Electric slewing mechanism safety control method and system based on fault prediction

Abstract

The invention discloses an electric slewing mechanism safety control method and system based on fault prediction, and the method comprises the steps: obtaining time series data including the motor winding temperature of an electric slewing mechanism, the U, V and W three-phase current of a motor, the rotating speed of the motor, the torque of the motor, and the X, Y and Z three-direction vibration acceleration of a transmission shaft, and corresponding fault type labels; constructing to obtain a training data set; wherein the fault types comprise motor over-temperature, motor overload, motor stalling, gear wear and output shaft breakage; training the LSTM network by using the training data set to obtain a trained LSTM network model; based on the trained LSTM network model, electric slewing mechanism fault prediction is carried out, a fault prediction result is obtained, and the fault prediction result comprises the probability value of each fault type; and performing fault grading processing according to the fault prediction result.

Description

TECHNICAL FIELD

[0001] The application relates to a safety control method and system for an electric slewing mechanism based on fault prediction, and belongs to the technical field of new energy engineering machinery. BACKGROUND

[0002] In the process of electrification, some engineering machinery with a slewing platform, such as excavators and cranes, have a traditional hydraulic slewing mechanism replaced by an electric slewing mechanism. The power source of the electric slewing mechanism is replaced by a slewing motor, and the slewing motor and a speed reducer are combined to form the electric slewing mechanism, which drives the platform to rotate and recovers brake energy through electric braking, so that the whole machine has higher energy utilization and power.

[0003] Due to the working characteristics of the electric slewing mechanism, such as frequent start-stop, large vibration, high impact and extreme working conditions (dust and gravel), the electric slewing mechanism is prone to mechanical problems such as motor over-temperature, motor stall, motor overload, gear wear and output shaft fracture. Light causes downtime, and heavy causes personal safety accidents. The traditional control method relies on threshold detection, which has the disadvantages of hysteresis (the protection can be triggered only after the fault occurs) and high false alarm rate (a single threshold is affected by environmental interference).

[0004] Therefore, it is necessary to accurately predict the mechanical problems of the electric slewing mechanism in advance, and then take safety control measures such as reducing the speed, twisting the torque, braking and instrument reminding, to effectively reduce the impact and improve the safety of the electric slewing mechanism. The control method of the prior art mainly focuses on the fault detection of the traditional hydraulic electric slewing mechanism, and combines various sensor detections to detect slewing gear faults. There is no fault prediction and detection method specifically for the electric slewing mechanism. SUMMARY

[0005] To solve at least one of the above technical problems, the application provides a safety control method and system for an electric slewing mechanism based on fault prediction, which can predict the occurrence of a fault in advance based on a trained LSTM network model within a certain period of time before the fault occurs.

[0006] The technical solution adopted by the application is as follows:

[0007] In a first aspect, the application provides a safety control method for an electric slewing mechanism based on fault prediction, comprising:

[0008] Obtaining time series data of a motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, transmission shaft X, Y, Z three-direction vibration acceleration of an electric slewing mechanism and corresponding fault type labels, and constructing a training data set; wherein the fault types include motor over-temperature, motor overload, motor stall, gear wear and output shaft fracture;

[0009] training the LSTM network by using the training data set, to obtain a trained LSTM network model;

[0010] performing fault prediction on the electric slewing mechanism based on the trained LSTM network model, to obtain a fault prediction result, wherein the fault prediction result comprises probability values of each fault type;

[0011] performing fault grading processing according to the fault prediction result.

[0012] In some embodiments, time series data of motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, transmission shaft X, Y, Z three-direction vibration acceleration of the electric slewing mechanism and corresponding fault type labels are obtained, including:

[0013] The motor winding temperature, motor three-phase current, motor speed, motor torque, transmission shaft three-direction vibration acceleration are collected by sensors installed on the electric slewing mechanism, the time and type of fault occurrence are recorded, and the time stamps are aligned.

[0014] In some embodiments, the time series data is also preprocessed, specifically including:

[0015] The motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, transmission shaft X, Y, Z three-direction vibration acceleration are subjected to eigenvalue filtering processing;

[0016] The motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, transmission shaft X, Y, Z three-direction vibration acceleration are subjected to eigenvalue normalization processing.

