Heavy fork electrical system fault diagnosis and intelligent repair method

Abstract

The invention discloses a heavy-fork electrical system fault diagnosis and intelligent repair method, and particularly relates to the technical field of electrical system intelligent operation and maintaining.The method comprises the steps that multi-source data of electrical parameters, mechanical parameters and environmental parameters are collected through a sensor, and fault spatio-temporal feature extraction is conducted through a spatio-temporal attention feature extraction network; then obtaining a fault intelligent repair strategy based on the similarity between the real-time fault spatio-temporal feature vector and the historical fault spatio-temporal feature vector, issuing a corresponding intelligent repair instruction through a CAN bus, monitoring an execution result, evaluating execution result data after repair operation, and feeding back an evaluation abnormal result to a management terminal for human-computer interaction; according to the method, feature extraction is carried out by means of the network comprising the space attention sub-network and the cavity convolution time sequence feature extractor, the feature information of the system operation state can be comprehensively and accurately captured, and rich feature basis is provided for subsequent fault diagnosis and intelligent repair.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for electrical systems, specifically a method for fault diagnosis and intelligent repair of a heavy forklift electrical system. Background Technology

[0002] In logistics transportation and warehousing operations, heavy-duty forklifts are core equipment in the industrial logistics field, and the reliability of their electrical systems directly determines production efficiency; however, with the increase in equipment complexity and the improvement of intelligence, fault diagnosis and repair face many challenges.

[0003] Traditional fault diagnosis mainly relies on the experience and intuitive judgment of maintenance personnel. They make preliminary judgments about the location and cause of faults by observing the operating status of forklifts, listening to abnormal sounds, and touching the temperature of components. Usually, only a few key physical quantities of electrical systems are tested, such as voltage and current, which is difficult to fully reflect the operating status of the system. Traditional repair strategies are often based on pre-set rules and experience, lacking specificity and flexibility. After the repair is completed, there is a lack of effective evaluation and feedback mechanisms for the repair effect.

[0004] Currently, most existing research in the field of fault diagnosis and repair focuses on a single aspect of fault diagnosis or repair, lacking a systematic approach that organically combines fault diagnosis, repair strategy generation, automatic repair execution, and repair effect evaluation and iteration. This makes it difficult to achieve intelligent processing of faults in heavy forklift electrical systems throughout the entire process. Moreover, heavy forklift electrical systems involve various types of sensor data, such as voltage, current, temperature, and vibration. These data have different characteristics and dimensions. How to effectively integrate these multi-source data and extract more valuable fault information is a problem that current technologies have not yet solved well. Therefore, developing an efficient, accurate fault diagnosis method for heavy forklift electrical systems with intelligent repair capabilities is of significant practical importance. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method for fault diagnosis and intelligent repair of a heavy forklift electrical system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system, comprising: S1: Collect and store historical operation and maintenance data of heavy forklift electrical system, including operation data under standard conditions, normal operation data, fault operation data and fault repair strategies. At the same time, receive and update the data generated during the operation of the heavy forklift electrical system in real time, and build a heavy forklift electrical system management database. S2: By deploying sensors at the monitoring locations of the heavy fork electrical system, multi-source data of electrical, mechanical and environmental parameters are collected in real time and synchronously, and then fused to construct a three-dimensional data matrix for fault diagnosis; S3: Using the spatiotemporal attention feature extraction network that includes a spatial attention subnet and a dilated convolutional temporal feature extractor, feature extraction is performed on the preprocessed 3D data matrix acquired in real time to obtain the spatiotemporal feature vector of the fault. S4: Construct a fault diagnosis and intelligent repair knowledge graph by using historical fault spatiotemporal feature vectors and corresponding intelligent repair strategies, and obtain intelligent fault repair strategies based on the similarity between real-time fault spatiotemporal feature vectors and historical fault spatiotemporal feature vectors. S5: Based on the acquired intelligent fault repair strategy, send out corresponding intelligent repair commands through the CAN bus, and monitor the execution results at the same time; S6: Evaluate the execution result data after the repair operation, and feed back any abnormal evaluation results to the management terminal for human-computer interaction.

