage cable joint monitoring uses the infrared thermal imager of Oudeli market for real-time monitoring; the high-voltage filter monitoring uses the patch temperature sensor of Hongsimai for real-time monitoring; the wall shaft and speed reducer monitoring uses the three-axis temperature and vibration sensor of Jianda Riken for real-time monitoring.

3. The method of claim 1, wherein the method is characterized by: In step S5, the fault prediction model uses K nearest neighbor-KNN algorithm; KNN algorithm selects K nearest neighbor samples in Euclidean space according to the similarity of sample Euclidean distance, and the regression value is the mean or weighted value of K nearest neighbor samples: the specific steps are as follows: (1) Yes Let the training samples be represented as... Each training sample can be represented as , Then the training samples , and the sample to be tested Euclidean distance between It can be represented as: （1）； wherein and are auxiliary variables, and are monitoring variables; (2) According to the sample to be tested The Euclidean distance between all training samples Find the nearest K nearest neighbor training samples , ;​ (3) Calculate the monitoring variable estimate value of the sample to be tested The calculation formula is:​ （2）。 4. The method of claim 1, wherein the method is characterized by: In step S5, the life prediction model uses improved recurrent neural network-RNN algorithm; it is completed by combining long short-term memory network model-LSTM in RNN with autoregressive conditional heteroscedasticity model-ARCH; wherein the state update equation of LSTM model is as follows: （3）； （4）； （5）； （6）； （7）； The ARCH model is expressed as: （8）。 5. The method of claim 1, wherein the method is characterized by: The calculation formula of DBOW model in step S6 is: （9）。

Images