[0017] In some embodiments, the input of the LSTM network model is 9 eigenvalues of motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, transmission shaft X, Y, Z three-direction vibration acceleration, and the output is probability values of each fault type;

[0018] The LSTM network model comprises an input layer, a first layer LSTM, a second layer LSTM, a full connection layer and an output layer, and the data processing process comprises:

[0019] The input layer is used to receive time series data of 9 eigenvalues, with a shape of (T, 9) ), T is the time window width, is the feature dimension, including 9 eigenvalues of motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, transmission shaft X, Y, Z three-direction vibration acceleration;

[0020] The first layer LSTM includes 64 LSTM memory units for extracting short-term and long-term dependency relationships between each feature value and a fault and extracting time sequence features, and the final output at each time is , with a shape of ( ) ;

[0021] The second layer LSTM includes 32 LSTM memory units for receiving the output of the first layer LSTM , analyzing and deciding the time sequence features, and the final output at each time is , with a shape of (1, 32) ;

[0022] The fully connected layer is used to map the features extracted by the second layer LSTM to a fault probability space.

[0023] The output layer is used to output probability values of each fault type to obtain a probability matrix ; wherein respectively represent the predicted probability values of motor over-temperature, motor overload, motor stall, gear wear, and output shaft breakage.

[0024] In some embodiments, a training data set is constructed, including:

[0025] Input sample definition: the dimension of each sample is ( , ), is the time window width, is the feature dimension, the time window sliding step is , and assuming that the total amount of collected time sequence data is seconds, the total amount of sample number N is:

[0026] ;

[0027] The recorded faults in the sample are defined as a fault label matrix , which respectively represents the occurrence of motor over-temperature, motor overload, motor stall, gear wear, and output shaft breakage, and takes a value of 0 or 1, 0 indicating normal and 1 indicating fault occurrence;

[0028] For each type of fault j, the warning time window is defined as ,

[0029] ;

[0030] wherein is the label of the jth fault at the kth time, is the actual time of occurrence of the fault.

[0031] ​​In some embodiments, during the LSTM network model training process, a loss function The binary cross-entropy is expressed as:

[0032] ;

[0033] In the formula is the total number of samples, is the true label of the jth fault of the ith sample, 0 represents no occurrence, and 1 represents occurrence; is the probability of the jth fault predicted by the LSTM network.

[0034] In some embodiments, according to the fault prediction result, a fault grading process is performed, including:

[0035] When the probability value is less than 0.5, it is defined as a normal state, and continuous monitoring is performed;

[0036] When the probability value is greater than or equal to 0.5 and less than 0.7, it is defined as a pre-warning level fault, and the instrument displays a reminder;

[0037] When the probability value is greater than or equal to 0.7 and less than 0.9, it is defined as a power limit level fault;

[0038] When the probability value is greater than or equal to 0.9, it is defined as an emergency protection level fault;

[0039] When the power limit level fault or the emergency protection level fault occurs, the instrument displays a reminder, and corresponding fault action restrictions are performed according to the fault type.

[0040] Further, according to the fault type, corresponding fault action restrictions are performed, specifically including:

[0041] If the fault type is motor over-temperature, when the power limit level fault occurs, the motor torque is limited to 70%, and the cooling water pump speed is increased to 80%; when the emergency protection level fault occurs, the motor torque is limited to 30%, and the cooling water pump speed is increased to 100%;

[0042] If the fault type is motor overload or motor locked-rotor, when the power limit level fault occurs, the motor torque is limited to 50%; when the emergency protection level fault occurs, the motor torque is limited to 0;

[0043] If the fault type is gear wear, when the power limit level fault occurs, the motor torque is limited to 50%; when the emergency protection level fault occurs, the motor torque is limited to 10%;

[0044] If the fault type is output shaft breakage, when the power limit level fault occurs, the motor torque is limited to 50%; when the emergency protection level fault occurs, the motor torque is limited to 0, and the parking brake is activated.

[0045] In a second aspect, the application provides a safety control system for an electric slewing mechanism based on fault prediction, comprising a processor and a storage medium.

[0046] The storage medium is configured to store instructions.

[0047] The processor is configured to operate according to the instructions to perform the method according to the first aspect.

[0048] In a third aspect, the application provides a computer-readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the method according to the first aspect.

[0049] In a fourth aspect, the application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the method according to the first aspect when executing the computer program.

[0050] In a fifth aspect, the application provides a computer program product comprising a computer program, which, when executed by a processor, implements the method according to the first aspect.

