Urban rail vehicle depot fixed type jacking machine fault diagnosis method based on digital twinning

Abstract

The present application relates to a kind of based on digital twinning urban rail vehicle base fixed car lift fault diagnosis method.The existing diagnostic method has the problems of poor diagnosis efficiency and low accuracy.This method establishes the digital twin of fixed car lift;Real-time operation data is collected by installing sensors and transmitted to the digital twin for synchronization;The correlation mechanism of each component of fixed car lift is sorted out, and the SDG model of fixed car lift is constructed and divided into multiple subsystems;Based on the subsystem, a fault monitoring model based on PCA algorithm is constructed, and a test algorithm is introduced to detect the health status of fixed car lift;By detecting the health status, it is judged whether there is a fault.The present application improves the fault diagnosis efficiency of fixed car lift through digital twinning technology, can monitor the health status in real time, and can diagnose faults and monitor health through simulation of virtual model and under the driving of twin data, with higher accuracy.

Description

Technical Field

[0001] This invention relates to the field of rail transit vehicle maintenance technology, specifically to a fault diagnosis method for a fixed vehicle jacking machine in an urban rail vehicle depot based on digital twins. Background Technology

[0002] Fixed bogie lifting machines are among the most important pieces of equipment in rail transit train maintenance. Their high reliability, high synchronization, high stability, and high safety are crucial for train maintenance, directly impacting the efficiency of maintenance and the safety of maintenance personnel. As the primary mode of transportation for rail transit, the reliability and safety of trains directly affect passenger safety and property, as well as social impact. To ensure good train performance, routine maintenance is essential, and the condition of the maintenance equipment is a vital guarantee for optimal train operation. Fixed bogie lifting machines are crucial equipment for both temporary and major train repairs, primarily consisting of bogie lifting devices, car body lifting devices, motor drive mechanisms, and control systems. Their main purpose is to lift or lower entire trains, single cars, or multiple cars while the train is either uncoupled or partially uncoupled. This is used for major or minor repairs of the entire train, replacing one or more bogies, or disassembling and assembling equipment such as undercarriage electrical boxes.

[0003] If a fixed jacking machine malfunctions during operation, it will severely impact train maintenance time and may even lead to major safety accidents (because jacking operations place high demands on the synchronization and stability of each lifting device on the fixed jacking machine; for example, the flatness of the four support points of the car body lifting column must be ≤4mm. If the synchronization and stability of each support point cannot meet the requirements, the train is prone to overturning on the jacking machine), resulting in vehicle damage and casualties. Ensuring the reliable, synchronized, stable, and safe operation of the fixed jacking machine is crucial. However, traditional fault diagnosis methods such as data acquisition, point monitoring, manual inspection, and experience-based fault diagnosis are no longer sufficient to meet the demands of the rapid development of intelligent rail transit in terms of diagnostic efficiency and accuracy.

[0004] Therefore, it is necessary to propose new fault diagnosis methods to overcome the above-mentioned shortcomings. Summary of the Invention

[0005] The purpose of this invention is to provide a fault diagnosis method for fixed-type vehicle lifting machines in urban rail vehicle depots based on digital twins, so as to solve the problems of poor diagnostic efficiency and low accuracy of existing diagnostic methods.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A fault diagnosis method for fixed-type vehicle lifting machines in urban rail vehicle depots based on digital twins, the method comprising:

[0008] Establish a digital twin of the fixed vehicle jacking machine;

[0009] Sensors are installed on various components of the fixed vehicle jacking machine to collect real-time operating data, which is then transmitted to the digital twin of the fixed vehicle jacking machine for synchronization.

[0010] The interrelationship mechanism of the operation of each component of the fixed car jacking machine was analyzed, and the SDG model of the fixed car jacking machine was constructed and divided into multiple subsystems;

[0011] A fault monitoring model based on the PCA algorithm is constructed based on the subsystem, and a verification algorithm is introduced to detect the health status of the fixed vehicle jacking machine.

[0012] By detecting the health status, it can be determined whether there is a malfunction.

[0013] Furthermore, establish a digital twin of the fixed-type vehicle jacking machine, including:

[0014] Establish BIM models of the maintenance depot's civil engineering, drainage, and lighting, and overlay and merge them based on a unified reference base point to obtain the maintenance depot's BIM model;

[0015] Establish BIM models of each component of the fixed vehicle jacking machine, and overlay and merge them based on a unified reference base point to obtain the BIM model of the fixed vehicle jacking machine;

[0016] Based on a unified reference point, the BIM model of the fixed vehicle rack machine is overlaid on the BIM model of the maintenance depot to obtain a combined BIM model;

[0017] Lightweighting of the composite BIM model;

[0018] In the composite BIM model, motion behavior rules that are mapped to each other are established for the related components in the fixed vehicle jacking machine.

[0019] Obtain a digital twin of the fixed vehicle jacking machine.