[0007] The technical effects and advantages of this invention are as follows: 1. This invention uses a multi-physical quantity synchronous acquisition module to deploy sensors at the monitoring position of the heavy fork electrical system, and collects multi-source data such as electrical parameters, mechanical parameters and environmental parameters in real time. This can more comprehensively reflect the operating status of the system, effectively avoid misjudgment or missed diagnosis of faults due to abnormal single parameter, and significantly improve the accuracy and reliability of fault diagnosis. 2. This invention utilizes a network that includes a spatial attention subnet and a dilated convolutional temporal feature extractor for feature extraction. This enables comprehensive and accurate capture of the feature information of the system's operating status. It can adaptively adjust the degree of attention to different spatial locations and temporal features, adapt to various complex operating conditions, and provide rich feature basis for subsequent identification of fault status labels and intelligent repair. 3. This invention utilizes historical fault spatiotemporal feature vectors and corresponding intelligent repair strategies to construct a knowledge graph. Based on the similarity between real-time fault spatiotemporal feature vectors and historical fault spatiotemporal feature vectors, it quickly obtains intelligent fault repair strategies, reducing the time for manual analysis and decision-making. Corresponding intelligent repair commands are issued through the CAN bus, and the execution results are monitored. The system is continuously optimized based on the monitoring results to improve its adaptive capabilities. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0009] Figure 2 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0011] Please see Figure 1 As shown, the present invention provides an AI-based production control system for offshore wind turbine nacelle structural components, including a heavy-duty electrical system management database, a multi-physical quantity synchronous acquisition module, a spatiotemporal attention feature extraction module, a fault-repair dynamic mapping module, an intelligent repair strategy execution module, and a repair effect evaluation and feedback module.

[0012] The heavy forklift electrical system management database is connected to the other modules, the multi-physical quantity synchronous acquisition module is connected to the spatiotemporal attention feature extraction module, the fault-repair dynamic mapping module is connected to the spatiotemporal attention feature extraction module and the intelligent repair strategy execution module respectively, and the repair effect evaluation and feedback module is connected to the fault-repair dynamic mapping module and the intelligent repair strategy execution module respectively.

[0013] Heavy-duty forklift electrical system management database: used to collect and store historical heavy-duty forklift electrical system operation and maintenance data, including operating data under standard conditions, normal operating data, fault operating data and fault repair strategies, and to receive and update data generated during the operation of the heavy-duty forklift electrical system in real time, thus building the heavy-duty forklift electrical system management database; Multi-physical quantity synchronous acquisition module: By using sensors placed at the monitoring position of the heavy fork electrical system, multi-source data of electrical parameters, mechanical parameters and environmental parameters are collected in real time and synchronously, and fused to construct a three-dimensional data matrix for fault diagnosis, and then transmitted to the spatiotemporal attention feature extraction module; Spatiotemporal attention feature extraction module: Through the constructed spatiotemporal attention feature extraction network containing a spatial attention subnet and a dilated convolutional temporal feature extractor, features are extracted from the preprocessed 3D data matrix acquired in real time to obtain real-time fault spatiotemporal feature vectors, which are then transmitted to the fault-repair dynamic mapping module. Fault-Repair Dynamic Mapping Module: Constructs a fault diagnosis and intelligent repair knowledge graph through historical fault spatiotemporal feature vectors and corresponding intelligent repair strategies. Obtains intelligent fault repair strategies based on the similarity between real-time fault spatiotemporal feature vectors and historical fault spatiotemporal feature vectors, and transmits them to the intelligent repair strategy execution module. Intelligent Repair Strategy Execution Module: Based on the acquired intelligent fault repair strategy, it sends corresponding intelligent repair commands through the CAN bus, monitors the execution results, and transmits them to the repair effect evaluation and feedback module. Repair effect evaluation and feedback module: evaluates the execution result data after the repair operation, and feeds back abnormal evaluation results to the management terminal for human-computer interaction.

[0014] Please see Figure 2 As shown, a method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system includes: S1: collecting and storing historical operation and maintenance data of the heavy-duty forklift electrical system, including operating data under standard conditions, normal operating data, fault operating data, and fault repair strategies, while simultaneously receiving and updating data generated during the operation of the heavy-duty forklift electrical system in real time, and constructing a heavy-duty forklift electrical system management database; S2: using sensors deployed at the monitoring locations of the heavy-duty forklift electrical system to synchronously collect multi-source data of electrical parameters, mechanical parameters, and environmental parameters in real time, and fusing them to construct a three-dimensional data matrix for fault diagnosis; S3: using the constructed temporal features including spatial attention subnets and dilated convolutions... The spatiotemporal attention feature extraction network of the extractor extracts features from the preprocessed 3D data matrix acquired in real time to obtain the spatiotemporal feature vector of the fault; S4: Construct a fault diagnosis and intelligent repair knowledge graph through historical fault spatiotemporal feature vectors and corresponding intelligent repair strategies, and obtain the fault intelligent repair strategy based on the similarity between the real-time fault spatiotemporal feature vector and the historical fault spatiotemporal feature vector; S5: Based on the obtained fault intelligent repair strategy, issue corresponding intelligent repair commands through the CAN bus, and monitor the execution results at the same time; S6: Evaluate the execution result data after the repair operation, and feed back the evaluation abnormal results to the management terminal for human-computer interaction.