[0051] Advantages: The safety control method and system for an electric slewing mechanism based on fault prediction provided by the application can design an LSTM network to analyze the coupling of time series data of multiple sensors, predict the fault probability in advance, and perform hierarchical processing on the fault. The occurrence of the fault can be identified in advance, and corresponding protection measures can be taken to avoid the occurrence of safety accidents. The method has the following advantages: (1) The time series data of multiple sensors are collected to identify the fault classification, and the coupling relationship in the time series data of multiple sensors is analyzed, which can effectively reduce the false alarm rate. (2) The LSTM network is used to analyze the trend of long-term and short-term sensor time series data, and the fault label alignment design can predict the fault probability in advance. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 FIG. 1 is a flowchart of a safety control method for an electric slewing mechanism based on fault prediction according to an embodiment of the application;

[0053] Figure 2 FIG. 3 is a schematic diagram of an LSTM network structure according to an embodiment of the application;

[0054] Figure 3 FIG. 4 is a schematic diagram of an LSTM memory cell structure according to an embodiment of the application;

[0055] Figure 4 FIG. 5 is a schematic diagram of a safety control system for an electric slewing mechanism based on fault prediction according to an embodiment of the application. DETAILED DESCRIPTION

[0056] The application will be further described below with reference to the drawings and embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the application, and cannot be used to limit the protection scope of the application.

[0057] In the description of the application, the meaning of several is more than one, the meaning of multiple is more than two, greater than, less than, more than, etc. are understood as not including the number, above, below, etc. are understood as including the number. If it is described as first, second, it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the sequence of indicated technical features.

[0058] In the description of the application, the description of the terms "one embodiment", "some embodiments", "exemplary embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. In the description, the exemplary description of the above terms does not necessarily mean the same embodiment or example. Moreover, the described specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0059] The term "and / or", only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, A and B together, and B alone. In addition, the character " / ", generally represents that the front and rear associated objects are in an "or" relationship.

[0060] Embodiment 1: The embodiment provides a safety control method for an electric slewing mechanism based on fault prediction, as shown in the following formula: Figure 1 As shown in the following formula:

[0061] S1, obtaining time series data of motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, transmission shaft X, Y, Z three-direction vibration acceleration of the electric slewing mechanism and corresponding fault type label, and constructing a training data set; wherein the fault types include motor overtemperature, motor overload, motor locked-rotor, gear wear and output shaft broken shaft.

[0062] In some embodiments, step S1 specifically comprises: collecting motor winding temperature, motor three-phase current, motor speed, motor torque, transmission shaft three-direction vibration acceleration through the sensors installed on the electric slewing mechanism, recording the time and type of fault occurrence, and aligning the time stamp.

[0063] In some embodiments, step S1 further comprises preprocessing the time series data, specifically comprising:

[0064] Perform eigenvalue filtering on the motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, and transmission shaft X, Y, Z three-axis vibration acceleration;

[0065] The characteristic value normalization processing is performed on the motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, and transmission shaft X, Y, Z three-axis vibration acceleration.

[0066] In this embodiment, sensor data collection and preprocessing include:

[0067] 1. Data Collection

[0068] The sensors installed on the electric rotating mechanism collect nine physical characteristics, including the motor winding temperature (unit: °C), the motor three-phase (U, V, W) current (unit: A), the motor speed (unit: rpm), the motor torque (unit: N•M), and the three-axis (X, Y, Z) vibration acceleration of the drive shaft (unit: g), including the time series of each characteristic.

[0069] Record the time and type of fault. The fault types are divided into five types: motor overtemperature, motor overload, motor stall, gear wear, and output shaft breakage.

[0070] 2. Preprocessing

[0071] a) Align timestamps to unify the collected sensor data time and fault recording time.

[0072] b) Eigenvalue filtering: Perform first-order low-pass filtering on the time series data collected from each sensor to remove high-frequency noise.

[0073] c) Eigenvalue normalization: In order to eliminate the influence of different units and dimensions of eigenvalues, the collected sensor data is normalized using Min-Max to limit the range of each eigenvalue to [0 1].

[0074] S2. Train the LSTM network using the training data set to obtain a trained LSTM network model;

[0075] It should be noted that the input of the LSTM network model is nine characteristic values: motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, and transmission shaft X, Y, Z three-axis vibration acceleration, and the output is the probability value of each fault type.

[0076] In this embodiment, the LSTM network training sample construction includes:

[0077] 1. Input sample definition

[0078] The time window width is =600, corresponding to a 600s time window, sampling frequency is 1 Hz.

[0079] Feature dimension = 9, corresponding to the above-mentioned 9 feature values collected.

[0080] Then the dimension of each sample is (F x T) , ).