[0020] Furthermore, sensors are installed on various components of the fixed jacking machine to collect real-time operational data, which is then transmitted to the fixed jacking machine's digital twin for synchronization. This includes:

[0021] Sensors are installed on various components of the fixed vehicle frame machine;

[0022] The collected real-time operation data is transmitted to the digital twin of the fixed vehicle jack.

[0023] The physical fixed vehicle jacking machine communicates with its digital twin. The digital twin updates its status based on real-time operating data and displays it in three dimensions.

[0024] The digital twin of the fixed vehicle jacking machine simulates and optimizes the operating status of the fixed vehicle jacking machine based on real-time operating data, and performs real-time control of the fixed vehicle jacking machine.

[0025] Furthermore, the operational mechanisms of the various components of the fixed car jacking machine were analyzed, and an SDG model of the fixed car jacking machine was constructed and divided into multiple subsystems, including:

[0026] Based on the operational relationships, fault phenomena, and causes of various components of the fixed vehicle jacking machine, the fault propagation path is analyzed to obtain the correlation mechanism of the operation of each component.

[0027] Based on the association mechanism, a fixed vehicle rack SDG model is constructed;

[0028] The SDG model of the fixed vehicle jack is divided into multiple subsystems to simplify the SDG model.

[0029] Furthermore, a fault monitoring model based on the PCA algorithm is constructed based on the subsystem, and a verification algorithm is introduced to detect the health status of the fixed car-mounting machine, including:

[0030] The real-time data collected by each sensor is normalized with historical data, and the principal components are obtained by calculating the covariance matrix and then compared with the historical data.

[0031] If it is historical data, it is stored in the threshold database. If it is not historical data, it is squared and the result is compared with the threshold. If it is greater than or equal to the threshold, the principal component and the result are input into the fault reasoning model for fault location. If it is less than the threshold, the principal component and the result are compared for mutation detection. If a mutation point exists, the principal component and the result are input into the fault reasoning model for fault location. If no mutation point exists, no fault is displayed.

[0032] Furthermore, by detecting the health status and determining whether there is a fault, the set of original faults is obtained using the SDG model of the fixed vehicle jack.

[0033] Furthermore, fault types are divided into sudden faults and cumulative faults;

[0034] A subsystem of the fixed scaffolding machine SDG model experienced a sudden and severe performance degradation due to stagnation, indicating a sudden and severe abnormal state.

[0035] The performance of a certain subsystem of the fixed scaffolding machine SDG model gradually declined due to aging. In the early stage, the abnormal state was vague and slight, and the overall operation was stable. The fault symptoms only became obvious in the middle and late stages, which is a cumulative fault.

[0036] Furthermore, the method also includes:

[0037] After establishing and synchronizing a digital twin of the fixed jacking machine, the digital twin is trained using historical operational data, including:

[0038] The system acquires operational data of various components of the fixed vehicle jack over a period of time, collected by sensors, to obtain historical operational data.

[0039] Perform feature calculations and filtering on historical operational data;

[0040] The entire lifecycle of the fixed vehicle rack is divided into different stages, and then the key degradation stages are characterized by feature calculation and screening, and data augmentation is performed.

[0041] The selected features are fused based on the combined BIM model, and the LSTM model with bias enhancement is used to learn the HS-RUL mapping relationship, thereby completing the training of the digital twin of the fixed vehicle jack.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] This invention provides a fault diagnosis method for fixed car-lifting machines in urban rail vehicle depots based on digital twins. By using digital twin technology, the method improves the fault diagnosis efficiency of fixed car-lifting machines, enables real-time monitoring of their health status, and performs fault diagnosis and health monitoring of fixed car-lifting machines through virtual model simulation and driven by twin data, resulting in higher accuracy.

[0044] This invention provides a fault diagnosis method for fixed car-lifting machines in urban rail vehicle depots based on digital twins. Based on digital twin technology, it utilizes full lifecycle data resources to establish a maintenance depot and a digital twin of the fixed car-lifting machine, realizing real-time mapping, simulation, and modeling of the physical entity and the virtual model. This achieves optimized resource allocation and management, optimized production planning, and coordinated operation of production factors, realizing full lifecycle digital management of the fixed car-lifting machine. It can create maximum benefits with minimal cost input, thereby improving overall production and maintenance efficiency. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0046] Figure 1 This is a flowchart of the method of the present invention.

[0047] Figure 2This is a flowchart of S3 and S4 of the present invention.

[0048] Figure 3 This is a schematic diagram of Example 1.

[0049] Figure 4 This is a schematic diagram of Example 2. Detailed Implementation

[0050] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0051] It should be noted that similar reference numerals and letters indicate similar items; therefore, once an item is defined in one embodiment, it does not need to be further defined and explained in subsequent embodiments. Furthermore, the terms "comprising" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0052] It should also be noted that although the order of steps is mentioned in the method description, in some cases, steps may be performed in a different order than that described here, and this should not be interpreted as a restriction on the order of steps.