[0015] S1: Collect and store historical operation and maintenance data of heavy forklift electrical system, including operation data under standard conditions, normal operation data, fault operation data and fault repair strategies. At the same time, receive and update the data generated during the operation of the heavy forklift electrical system in real time, and build a heavy forklift electrical system management database. S2: By deploying sensors at the monitoring locations of the heavy forklift electrical system, multi-source data of electrical, mechanical, and environmental parameters are collected in real time and synchronously, and then fused to construct a three-dimensional data matrix for fault diagnosis, including the following steps: S2.1: Using sensor technology, electrical parameters are acquired at a sampling frequency f1 (e.g., 1kHz), mechanical parameters are acquired at a sampling frequency f2 (e.g., 100Hz), and environmental parameters are acquired at a sampling frequency f2 (e.g., 10Hz); electrical parameters include three-phase voltage U. a,b,c (t), Three-phase current I a,b,c (t) and three-phase current harmonic distortion rate (THD) a,b,c (t), mechanical parameters include vibration acceleration a in the x-direction. x (t) and y-direction vibration acceleration a y The vibration acceleration a in the (t) and z directions z(t), where environmental parameters include temperature T(t), humidity H(t), and dust concentration D(t); This embodiment requires specific explanation of the monitoring locations of the heavy forklift electrical system, including the motor end, controller end, hydraulic pump end, and cab; the sensors include, but are not limited to, voltage sensors, current sensors, vibration sensors, temperature and humidity composite sensors, and dust concentration sensors.

[0016] This embodiment specifically illustrates the use of a current sensor to collect the current in the electric system of the heavy forklift in real time, obtaining a discrete sequence I(n) of the current changing over time. Let n be the number of samples at time point t. The collected discrete current sequence is processed using FFT (Fast Fourier Transform) to calculate the current harmonic distortion rate THD(t) (using an existing formula). This process is repeated for all three phase currents to obtain the three-phase current harmonic distortion rate THD. a,b,c (t); To accurately describe the mechanical vibration of a heavy-duty forklift system, three mutually perpendicular directions are established. For example, during the operation of a heavy-duty forklift in a heavy-duty forklift system, its body may vibrate simultaneously in three directions: front-to-back (assuming the x-direction), left-to-right (assuming the y-direction), and vertical (assuming the z-direction). x This reflects the intensity of the object's vibration along the x-axis, a y This reflects the intensity of the object's vibration along the y-axis, a z It reflects the intensity of the object's vibration along the z-axis.

[0017] S2.2: Use the PTP precision clock protocol (IEEE 1588) to synchronize the timestamps of each sensor, with the error less than or equal to a threshold Δt (e.g., 50 ns). Calculate the synchronization time t of each sensor. sync , t sync =t mas +Δt pro -Δt deL , t mas Master clock reference time, Δt pro For propagation delay, Δt deL This is the clock offset compensation value; S2.3: The synchronized electrical, mechanical, and environmental parameters are fused to construct a three-dimensional data matrix M(t)∈R. 3×3×n n is the number of samples at time point t. ; This implementation requires specific explanation regarding the three-phase voltage U in the electrical parameters. a,b,c (t) as a whole, the three-phase current I a,b,c (t) as a whole and the three-phase current harmonic distortion rate (THD) a,b,c (t) as a whole.

[0018] S3: Using the constructed spatiotemporal attention feature extraction network, which includes a spatial attention subnetwork and a dilated convolutional temporal feature extractor, features are extracted from the preprocessed 3D data matrix acquired in real time to obtain the spatiotemporal feature vector of the fault. This includes the following steps: S3.1: Constructing a spatiotemporal attention feature extraction network includes: S3.1.1: First, obtain N sets (e.g., 5000 sets) of historical three-dimensional data matrix M(t) of different fault states of the heavy forklift electrical system from the heavy forklift electrical system management database. Perform preprocessing operations on the three-dimensional data matrix, including noise reduction, outlier handling, and data normalization, to obtain the preprocessed three-dimensional data matrix M. y (t)∈R 3×3×n Then, N sets of preprocessed three-dimensional data matrices M y (t)∈R 3×3×n And the corresponding one-hot encoded vectors of fault states are used as the training dataset D. [M] N [t), yN] represents the Nth group of three-dimensional data matrix M. N The fault status label of (t) is yN. The fault status label is a one-hot encoded vector, which consists of three fault types: no fault, mechanical fault, and environmental fault. The severity levels are mild, moderate, and severe. There are a total of 3×3+1=10 fault status label categories. For example, the one-hot encoded vector of no fault is [1,0,0,0,0,0,0,0,0,0], and the one-hot encoded vector of mild electrical fault is [0,1,0,0,0,0,0], where 1 represents electrical fault and 0.5 represents the corresponding fault severity coefficient. In this implementation, it should be noted that the preprocessing operation is an existing technology, noise reduction can be achieved using filtering algorithms such as median filtering and mean filtering, outlier handling can be achieved using statistical methods such as the 3σ principle, and normalization can map the data to the [0,1] interval. Commonly used methods include min-max normalization.