[0081] The time window sliding step size is = 10, assuming that the total amount of data collected is seconds, then the total amount of samples N is:

[0082]

[0083] 2. Fault label definition and alignment rules

[0084] The record fault in the sample is defined as a fault label matrix , respectively representing the occurrence of motor over-temperature, motor overload, motor stall, gear wear, output shaft breakage, taking values of 0 or 1, 0 indicating normal and 1 indicating fault.

[0085] In order to make the LSTM network have the fault probability prediction ability, if a certain fault occurs within time after a certain time point, the fault label of the sample in this time period is set to 1. For each type of fault j, define its early warning time window as ,

[0086]

[0087] is the actual time of occurrence of this type of fault.

[0088] Take different values according to the type of fault, such as 60s for motor over-temperature fault, 10s for motor overload fault, 5s for motor stall fault, 300s for gear wear fault, 5s for output shaft breakage fault, etc.

[0089] The structure of the LSTM network is shown in Figure 2 .

[0090] 1. Input layer (Input Layer)

[0091] The role of this input layer is to receive the sequence data of the above-mentioned 9 feature values collected by the original sensor, without any calculation, the shape of this input layer is (F x T), the dimension of each time input sample is F.

[0092] ​2. First LSTM layer (LSTM Layer 1)

[0093] The main role of this layer of LSTM is to extract the short-term and long-term dependencies between each feature value and the fault, and extract the time sequence features. The number of LSTM memory cells is designed to be 64.

[0094] The internal structure of each LSTM memory cell is as shown in Figure 3 For each time step, the hidden state and memory cell are calculated step by step, and the internal calculation formula is as follows:

[0095] Forget gate:

[0096] Input gate:

[0097] Candidate memory cell:

[0098] Update memory cell:

[0099] Output gate:

[0100] Update hidden state:

[0101] In the formula, represents a sigmoid activation function; represents a tanh activation function; , , , , , , are trainable parameter matrices and biases of the forget gate, the input gate, the candidate memory cell, and the output gate, respectively; represents element-wise multiplication;

[0102] This layer retains the output of each time step for use by the lower layer, and the final output at each time is , whose shape is ( ).

[0103] 3. Second LSTM layer (LSTM Layer 2)

[0104] The role of the second LSTM layer is to receive the output of the first LSTM layer and further analyze and decide the time sequence features. The number of LSTM layer memory cells is designed to be 32.

[0105] The internal memory cell calculation formula of this layer of LSTM is the same as that of the first layer of LSTM.

[0106] The layer only retains the output of the last time step for the lower layer to use, and the final output at each time with shape (1, 32).

[0107] 4. Full connection layer (Dense Layer)

[0108] The role of this layer is to map the features extracted by the LSTM layer to the fault probability space. The number of units designed is 16. The calculation formula of this layer is:

[0109]

[0110] where -ReLu activation function, which can effectively alleviate the gradient vanishing problem and improve computational efficiency. is the trainable parameter matrix and bias of this layer.

[0111] 5. Output layer (Output Layer)

[0112] The role of this layer is to output the classification fault probability. Since the fault classification to be predicted is five categories, the number of units designed in this layer is five, corresponding to the probability matrix of motor over-temperature, motor overload, motor stall, gear wear, and output shaft breakage . The calculation formula of this layer is:

[0113]

[0114] where - sigmoid activation function, which realizes independent multi-label probability prediction. is the trainable parameter matrix and bias of this layer.

[0115] 6. Loss function

[0116] In the sample training stage, the training goal is to optimize the loss function to make the model accurately predict the joint occurrence probability of the five types of faults (over-temperature, overload, stall, wear, and shaft breakage). The loss function The binary cross-entropy is selected, defined as shown below:

[0117] ;

[0118] where is the total number of samples, is the true label of the jth fault of the ith sample, 0 indicates that it does not occur, and 1 indicates that it occurs; is the probability of the jth fault predicted by the LSTM network.

[0119] In the training process, the loss function is optimized by iteration of samples, such as selecting Adam to optimize the parameter matrix and bias of each layer network, so that the loss function tends to 0.

[0120] S3, based on the trained LSTM network model, the fault prediction of the electric slewing mechanism is carried out, and a fault prediction result is obtained, wherein the fault prediction result includes probability values of each fault type;

[0121] S4, according to the fault prediction result, fault grading processing is carried out.