[0053] This invention provides a fault diagnosis method for fixed vehicle jacking machines in urban rail vehicle depots based on digital twins. It can be applied to the maintenance of subway and high-speed rail vehicles, effectively reducing maintenance costs, improving maintenance efficiency, and mitigating safety risks.

[0054] like Figure 1 This method specifically includes:

[0055] S1: Establish a digital twin of the fixed vehicle jacking machine. This includes:

[0056] S101: Using 3D visualization software such as Revit, a 1:1 BIM model of the maintenance depot's civil engineering, drainage, and lighting is created in the same coordinate system. The models are then overlaid and merged based on a unified reference point to obtain the maintenance depot's BIM model, laying the foundation for later use of digital twin technology to diagnose faults in fixed vehicle racking machines.

[0057] S102: Use 3D visualization software such as Revit (or SolidWorks) to create 1:1 BIM models of each component of the fixed vehicle jacking machine, and then overlay and merge them based on a unified reference point to obtain the BIM model of the fixed vehicle jacking machine.

[0058] The components include the body support lifting device (also known as the body lifting column), the bogie support lifting device (also known as the bogie lifting column), the motor, the gearbox (also known as the reducer), the coupling, the motor and its mounting feet being loose, the bearing nuts being worn, the limit device, etc.

[0059] S103: Based on a unified reference point, the fixed vehicle rack BIM model is overlaid onto the maintenance depot BIM model to obtain a combined BIM model.

[0060] S104: To improve the smoothness of visualization, the combined BIM model is imported into BIMFACE software for lightweight processing.

[0061] S105: Using Unity 3D software, in the combined BIM model, establish mutual mapping motion behavior rules for related components in the fixed vehicle jacking machine, so that the motion states of each component collected by each sensor can be matched one by one and mapped to achieve consistency between the motion state of the virtual fixed vehicle jacking machine digital twin and the physical entity.

[0062] S106: Obtain a digital twin of the fixed-type vehicle jack.

[0063] S2: Sensors are installed on various components of the fixed car jacking machine to collect real-time operating data and transmit it to the fixed car jacking machine's digital twin for synchronization. This includes:

[0064] S201: Sensors are installed on various components of the fixed jacking machine to collect real-time operating data and obtain status information for each component. Examples include position sensors on the bogie lifting column, body lifting column, voltage, current, and temperature sensors on the motors, image recognition sensors (cameras with edge computing image recognition) near the motor mounting bolts to prevent loosening, wear sensors on the bearing nut, limit device sensors, and acceleration sensors on the couplings and motors. Furthermore, the control system of the fixed jacking machine can use a Siemens S7-1500 series PLC controller, but is not limited to this.

[0065] S202: Transmit the collected real-time operation data to the digital twin of the fixed jacking machine. To achieve real-time monitoring and collection of the fixed jacking machine's operation data, the Ethernet Profinet communication protocol is used to realize communication between the physical fixed jacking machine and its digital twin.

[0066] S203: The fixed jacking machine entity and its digital twin communicate with each other. The digital twin updates its status based on real-time operating data and displays it in 3D. The digital twin maintains consistency with the fixed jacking machine entity at all times, realistically reproducing any state changes that occur in the fixed jacking machine entity through 3D visualization.

[0067] S204: The digital twin of the fixed vehicle jacking machine simulates and optimizes the operating status of the fixed vehicle jacking machine entity based on real-time operating data, and performs real-time control of the fixed vehicle jacking machine entity.

[0068] The information of the fixed-type car frame that is perceived and synchronized by the physical and digital twins includes: bogie height, car body height, bogie lifting speed, car body lifting speed, nut position, lifting asynchronous fault, upper limit fault, lower limit fault, nut wear fault, nut detachment fault, sensor fault, overload fault, etc.

[0069] S3: Analyze the interrelationships of the various components of the fixed car jacking machine, construct an SDG model of the fixed car jacking machine, and divide it into multiple subsystems. For example... Figure 2 ,include:

[0070] S301: Based on the operational relationships, fault phenomena, and causes of the various components of the fixed vehicle jacking machine, the fault propagation path is analyzed to obtain the correlation mechanism of the operation of each component. Unmeasurable state parameters are avoided as much as possible without affecting the positioning results to facilitate the subsequent construction of the SDG model.

[0071] S302: Based on the association mechanism, construct the fixed vehicle rack SDG model.

[0072] S303: The fixed car-lifting machine consists of numerous subsystems. Its operation mechanism relies on the interaction and coordination between these subsystems to complete car body lifting or lowering tasks. This mechanism makes fault propagation possible within and between the subsystems of the fixed car-lifting machine; that is, an anomaly at one point can lead to anomalies at related points. The operation and maintenance of the fixed car-lifting machine needs to consider the interrelationships between and within each subsystem as much as possible to quickly and accurately resolve the root cause of the fault. Simply addressing derivative faults cannot fundamentally solve the fault source; frequent derivative faults will delay normal production, and long-term aging at a certain point can even lead to the expansion and deterioration of the entire equipment's malfunction. Therefore, the SDG model of the fixed car-lifting machine is divided into multiple subsystems to simplify the SDG model and address the root cause of the fault.