[0019] This embodiment specifically explains that the fault severity is obtained by summing the absolute values ​​of the differences between all fault elements and their corresponding standard condition elements in the three-dimensional data matrix for each fault type, and the ratio of these differences to the standard condition elements. Then, the fault severity value of each fault element is compared with preset severity value ranges for mild, moderate, and severe faults to obtain the severity of the fault category corresponding to the fault element. For example, mechanical fault severity = [(|a x -a x,0 | / a x,0 )+(|a y -a y,0 | / a y,0)+(|a z -a z,0 | / a z,0 )],a x,0 a y,0 and a z,0 These represent the accelerations in the x, y, and z directions, respectively, under standard conditions.

[0020] S3.1.2: First, the spatial attention weights Ms(M) in the CBAM attention module structure are... y (t) Add adaptive gating function ΘM y (t), calculate the spatial attention weights Ms after adding the adaptive gating function. t (M y (t)), Ms t (M y (t))=Ms(M y (t))×ΘM y (t), Ms(M) y (t))=σ(Conv 7×7 ([AvgPool(M y (t));MaxPool(M y (t))])),ΘM y (t)=σ(k×M y (t)+b), k is a learnable weight parameter, b is a learnable bias parameter, Ms(M y (t)) through the input three-dimensional data matrix M y (t) First perform average pooling AvgPool(M) y (t) and MaxPool(M) y The two pooling results are then concatenated along the channel dimension and followed by a 7×7 convolution Conv. 7×7 The spatial attention weights Ms are obtained by using the sigmoid activation function σ; then... t (M y (t) and the input three-dimensional data matrix M y (t) Element-wise multiplication yields the output spatial features Fs of the spatial attention subnet, Fs = Ms t (M y (t))⊙M y (t); This embodiment specifically explains that CBAM (Convolutional Block Attention Module) is a convolutional block attention module composed of two sub-modules: channel attention and spatial attention. The design goal of CBAM is to improve the feature representation capability of convolutional neural networks by explicitly modeling the attention in both channel and spatial dimensions. In this application, it improves the spatiotemporal feature representation capability of convolutional neural networks for faults.

[0021] S3.1.3: First, the output Fs of the spatial attention subnet is used as the input of the dilated convolution temporal feature extractor. The dilated convolution structure consists of three stacked dilated convolution layers with dilation rates of 2, 4, and 8, respectively. K represents the dilation rate, and the receptive field of the dilated convolution is calculated using the formula RF. RF pre The size of the receptive field of the previous layer is k, the initial receptive field size is the kernel size, k is the kernel size (e.g., k=3), stride is the stride (e.g., stride=1), and n is the number of dilated convolution layers. i Let Fs be the dilation rate of the i-th layer; then, the input Fs is processed by dilated convolution to extract temporal features, resulting in the output temporal features Ft of the dilated convolution temporal feature extractor, where Ft = Conv d-8 (Conv d-4 (Conv d-2 (Fs))), Conv d-K For a convolution operation with a dilation rate of K; S3.1.4: First, the spatial features Fs obtained from the spatial attention subnet and the temporal features Ft obtained from the dilated convolution temporal extractor are fused to obtain the fault spatiotemporal feature vector F, which has a dimension of 512. F = Flatten([Fs;Ft]), where Flatten() is the flattening operation. Then, the fault spatiotemporal feature vector F is input into the fully connected layer and the Softmax function to obtain the fault state label prediction probability y. y y y =Softmax(W F T F+b F ), W F W is an m1×n1 dimensional vector, where n1 is the dimension of the fault spatiotemporal feature vector F, m1 is the number of classification categories, i.e., 10 fault state label categories. F T For W F transpose, b F The bias vector is a learnable vector with m1 dimensions; then, the prediction result y is constructed using the cross-entropy loss function. yThe loss function L(θ) between the true label y and the actual label y is... y c Let y be the c-th element in the one-hot encoding of label y. c y To predict the probability y y The c-th element in the algorithm represents the learnable parameters of the network. Then, a backpropagation algorithm (e.g., stochastic gradient descent, Adam optimization algorithm) is used to calculate the gradient ∇ of the network parameters. θ L adjusts the value of θ and finds the minimum value of the loss function L(θ). When the value of the loss function changes less than the threshold for d1 consecutive times (e.g., 5 times), it indicates that the training is over, d1 < d; finally, the spatiotemporal attention feature extraction network is output. In this embodiment, it should be specifically noted that in the fault diagnosis scenario of heavy fork electrical system, the larger the sensing field, the more long-distance dependencies can be captured in the time series data, which helps to discover features such as the development of intermittent faults over time and the accumulation process of gradual faults.