[0122] In some embodiments, step S4, according to the fault prediction result, fault grading processing is carried out, including:

[0123] When the probability value is less than 0.5, it is defined as a normal state, and continuous monitoring is carried out;

[0124] When the probability value is greater than or equal to 0.5 and less than 0.7, it is defined as a pre-warning level fault, and the instrument displays a reminder;

[0125] When the probability value is greater than or equal to 0.7 and less than 0.9, it is defined as a limited power level fault;

[0126] When the probability value is greater than or equal to 0.9, it is defined as an emergency protection level fault;

[0127] When the limited power level fault and the emergency protection level fault occur, the instrument displays a reminder, and corresponding fault action restrictions are carried out according to the fault type, such as Table 1, which specifically includes:

[0128] If the fault type is motor over-temperature, when the limited power level fault occurs, the motor torque is limited to 70%, and the speed of the cooling water pump is increased to 80%; when the emergency protection level fault occurs, the motor torque is limited to 30%, and the speed of the cooling water pump is increased to 100%;

[0129] If the fault type is motor overload or motor locked-rotor, when the limited power level fault occurs, the motor torque is limited to 50%; when the emergency protection level fault occurs, the motor torque is limited to 0;

[0130] If the fault type is gear wear, when the limited power level fault occurs, the motor torque is limited to 50%; when the emergency protection level fault occurs, the motor torque is limited to 10%;

[0131] If the fault type is output shaft breakage, when the limited power level fault occurs, the motor torque is limited to 50%; when the emergency protection level fault occurs, the motor torque is limited to 0, and the parking brake is applied.

[0132] When multiple fault types are predicted, the speed, torque and power of the motor can be limited in a superimposed manner.

[0133] Table 1: Fault limiting action

[0134]

[0135] Embodiment 2: Based on embodiment 1, this embodiment provides a fault prediction based electric slewing mechanism safety control system, comprising a processor and a storage medium;

[0136] The storage medium is used to store instructions;

[0137] The processor is used to operate according to the instructions to perform the method according to embodiment 1.

[0138] In some specific embodiments, as shown in Figure 4 The electric slewing mechanism safety control system comprises:

[0139] 1. Data perception layer

[0140] Through the sensors installed on the electric slewing mechanism, the motor winding temperature size (unit: ℃), the motor three-phase (U, V, W) current size (unit: A), the motor speed size (unit: rpm), the motor torque size (unit: N•M), and the transmission shaft three-direction (X, Y, Z) vibration size (unit: g) are collected; and the data is filtered and normalized.

[0141] 2. LSTM network prediction layer

[0142] The data processed by the data perception layer is provided to the trained LSTM network, and the LSTM network performs fault probability prediction according to the time series data and outputs the prediction result.

[0143] 3. Fault grading control layer

[0144] According to the probability prediction result, the fault is graded, when the probability value is lower than 0.5, it is defined as normal state, and continuous monitoring is performed; when the probability value is between 0.5~07, it is defined as pre-warning level fault, only instrument reminder is performed; when the probability value is between 0.7~0.9, it is defined as limited power level fault; when the probability value is between 0.9~1, it is defined as emergency protection level fault.

[0145] 4. Execution layer

[0146] According to the fault grading result, the corresponding control signal conversion is performed. The safety components involved include the cooling water pump, the motor, the parking brake, the instrument, the output duty cycle PWM, the speed / torque limit value, the brake signal, and the display signal.

[0147] Embodiment 3: Based on embodiment 1, this embodiment provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the method of embodiment 1.

[0148] Example 4: Based on Example 1, this example provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the method described in Example 1 when executing the computer program.

[0149] Example 5: Based on Example 1, this example provides a computer program product, including a computer program, which implements the method described in Example 1 when executed by a processor.

[0150] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0152] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0153] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0154] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A safety control method for an electric rotary mechanism based on fault prediction, characterized in that: include: Obtain time series data including the motor winding temperature of the electric rotary mechanism, the motor's three-phase currents (U, V, and W), the motor speed, the motor torque, and the transmission shaft's X, Y, and Z-axis vibration accelerations, as well as the corresponding fault type labels, to construct a training dataset. The fault types include motor overheating, motor overload, motor stall, gear wear, and output shaft breakage; Using the training data set to train the LSTM network to obtain a trained LSTM network model; Perform fault prediction on the electric rotary mechanism based on the trained LSTM network model to obtain a fault prediction result, wherein the fault prediction result includes a probability value of each fault type; Fault classification processing is performed according to the fault prediction result.