[0073] To address the challenges of complex structure and high fault uncertainty in fixed car-laying machines, a method for locating the intrinsic faults of these machines based on MPCA-DSDG is proposed. This method primarily comprises four steps: ① clarifying equipment correlation mechanisms; ② constructing a fault reasoning model; ③ building a fault monitoring system; and ④ real-time fault monitoring and location. Specifically:

[0074] ① Equipment association mechanism analysis. The association mechanism of the fixed vehicle jack, including the operation relationship of components, fault phenomena and causes, was analyzed. In addition, unmeasurable state parameters were avoided as much as possible without affecting the positioning results, so as to facilitate the subsequent construction of the SDG model.

[0075] ② Fault Reasoning Model Construction. Based on the fixed-type vehicle jack association mechanism identified in step ①, an equipment SDG model is constructed. The equipment is divided into subsystems according to the model to facilitate the subsequent construction of a multi-subsystem monitoring system. The reasoning method of SDG is improved by combining the characteristics of early anomalies.

[0076] ③ Equipment monitoring system construction. Based on the systems defined in step ②, fault monitoring models based on the PCA algorithm are constructed, and a verification algorithm is introduced to detect changes in the health status trend of the system, thereby narrowing the fault scope to the system level and identifying anomalies in the early stages of the fault.

[0077] ④ Real-time fault monitoring and location of equipment. Based on the real-time data of system segmentation in step ③, the health status of the system is calculated. By detecting the trend of the health status, it is determined whether there is a fault in the system. Finally, the set of original faults is obtained by reasoning through the SDG model constructed in step ②.

[0078] S4: Based on the subsystem, a fault monitoring model based on the PCA algorithm is constructed, and a verification algorithm is introduced to detect the health status of the fixed vehicle jacking machine. For example... Figure 2 ,include:

[0079] Data is normalized by combining real-time and historical data collected from various sensors. Principal components are calculated using the covariance matrix and compared with historical data. If the data is historical, it is stored in a threshold database. If not, it is squared and compared with a threshold. If the result is greater than or equal to the threshold, the principal component and its calculation result are input into the fault reasoning model for fault location. If the result is less than the threshold, the principal component and its calculation result are compared for abrupt change detection. If an abrupt change is found, the principal component and its calculation result are input into the fault reasoning model for fault location. If no abrupt change is found, no fault is displayed.

[0080] The PCA-based method determines the system T. 2Fault monitoring of the system is achieved by checking whether a statistic exceeds a given threshold Ta. This method can effectively identify sudden faults with severe severity and also has the ability to identify the later stages of cumulative faults. However, this method has low accuracy in identifying early anomalies in cumulative faults because T... 2 The statistic shows a small upward trend in the early stages of the fault and does not reach the threshold Ta.

[0081] Health assessment models are constructed based on fixed state parameters. However, because the system's state differs at different operational stages, the T values ​​at different stages will vary. 2 The statistical amplitudes differ, so simply reducing Ta cannot solve the above problem. However, in the early stages of a system failure, T... 2 The statistic will show a gradually increasing trend, a phenomenon that allows for monitoring T... 2 The trend of statistical changes can be used to identify the possibility of early anomalies. Because of inherent measurement biases in the sensor itself, state parameters fluctuate within a certain range, leading to T... 2 The statistics also fluctuate; the order series aims to measure whether the current value and historical process values ​​are on the same order of magnitude, while T... 2 Fluctuations in the statistic can interfere with the accuracy of the ordered series; therefore, a threshold t needs to be defined to enhance the versatility of the ordered series. 2 The order sequence {r2,r3,…,rn} corresponding to the statistical sequence {t1,t2,…,tn} is also calculated according to the relevant formula. Furthermore, to suppress T due to noise... 2 A sudden increase followed by a sharp drop in statistical values ​​can lead to misjudgments of faults. Points showing three consecutive increases in the order column are considered as T-values ​​caused by early anomalies. 2 A mutation point in the statistics indicates an anomaly in the system.

[0082] This step narrows down the fault scope to the system level and identifies anomalies at an early stage. Equipment consists of numerous components; for example, the motor and transmission beam are closely connected and can therefore be classified into the same system. However, the temperature of the load-bearing nut is less related to the motor and can be classified into a different system. Since equipment anomaly detection methods are derived through the analysis of multidimensional data, classifying closely related components into the same system is reasonable.

[0083] The PCA algorithm aims to detect early anomalies in equipment. After detecting an early anomaly, it uses a fault reasoning model based on SDG to infer the cause of the fault.

[0084] S5: By detecting the health status, determine whether there is a fault, and then use the fixed scaffolding machine SDG model to obtain the set of source faults.