[0022] S3.2: Using the constructed spatiotemporal attention feature extraction network, features are extracted from the preprocessed 3D data matrix acquired in real time to obtain the real-time fault spatiotemporal feature vector F. now Then, the real-time fault spatiotemporal feature vector F now The input is fed into a fully connected layer and a softmax function to obtain the fault state label; if the real-time fault spatiotemporal feature vector F now If the corresponding fault status label is "no fault", no further action is required; otherwise, further action is required. S4: Construct a fault diagnosis and intelligent repair knowledge graph using historical fault spatiotemporal feature vectors and corresponding intelligent repair strategies. Obtain intelligent fault repair strategies based on the similarity between real-time fault spatiotemporal feature vectors and historical fault spatiotemporal feature vectors, including the following steps: S4.1: First, define the set of fault feature nodes V using historical fault spatiotemporal feature vectors and corresponding intelligent repair strategies. F and the set of intelligent repair strategy nodes V R , F N Let N be the spatiotemporal feature vector of the Nth group of historical faults, where N is the number of historical fault cases. Then, the intelligent repair strategy text is hierarchically encoded using the BERT model to obtain the text feature vector R, R=BERT([w1,w2,...wm1]), where wm1 represents the m1th word unit after word segmentation and part-of-speech tagging. By traversing N groups of intelligent repair strategy texts, the set of intelligent repair strategy nodes V is obtained. R , R N For F NThe corresponding intelligent repair strategy text vector; finally, for each group of historical fault cases (F) N ,R N Establish from F N To R N The directed edge E N Construct a preliminary knowledge graph for fault diagnosis and intelligent repair (READ1); This implementation requires a specific explanation: BERT (Bidirectional Encoder Representations from Transformers) is a pre-trained language model based on the Transformer architecture. Through unsupervised learning on a large-scale intelligent repair strategy text corpus, it learns rich language representations and semantic information. By using a bidirectional Transformer encoder to capture contextual information in the text, it can better understand the meaning of words in different contexts.