2. The method according to claim 1, characterized in that Obtain time series data including the motor winding temperature of the electric rotating mechanism, the motor's three-phase currents U, V, and W, the motor speed, the motor torque, and the vibration acceleration of the transmission shaft in the X, Y, and Z directions, as well as the corresponding fault type labels, including: Sensors installed on the electric slewing mechanism collect motor winding temperature, motor three-phase current, motor speed, motor torque, and three-dimensional vibration acceleration of the drive shaft, record the time and type of fault occurrence, and align the timestamps.

3. The method according to claim 2, characterized in that It also includes preprocessing of time series data, including: Perform eigenvalue filtering on the motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, and transmission shaft X, Y, Z three-axis vibration acceleration; The characteristic value normalization processing is performed on the motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, and transmission shaft X, Y, Z three-axis vibration acceleration.

4. The method according to claim 1, wherein The input of the LSTM network model is nine characteristic values: motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, and transmission shaft X, Y, Z three-axis vibration acceleration. The output is the probability value of each fault type. The LSTM network model includes an input layer, a first LSTM layer, a second LSTM layer, a fully connected layer, and an output layer. The data processing process includes: The input layer is used to receive time series data of 9 eigenvalues, with a shape of ( ), is the time window width, It is a feature dimension, including 9 feature values: motor winding temperature, motor U, V, W three-phase current, motor speed, motor torque, and transmission shaft X, Y, Z three-axis vibration acceleration; The first LSTM layer includes 64 LSTM memory units, which are used to extract the short-term and long-term dependencies between each feature value and the fault, and extract the time series features. The final output at each moment is , the shape is ( ); The second LSTM layer includes 32 LSTM memory units, which are used to receive the output of the first LSTM layer. , analyze and make decisions on the time series features, and the final output at each moment is , shape is (1, 32); The fully connected layer is used to extract the features of the second layer LSTM Mapping to the fault probability space; The output layer is used to output the probability value of each fault type and obtain the probability matrix ;in They represent the predicted probability values ​​of motor overtemperature, motor overload, motor stall, gear wear, and output shaft breakage failures respectively.

5. The method according to claim 1, wherein The training dataset is constructed, including: Input sample definition: The dimension of each sample is ( , ), is the time window width, is the feature dimension, and the time window sliding step is , assuming that the total amount of time series data collected is seconds, the total number of samples N is: ； The recorded faults in the sample are defined as a fault label matrix in sequence , respectively indicating the occurrence of motor overtemperature, motor overload, motor stall, gear wear, and output shaft breakage. The value is 0 or 1, 0 indicates normal, and 1 indicates a fault. For each type of fault j, the warning time window is defined as , ； in for The label of the j-th fault at time instant, The actual time when the fault occurs.

6. The method according to claim 1, characterized in that During the training process of the LSTM network model, the loss function Using binary cross entropy, it is expressed as: ； In the formula is the total number of samples, is the true label of the jth type of fault of the i-th sample, 0 means it did not occur, and 1 means it occurred; The probability of the j-th type of failure predicted by the LSTM network.

7. The method according to claim 1, characterized in that Performing fault classification processing according to the fault prediction result includes: When the probability value is lower than 0.5, it is defined as a normal state and is continuously monitored; When the probability value is greater than or equal to 0.5 and less than 0.7, it is defined as a warning level fault and the instrument displays a reminder; When the probability value is greater than or equal to 0.7 and less than 0.9, it is defined as a power-limiting level fault; When the probability value is greater than or equal to 0.9, it is defined as an emergency protection level failure; When a power limit level fault or an emergency protection level fault occurs, the instrument will display a reminder and perform corresponding fault action restrictions based on the fault type.

8. The method according to claim 7, characterized in that The corresponding fault action restrictions are implemented according to the fault type, including: If the fault type is motor overtemperature, when a power limit level fault occurs, the motor torque is limited to 70% and the cooling water pump speed is increased to 80%; when an emergency protection level fault occurs, the motor torque is limited to 30% and the cooling water pump speed is increased to 100%; If the fault type is motor overload or motor stall, when a power limit fault occurs, the motor torque is limited to 50%; when an emergency protection fault occurs, the motor torque is limited to 0; If the fault type is gear wear, when a power limit level fault occurs, the motor torque is limited to 50%; when an emergency protection level fault occurs, the motor torque is limited to 10%; If the fault type is output shaft breakage, when a power limit level fault occurs, the motor torque is limited to 50%; when an emergency protection level fault occurs, the motor torque is limited to 0 and the parking brake is applied.

9. A safety control system for an electric rotary mechanism based on fault prediction, characterized in that: including processors and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images