[0085] Faults can be categorized into primary faults and derivative faults based on their causes. Taking the entry of a foreign object into a lifting beam as an example, the lifting beam's cessation of operation could potentially cause the motor to jam. Since the motor jamming is not due to its own inherent fault, the entry of the foreign object into the lifting beam is the primary fault, and the motor jamming is the derivative fault. Due to the ambiguity of this relationship, reverse reasoning may reveal multiple primary faults. Although some of these may be spurious solutions, it is still necessary to ensure the completeness of the faults and include them in the primary fault category for inspection. This is the origin of the term "set." The SDG model first infers based on whether nodes are abnormal, then performs reverse reasoning based on directed edges, and finally uses the resulting set of primary faults for forward reasoning to eliminate spurious solutions.

[0086] The fault types diagnosed by this method are divided into sudden faults and cumulative faults. A subsystem of the fixed scaffolding machine SDG model experienced a sharp decline in performance due to stagnation, with obvious and severe abnormal conditions, which is a sudden fault. Another subsystem of the fixed scaffolding machine SDG model experienced a gradual decline in performance due to aging, with early abnormal conditions being vague and minor, and overall stable operation, until the fault symptoms became obvious in the middle and later stages, which is a cumulative fault.

[0087] This method also includes a digital twin training step. After establishing and synchronizing a digital twin of the fixed jacking machine, the digital twin is trained using historical operating data, including:

[0088] Step 1: Obtain the operating data of each component of the fixed vehicle jacking machine collected by the sensors over a period of time to obtain historical operating data;

[0089] Step 2: Perform feature calculations and filtering on historical operational data;

[0090] Step 3: Divide the entire life cycle of the fixed vehicle rack into different stages, and then perform feature calculation and screening for the key degradation stages and perform data augmentation;

[0091] Step 4: Based on the combined BIM model, the selected features are fused, and an LSTM model with a bias mechanism is used to learn the HS-RUL mapping relationship, thereby completing the training of the digital twin of the fixed vehicle jack.

[0092] The purpose of feature selection is to find the optimal subset of features. By eliminating irrelevant features and selecting those truly relevant to the problem, the upper limit of model accuracy can be increased. Failure to eliminate irrelevant features will cause the HI extraction model to sacrifice information from some effective features in order to retain information from inefficient features; therefore, feature selection is particularly important.

[0093] After training, remaining lifespan prediction can also be performed. For example, the acceleration values ​​for the entire lifespan of a fixed-position jacking machine can be extracted from its historical operating data. These values ​​are obtained from an accelerometer, and acceleration is a physical quantity characterizing the vibration of an object. Vibration signals are time-series signals, and time-domain features are important indicators for measuring the waveform information of time-series signals. There are 16 time-domain features, including maximum value, minimum value, peak value, peak-to-peak interval, absolute average value, square root amplitude, variance, standard deviation, effective value, steepness, skewness, waveform factor, peak factor, impulse factor, margin factor, and residual gap factor. Feature filtering can eliminate irrelevant features and select features relevant to the actual problem.

[0094] This method enables digital interconnection between the digital twin of the fixed vehicle jacking machine and its physical counterpart, applying digital twin technology to fault diagnosis of the fixed vehicle jacking machine to achieve full lifecycle digital management. Simultaneously, based on real-time collection of operational data and analysis of the digital twin, it can not only achieve the purposes of fault analysis and remote management of the vehicle jacking machine, but also improve equipment safety, quickly locate faults, and automatically provide solutions.

[0095] The following are specific examples:

[0096] This method can be used to diagnose faults such as loose mounting feet of motors in fixed car frame systems. Figure 3 The specific method is as follows:

[0097] S1: Establish a digital twin of the fixed vehicle frame machine (including components such as motor mounting feet, bolts, and gaskets).

[0098] S101: Using 3D visualization software such as Revit, a 1:1 BIM model of the maintenance warehouse is established in the same coordinate system, laying the foundation for the later use of digital twin technology to diagnose faults of fixed vehicle racking machines.

[0099] S102: Use 3D visualization software such as Revit (or SolidWorks) to create 1:1 BIM models of each component of the fixed vehicle jacking machine, and then overlay and merge them based on a unified reference point to obtain the BIM model of the fixed vehicle jacking machine.

[0100] S103: Based on a unified reference point, the fixed vehicle rack BIM model is overlaid onto the maintenance depot BIM model to obtain a combined BIM model.

[0101] S104: To improve the smoothness of visualization, the combined BIM model is imported into BIMFACE software for lightweight processing.

[0102] S105: Using Unity 3D software, in the combined BIM model, establish mutual mapping motion behavior rules for related components in the fixed vehicle jacking machine, so that the motion states of each component collected by each sensor can be matched one by one and mapped to achieve consistency between the motion state of the virtual fixed vehicle jacking machine digital twin and the physical entity.

[0103] S106: Obtain a digital twin of the fixed jack motor base and its bolts.