[0023] S4.2: For each directed edge E of the initial fault diagnosis and intelligent repair knowledge graph READ1 N weight W N The edge weight contribution Δw of the remaining N-1 groups of historical fault cases is calculated. ij Accumulated, , Δw ij Let be the edge weight contribution between node i and node j, i ≠ j, i ∈ N, j ∈ N, and α be the balance coefficient (e.g., α = 0.75). Sim(F i ,F j F represents the spatiotemporal feature vectors of the i-th and j-th historical faults. i and F j cosine similarity, ||R i -R j || represents the distance norm between the corresponding i-th and j-th smart repair strategy text vectors (measuring the distance between vectors; the smaller the distance, the closer the exponent is to 1), SF i th1 Sim(F) i ,F j The corresponding threshold (e.g., 0.8); then iterate through all directed edges E. N weight W N Construct a complete knowledge graph for fault diagnosis and intelligent repair (READ2); S4.3: Based on the similarity function Sim, traverse C N 2 Group (C) N 2To obtain the similarity SF between any two sets of historical fault spatiotemporal feature vectors (represented by the number of combinations, where 2 sets are selected from N sets of historical fault cases), we need to calculate the similarity SF between each pair of historical fault spatiotemporal feature vectors. i If SF i ≥threshold SF i th1 If they are not identical, they are merged into a single node; otherwise, they exist as independent nodes. Traverse C. N 2 The similarity of spatiotemporal feature vectors of historical faults is used to evaluate the set of fault feature nodes V. F Update to V F g Traverse C based on a distance function (e.g., Euclidean distance). N 2 Group intelligent repair strategy nodes, obtain the distance function value SR between the text vectors of every 2 groups of intelligent repair strategies. i If SR i <threshold SR i th (For example, 0.3), then they are merged into one node; otherwise, they exist as independent nodes. Traverse C. N 2 The distance function value of the text vector of the group intelligent repair strategy, for the repair strategy node set V R Update to V R g Finally, the updated fault diagnosis and intelligent repair knowledge graph READ3 is obtained; S4.4: Calculate the real-time fault spatiotemporal feature vector F now The set of fault nodes V in the Fault Diagnosis and Intelligent Repair Knowledge Graph READ3 F g Each node F in N cosine similarity Sim(F) now ,F N ), filter out those with similarity ≥ threshold SF i th1 The faulty nodes are used as the candidate fault feature node set VF now Then, for candidate fault feature nodes F... N ∈VF now To the intelligent repair strategy node R N Given the path, obtain the set of path edge weights W. N (VF now The optimal repair strategy R is selected by filtering the repair strategy text corresponding to the path with the highest path edge weight. opt This refers to intelligent fault repair strategies (e.g., motor winding temperature > 150℃ AND abnormal vibration acceleration THEN generation strategy: shutdown inspection + winding insulation test + stator replacement if necessary); if the cosine similarity Sim(Fnow ,F N ) < threshold SF i th1 This triggers the management terminal decision interface; S4.5: If the real-time fault spatiotemporal feature vector F now With the set of fault characteristic nodes V F g node F in N cosine similarity Sim(F) now ,F N )≥threshold SF i th2 (e.g., 0.95), threshold SF i th2 > Threshold SF i th1 Update the fault spatiotemporal feature vector of the node; otherwise, add a fault feature node and a corresponding intelligent repair strategy node. If the freshness fh of historical fault cases is less than the corresponding threshold fh... th (For example, if fhth=0.2), then the faulty case is eliminated; otherwise, it is retained, fh=f(t) case )×g(up),f(t) case The basic time decay function is... , t case T represents the number of days since a historical failure case. max This represents the maximum number of days since a historical failure (e.g., 730 days), where g(up) represents the upgrade and correction items for the heavy forklift system equipment. up=1 represents historical fault cases after the equipment upgrade, up=0 represents historical fault cases before the equipment upgrade, and β is a correction factor (e.g., β=0.5); for example, case A: t case =365 days, up=1 (upgraded case), fh=(1-365 / 730)×1=0.5, 0.5≥0.2, case retained; S5: Based on the acquired intelligent fault repair strategy, send the corresponding intelligent repair command through the CAN bus, and monitor the execution result. If the execution fails, trigger S4 to reacquire the intelligent fault repair strategy. The intelligent repair command is sent through the CAN bus message, which includes the fault status label category and the intelligent repair strategy text. In this embodiment, it should be specifically explained that, for example, if the fault is an electrical fault and the power module control mechanism fails to successfully adjust the output voltage waveform, the monitoring system will send a command or switch to a backup control strategy, such as activating a backup filter circuit.

[0024] S6: Evaluate the execution result data after the repair operation, and feed back any abnormal evaluation results to the management terminal for human-computer interaction, including the following steps: S6.1: Monitor the execution results of the repair strategy and the real-time status data of the system via the CAN bus. The data on the execution results of the repair strategy includes the number of successful fault repairs (sc), the total number of repairs (ta), and the time (t) for each repair. I The spatiotemporal feature vector F of the I-th fault after repair I h And the spatiotemporal feature vector F of the I-th fault under standard conditions I b Calculate the performance evaluation index η. , t I,h Let ||·||2 be the average repair time of the three most recent repair operations that had the same repair operation as the I-th repair operation, and let ||·||2 be the L2 norm, where I∈ta; S6.2: The performance evaluation metric η is less than the corresponding threshold η th This triggers incremental learning of the fault diagnosis and intelligent repair knowledge graph READ3. new =READ3∪[(F now ,F h )], F now and F h These are the real-time fault spatiotemporal feature vector and the repaired fault spatiotemporal feature vector, respectively, and are fed back to the management terminal. Otherwise, it indicates that the execution result is good.

[0025] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be considered as such. It is included within the scope of protection of this invention.