[0104] S2: Sensors are installed on components such as the motor and universal joint of the fixed jacking machine to collect real-time operating data and transmit it to the digital twin of the fixed jacking machine for synchronization.

[0105] Loose motor mounting feet can cause vibration in both the universal joint and the motor. An accelerometer is used to monitor motor vibration, and a photoelectric speed sensor is used to monitor universal joint vibration.

[0106] S3: Analyze the interrelationship mechanism of the operation of each component of the fixed car jacking machine and construct the SDG model of the fixed car jacking machine.

[0107] The motor mounting base may loosen due to reduced bolt tightening force or aging of gaskets. Once the mounting base is loose, the motor will vibrate, causing the coupling to shake. This will result in irregular vibration of the universal joint, and the lifting columns on both sides will rise asynchronously, potentially leading to a subway overturning accident.

[0108] ① Analyze the correlation mechanism of the fixed car jacking machine, including the operating relationship, fault phenomena and causes of components such as anchor bolts, couplings, and universal joints. Under the premise of not affecting the positioning results, try to avoid unmeasurable state parameters so as to facilitate the construction of the subsequent SDG model.

[0109] ② Fault Reasoning Model Construction. Based on the fixed-type vehicle jack association mechanism identified in step ①, an equipment SDG model is constructed. The equipment is divided into subsystems according to the model to facilitate the subsequent construction of a multi-subsystem monitoring system. The reasoning method of SDG is improved by combining the characteristics of early anomalies.

[0110] ③ Equipment monitoring system construction. Based on the systems defined in step ②, fault monitoring models based on the PCA algorithm are constructed, and a verification algorithm is introduced to detect changes in the health status trend of the system, thereby narrowing the fault scope to the system level and identifying anomalies in the early stages of the fault.

[0111] ④ Real-time fault monitoring and location of equipment. Based on the real-time data of system segmentation in step ③, the health status of the system is calculated. By detecting the trend of the health status, it is determined whether there is a fault in the system. Finally, the set of original faults is obtained by reasoning through the SDG model constructed in step ②.

[0112] S4: Construct a fault monitoring model based on subsystems and introduce a verification algorithm to detect the health status of the fixed vehicle jacking machine.

[0113] Motor vibration detection is achieved through machine learning methods. Specifically, motor vibration signal data can be collected, and machine learning algorithms can be used to process and analyze the data, thereby enabling the detection and analysis of motor vibration signals.

[0114] Data preprocessing: The collected raw vibration signal data is preprocessed by filtering, noise reduction and other methods to remove noise and extract effective information.

[0115] Feature extraction: Effective features are extracted from the preprocessed vibration signal data, such as time-domain features (e.g., root mean square, standard deviation, peak-to-peak value, etc.) and frequency-domain features (e.g., power spectral density, peak frequency, etc.), which can be used to describe the characteristics and properties of the vibration signal.

[0116] Feature dimensionality reduction: Feature variables often exhibit varying degrees of correlation, making computation quite challenging. The purpose of PCA preprocessing is to reduce the number of variables by considering their correlations, thus reducing the number of variables to a smaller set. These new variables are not the original variables and are uncorrelated, preserving most of the information from the original data. Subsequent model training also employs the K-nearest neighbor algorithm.

[0117] Data labeling: Labeling the data after feature extraction, that is, labeling each data sample with labels such as "normal" or "abnormal" to facilitate the subsequent training of the K-nearest neighbor algorithm.

[0118] Model training: The model is trained on labeled data to learn the patterns and regularities of vibration signals. The trained machine learning model is then evaluated, for example, by calculating metrics such as accuracy and recall, to assess its performance and effectiveness.

[0119] Loosening fault diagnosis: The trained K-nearest neighbor model is used to classify new vibration signal data to determine whether the motor is loose, as well as the location and degree of loosening.

[0120] Example 2:

[0121] This method can be used to diagnose wear faults in the load-bearing nuts of fixed car frame machines, such as... Figure 4 The specific method is as follows:

[0122] S1: Establish a digital twin of the fixed-type jacking machine load-bearing nut grinding.

[0123] S101: Using 3D visualization software such as Revit, a 1:1 BIM model of the maintenance warehouse is established in the same coordinate system, laying the foundation for the later use of digital twin technology to diagnose faults of fixed vehicle racking machines.

[0124] S102: Use 3D visualization software such as Revit (or SolidWorks) to create BIM models of various components such as the load-bearing nut grinder of the fixed vehicle jack at a 1:1 scale. Then, overlay and merge them based on a unified reference point to obtain the BIM model of the fixed vehicle jack.

[0125] S103: Based on a unified reference point, the fixed vehicle rack BIM model is overlaid onto the maintenance depot BIM model to obtain a combined BIM model.

[0126] S104: To improve the smoothness of visualization, the combined BIM model is imported into BIMFACE software for lightweight processing.

[0127] S105: Using Unity 3D software, in the combined BIM model, establish mutual mapping motion behavior rules for related components in the fixed vehicle jacking machine, so that the motion states of each component collected by each sensor can be matched one by one and mapped to achieve consistency between the motion state of the virtual fixed vehicle jacking machine digital twin and the physical entity.

[0128] S106: Obtain a digital twin of the fixed-type jacking machine load-bearing nut grinding.

[0129] S2: Sensors are installed on components such as the load-bearing nut grinder of the fixed jacking machine to collect real-time operating data and transmit it to the digital twin of the fixed jacking machine for synchronization.

[0130] During long-term lifting operations, most of the weight of a fixed vehicle lifting machine is applied to the load-bearing nut. If the load-bearing nut wears out too much, it will fall off, causing the entire lifting column to detach from the equipment and resulting in a vehicle overturning accident.

[0131] S3: Analyze the interrelationship mechanism of the operation of each component of the fixed car jacking machine and construct the SDG model of the fixed car jacking machine.

[0132] Under normal circumstances, there is a certain safe distance between the load-bearing nut and the bottom iron plate. As the load-bearing nut gradually wears down, the distance between the nut and the bottom iron plate will gradually decrease. In severe cases, this will cause the load-bearing nut to fall off.

[0133] ① Analyze the correlation mechanism of the fixed car jacking machine, including the operating relationship, fault phenomena and causes of components such as bearing nuts, rollers and lead screws, and try to avoid unmeasurable state parameters without affecting the positioning results, so as to facilitate the construction of the SDG model.

[0134] ② Fault Reasoning Model Construction. Based on the fixed-type vehicle jack association mechanism identified in step ①, an equipment SDG model is constructed. The equipment is divided into subsystems according to the model to facilitate the subsequent construction of a multi-subsystem monitoring system. The reasoning method of SDG is improved by combining the characteristics of early anomalies.

[0135] ③ Equipment monitoring system construction. Based on the systems defined in step ②, fault monitoring models based on the PCA algorithm are constructed, and a verification algorithm is introduced to detect changes in the health status trend of the system, thereby narrowing the fault scope to the system level and identifying anomalies in the early stages of the fault.

[0136] ④ Real-time fault monitoring and location of equipment. Based on the real-time data of system segmentation in step ③, the health status of the system is calculated. By detecting the trend of the health status, it is determined whether there is a fault in the system. Finally, the set of original faults is obtained by reasoning through the SDG model constructed in step ②.

[0137] S4: Construct a fault monitoring model based on subsystems and introduce a verification algorithm to detect the health status of the fixed vehicle jacking machine.

[0138] A wear mechanism model for load-bearing nuts was constructed, and simulation data under different working conditions was simulated based on this model. The simulation data and the collected data were used together as a dataset.

[0139] Data preprocessing: First, a certain amount of data related to nut wear needs to be collected. The data should include relevant characteristics of the nuts and bolts, such as material, geometry, load, and usage environment. Before training the model, the data needs to be cleaned and preprocessed. Data cleaning aims to remove outliers, missing values, and duplicate data. Preprocessing involves feature selection, feature scaling, and data transformation to improve the accuracy and efficiency of the algorithm.

[0140] Feature engineering: It is necessary to determine which features are most important for predicting nut thread wear. The purpose of feature engineering is to extract useful features from the raw data to improve the accuracy and interpretability of the algorithm. Considering the temporal and periodic characteristics of the signal itself, feature extraction mainly includes two aspects: time domain and frequency domain extraction. Time domain indicators include peak-to-peak value, kurtosis, etc.; frequency domain indicators include empirical mode decomposition (EMD), etc. Frequency domain and time domain indicators will have some degree of correlation, making calculation difficult. PCA preprocessing method is used to reduce the number of variables while retaining most of the information in the original data.

[0141] Model Training: The SVR algorithm was chosen to build the prediction model. The dataset was divided into training and test sets. The model was trained using the training set and its accuracy and generalization ability were verified using the test set. For the existing data, the degradation values ​​at n consecutive time points were used as a set of data, with the degradation value at time point i as the result. The degradation values ​​at the previous n-1 time points were also considered, resulting in a rough degradation curve. During training, techniques such as cross-validation and grid search could be used to optimize the model parameters.

[0142] Model tuning: Evaluate the model's accuracy and generalization ability, and fine-tune the model. Common metrics such as mean squared error (MSE), mean absolute error (MAE), and R-squared value can be used to evaluate model performance.

[0143] Model Application: A trained model is used to predict the wear of nut threads. Relevant features of the nut and bolt are input into the model to obtain the prediction results. Data collected in real-time at n consecutive time points is imported into the trained model to obtain the degradation amount at the next time point. This is compared with data from the entire lifespan to determine the remaining service life of the nut. When applying the model, attention should be paid to its confidence level and interpretability to ensure the accuracy and reliability of the prediction results.