Claims

1. A method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system, characterized in that: include: S1: Collect and store historical operation and maintenance data of heavy forklift electrical system, including operation data under standard conditions, normal operation data, fault operation data and fault repair strategies. At the same time, receive and update the data generated during the operation of the heavy forklift electrical system in real time, and build a heavy forklift electrical system management database. S2: By deploying sensors at the monitoring locations of the heavy fork electrical system, multi-source data of electrical, mechanical and environmental parameters are collected in real time and synchronously, and then fused to construct a three-dimensional data matrix for fault diagnosis; S3: Using the spatiotemporal attention feature extraction network that includes a spatial attention subnet and a dilated convolutional temporal feature extractor, feature extraction is performed on the preprocessed 3D data matrix acquired in real time to obtain the spatiotemporal feature vector of the fault. S4: Construct a fault diagnosis and intelligent repair knowledge graph by using historical fault spatiotemporal feature vectors and corresponding intelligent repair strategies, and obtain intelligent fault repair strategies based on the similarity between real-time fault spatiotemporal feature vectors and historical fault spatiotemporal feature vectors. S5: Based on the acquired intelligent fault repair strategy, send out corresponding intelligent repair commands through the CAN bus, and monitor the execution results at the same time; S6: Evaluate the execution result data after the repair operation, and feed back any abnormal evaluation results to the management terminal for human-computer interaction.

2. The method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system according to claim 1, characterized in that: The construction of the spatiotemporal attention feature extraction network in S3 includes: S3.1.1: First, obtain N sets of historical three-dimensional data matrices M(t) of different fault states of the forklift electrical system from the forklift electrical system management database, and perform preprocessing operations on the three-dimensional data matrices to obtain the preprocessed three-dimensional data matrix M. y (t)∈R 3×3×n Then, N sets of preprocessed three-dimensional data matrices M y (t)∈R 3×3×n And the corresponding one-hot encoded vectors of fault states are used as the training dataset D. [M] N [t), yN] represents the Nth group of three-dimensional data matrix M. N The fault status label of (t) is yN. The fault status label is a one-hot encoded vector, consisting of three fault types: no fault, no fault, and three corresponding severity levels. The three fault types include electrical fault, mechanical fault, and environmental fault. The severity levels include mild, moderate, and severe. There are a total of 3×3+1=10 fault status label categories.

3. The method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system according to claim 1, characterized in that: The construction of the spatiotemporal attention feature extraction network in S3 also includes: S3.1.2: First, the spatial attention weights Ms(M) in the CBAM attention module structure are... y (t) Add adaptive gating function ΘM y (t), calculate the spatial attention weights Ms after adding the adaptive gating function. t (M y (t)), Ms t (M y (t))=Ms(M y (t))×ΘM y (t), Ms(M) y (t))=σ(Conv 7×7 ([AvgPool(M y (t));MaxPool(M y (t))])),ΘM y (t)=σ(k×M y (t)+b), k is a learnable weight parameter, b is a learnable bias parameter, Ms(M y (t)) through the input three-dimensional data matrix M y (t) First perform average pooling AvgPool(M) y (t) and MaxPool(M) y The two pooling results are then concatenated along the channel dimension and followed by a 7×7 convolution Conv. 7×7 The spatial attention weights Ms are obtained by using the sigmoid activation function σ; then... t (M y (t) and the input three-dimensional data matrix M y (t) Element-wise multiplication yields the output spatial features Fs of the spatial attention subnet, Fs = Ms t (M y (t))⊙M y (t); S3.1.3: First, the output Fs of the spatial attention subnet is used as the input of the dilated convolution temporal feature extractor. The dilated convolution structure consists of three stacked dilated convolution layers with dilation rates of 2, 4, and 8, respectively. K represents the dilation rate, and the receptive field of the dilated convolution is calculated using the formula RF. RF pre is the receptive field size of the previous layer, k is the kernel size, stride is the stride, n is the number of dilated convolution layers, and dilation... i Let Fs be the dilation rate of the i-th layer; then, the input Fs is processed by dilated convolution to extract temporal features, resulting in the output temporal features Ft of the dilated convolution temporal feature extractor, where Ft = Conv d-8 (Conv d-4 (Conv d-2 (Fs))), Conv d-K This is a convolution operation with a dilation rate of K.

4. The method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system according to claim 1, characterized in that: In step S3, the spatiotemporal feature vector of the fault is obtained by extracting features from the pre-processed 3D data matrix acquired in real time using the constructed spatiotemporal attention feature extraction network to obtain the real-time spatiotemporal feature vector F. now ; Then the real-time fault spatiotemporal feature vector F now The input is fed into a fully connected layer and a softmax function to obtain the fault state label; if the real-time fault spatiotemporal feature vector F now If the corresponding fault status label is "no fault," no further action is required; otherwise, further action is required.

5. The method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system according to claim 1, characterized in that: The fault diagnosis and intelligent repair knowledge graph constructed in S4 includes: S4.1: First, define the set of fault feature nodes V using historical fault spatiotemporal feature vectors and corresponding intelligent repair strategies. F and the set of intelligent repair strategy nodes V R , F N Let N be the spatiotemporal feature vector of the Nth group of historical faults, where N is the number of historical fault cases. Then, the intelligent repair strategy text is hierarchically encoded using the BERT model to obtain the text feature vector R, R=BERT([w1,w2,...wm1]), where wm1 represents the m1th word unit after word segmentation and part-of-speech tagging. By traversing N groups of intelligent repair strategy texts, the set of intelligent repair strategy nodes V is obtained. R , R N For F N The corresponding intelligent repair strategy text vector; finally, for each group of historical fault cases (F) N ,R N Establish from F N To R N The directed edge E N READ1: Construct a preliminary knowledge graph for fault diagnosis and intelligent repair.

6. The method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system according to claim 1, characterized in that: The construction of the fault diagnosis and intelligent repair knowledge graph in S4 also includes: S4.2: For each directed edge E of the initial fault diagnosis and intelligent repair knowledge graph READ1 N weight W N The edge weight contribution Δw of the remaining N-1 groups of historical fault cases is calculated. ij Accumulated, Then iterate through all directed edges E. N weight W N Construct a complete knowledge graph for fault diagnosis and intelligent repair (READ2); S4.3: Based on the similarity function Sim, traverse C N 2 Collect the spatiotemporal feature vectors of historical faults, and obtain the similarity SF between every two sets of spatiotemporal feature vectors of historical faults. i If SF i ≥threshold SF i th1 If they are not identical, they are merged into a single node; otherwise, they exist as independent nodes. Traverse C. N 2 The similarity of spatiotemporal feature vectors of historical faults is used to evaluate the set of fault feature nodes V. F Update to V F g Traversing C based on distance function N 2 Group intelligent repair strategy nodes, obtain the distance function value SR between the text vectors of every 2 groups of intelligent repair strategies. i If SR i <threshold SR i th If they are not identical, they are merged into a single node; otherwise, they exist as independent nodes. Traverse C. N 2 The distance function value of the text vector of the group intelligent repair strategy, for the repair strategy node set V R Update to V R g Finally, the updated fault diagnosis and intelligent repair knowledge graph READ3 is obtained.

7. The method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system according to claim 1, characterized in that: The construction of the fault diagnosis and intelligent repair knowledge graph in S4 also includes: S4.4: Calculating the real-time fault spatiotemporal feature vector F now The set of fault nodes V in the Fault Diagnosis and Intelligent Repair Knowledge Graph READ3 F g Each node F in N cosine similarity Sim(F) now ,F N ), filter out those with similarity ≥ threshold SF i th1 The faulty nodes are used as the candidate fault feature node set VF now Then, for candidate fault feature nodes F... N ∈VF now To the intelligent repair strategy node R N Given the path, obtain the set of path edge weights W. N (VF now The optimal repair strategy R is selected by filtering the repair strategy text corresponding to the path with the highest path edge weight. opt That is, the intelligent fault repair strategy; if the cosine similarity Sim(F now ,F N ) < threshold SF i th1 This triggers the management terminal decision interface; S4.5: If the real-time fault spatiotemporal feature vector F now With the set of fault characteristic nodes V F g node F in N cosine similarity Sim(F) now ,F N )≥threshold SF i th2 Threshold SF i th2 > Threshold SF i th1 Update the fault spatiotemporal feature vector of the node; otherwise, add a fault feature node and a corresponding intelligent repair strategy node. If the freshness fh of historical fault cases is less than the corresponding threshold fh... th If the faulty case is positive, it will be eliminated; otherwise, it will be retained. fh = f(t) case )×g(up),f(t) case ) is the base time decay function, and g(up) is the upgrade correction term for the heavy forklift system equipment.

8. The method for fault diagnosis and intelligent repair of a heavy-duty forklift electrical system according to claim 1, characterized in that: The S6 implementation includes the following steps: S6.1: Monitor the execution results of the repair strategy and the real-time status data of the system via the CAN bus. The data on the execution results of the repair strategy includes the number of successful fault repairs (sc), the total number of repairs (ta), and the time (t) for each repair. I The spatiotemporal feature vector F of the I-th fault after repair I h And the spatiotemporal feature vector F of the I-th fault under standard conditions I b Calculate the performance evaluation index η; S6.2: The performance evaluation metric η is less than the corresponding threshold η th This triggers incremental learning of the fault diagnosis and intelligent repair knowledge graph READ3. new =READ3∪[(F now ,F h )], F now and F h These are the real-time fault spatiotemporal feature vector and the repaired fault spatiotemporal feature vector, respectively, and are fed back to the management terminal. Otherwise, it indicates that the execution result is good.

Images