[0144] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. A fault diagnosis method for fixed-type vehicle jacking machines in urban rail vehicle depots based on digital twins, characterized in that: The method includes: Establish a digital twin of the fixed vehicle jacking machine; Sensors are installed on various components of the fixed vehicle jacking machine to collect real-time operating data, which is then transmitted to the digital twin of the fixed vehicle jacking machine for synchronization. The interrelationship mechanism of the operation of each component of the fixed car jacking machine was analyzed, and the SDG model of the fixed car jacking machine was constructed and divided into multiple subsystems; A fault monitoring model based on the PCA algorithm is constructed based on the subsystem, and a verification algorithm is introduced to detect the health status of the fixed vehicle jacking machine. By detecting the health status, it can be determined whether a fault exists; in: The operational mechanisms of various components of a fixed car jacking machine are analyzed, and an SDG model of the fixed car jacking machine is constructed and divided into multiple subsystems, including: Based on the operational relationships, fault phenomena, and causes of various components of the fixed vehicle jacking machine, the fault propagation path is analyzed to obtain the correlation mechanism of the operation of each component. Based on the association mechanism, a fixed vehicle rack SDG model is constructed; The SDG model of the fixed vehicle jacking machine is divided into multiple subsystems to simplify the SDG model; A fault monitoring model based on the PCA algorithm is constructed based on the subsystem, and a verification algorithm is introduced to detect the health status of the fixed car-mounting machine, including: Data normalization is performed by collecting real-time data and historical data from various sensors. Principal components are obtained by calculating the covariance matrix and then compared with historical data. If it is historical data, it is stored in the threshold database. If it is not historical data, the principal component is squared, and the result is compared with the threshold. If it is greater than or equal to the threshold, the principal component and the result are input into the fault reasoning model for fault location. If it is less than the threshold, the principal component and the result are compared for mutation detection. If a mutation point exists, the principal component and the result are input into the fault reasoning model for fault location. If no mutation point exists, no fault is displayed. After detecting the health status and determining whether there is a fault, the set of root faults is obtained using the SDG model of the fixed vehicle jacking machine. The method further includes: After establishing and synchronizing a digital twin of the fixed jacking machine, the digital twin is trained using historical operational data, including: The system acquires operational data of various components of the fixed vehicle jack over a period of time, collected by sensors, to obtain historical operational data. Perform feature calculations and filtering on historical operational data; The entire lifecycle of the fixed vehicle rack is divided into different stages, and then the key degradation stages are characterized by feature calculation and screening, and data augmentation is performed. The selected features are fused based on the combined BIM model, and the LSTM model with bias enhancement is used to learn the HS-RUL mapping relationship, thereby completing the training of the digital twin of the fixed vehicle jack.

2. The fault diagnosis method for fixed-type vehicle jacking machines in urban rail vehicle depots based on digital twins according to claim 1, characterized in that: Establish a digital twin of the fixed vehicle jacking machine, including: Establish BIM models of the maintenance depot's civil engineering, drainage, and lighting, and overlay and merge them based on a unified reference base point to obtain the maintenance depot's BIM model; Establish BIM models of each component of the fixed vehicle jacking machine, and overlay and merge them based on a unified reference base point to obtain the BIM model of the fixed vehicle jacking machine; Based on a unified reference point, the BIM model of the fixed vehicle rack machine is overlaid on the BIM model of the maintenance depot to obtain a combined BIM model; Lightweighting of the composite BIM model; In the composite BIM model, motion behavior rules that are mapped to each other are established for the related components in the fixed vehicle jacking machine. Obtain a digital twin of the fixed vehicle jacking machine.

3. The fault diagnosis method for fixed-type vehicle jacking machines in urban rail vehicle depots based on digital twins according to claim 2, characterized in that: Sensors are installed on various components of the fixed jacking machine to collect real-time operating data, which is then transmitted to the fixed jacking machine's digital twin for synchronization. This includes: Sensors are installed on various components of the fixed vehicle frame machine; The collected real-time operation data is transmitted to the digital twin of the fixed vehicle jack. The physical fixed vehicle jacking machine communicates with its digital twin. The digital twin updates its status based on real-time operating data and displays it in three dimensions. The digital twin of the fixed vehicle jacking machine simulates and optimizes the operating status of the fixed vehicle jacking machine based on real-time operating data, and performs real-time control of the fixed vehicle jacking machine.

4. The fault diagnosis method for fixed-type vehicle jacking machines in urban rail vehicle depots based on digital twins according to claim 3, characterized in that: Fault types are divided into sudden faults and cumulative faults; A subsystem of the fixed scaffolding machine SDG model experienced a sudden and severe performance degradation due to stagnation, indicating a sudden and severe abnormal state. The performance of a certain subsystem of the fixed scaffolding machine SDG model gradually declined due to aging. In the early stage, the abnormal state was vague and slight, and the overall operation was stable. The fault symptoms only became obvious in the middle and late stages, which is a cumulative fault.

Images