Motor abnormality detection apparatus and method, and non-transitory computer readable storage medium

Abstract

The invention relates to a motor abnormality detection device, a motor abnormality detection method, and a non-transitory computer-readable storage medium storing a program for executing the method. A motor abnormality detection device according to one aspect of the present invention is used for detecting an abnormality of a motor generating a steering assist torque in an electric power steering system of a vehicle, and may include: a memory; and a processor that stores an artificial neural network model that receives state data related to a state of the vehicle and operation data related to a steering operation by a driver of the vehicle, and estimates a physical quantity related to an output of the motor, the processor inputs the state data and the operation data into the artificial neural network model and performs calculation to output an estimated value of the physical quantity, and the processor detects whether the motor is abnormal or not by comparing the estimated value with a measured value of the physical quantity.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the priority of Korean Patent Application No. 10-2023-0104199 filed with the Korean Intellectual Property Office on August 9, 2023, the disclosure of which is incorporated herein by reference in its entirety. Technical Field

[0003] The present invention relates to a motor abnormality detection device, method and non-transitory computer-readable storage medium, and more particularly, to a motor abnormality detection device for detecting an abnormality of a motor that generates steering assist torque in an electric power steering (EPS) system of a vehicle, a motor abnormality detection method and a non-transitory computer-readable storage medium storing a program for executing the motor abnormality detection method. Background Art

[0004] The Electric Power Steering (EPS) system is a device that uses a motor to assist the vehicle's steering. The EPS system analyzes the detection data related to the operation of the steering wheel, the vehicle speed, etc. by controlling the motor and provides appropriate steering assist torque. Unlike the traditional hydraulic power steering system, the EPS system can also be applied to electric vehicles and is also highly efficient in terms of space and fuel efficiency.

[0005] The EPS system has the advantages described above. However, during driving, a failure of the EPS system, i.e., loss-of-assist (LoA), has a significant impact on the steering performance of the vehicle and may develop into a fatal accident. The main cause of LoA is the misoperation of the motor that generates the steering assist torque. Therefore, in order to ensure the safety of the vehicle, functional safety (fail safety) related to the motor used to generate the steering torque in the EPS system is required.

[0006] However, conventional trial methods related to the functional safety of motors in EPS systems focus on follow-up measures based on failures rather than preventing failures in advance. On the other hand, although research and development for early detection of defects in EPS motors is partially underway, model-based trial methods that simultaneously estimate the state and defects of the system are mainly used. This trial method has the limitation that it cannot be expanded to vehicles in the form of embedded applications. Summary of the invention

[0007] Problem that the invention aims to solve

[0008] The present invention aims to solve the problems mentioned above. The purpose of the present invention is to provide a motor abnormality detection device, a motor abnormality detection method and a non-temporary computer-readable storage medium storing a program for executing the motor abnormality detection method, which can predict the reduction of motor function before the motor fails in the EPS system.

[0009] Another object of the present invention is to provide a motor abnormality detection device, a motor abnormality detection method, and a non-temporary computer-readable storage medium storing a program for executing the motor abnormality detection method, which can predict the decrease in motor output in advance based on a digital twin algorithm based on artificial intelligence before LoA (Loss-of-Assist) or a serious decrease in motor torque output occurs in the EPS system.

[0010] Another object of the present invention is to provide a motor abnormality detection device, a motor abnormality detection method, and a non-temporary computer-readable storage medium storing a program for executing the motor abnormality detection method, which are capable of predicting the output drop of a motor in an EPS system from a signal that can be obtained through a vehicle's Controller Area Network (CAN).

[0011] The objects of the present invention are not limited to the above objects, and other objects not mentioned will be clearly understood by those skilled in the art to which the present invention pertains through the following description.

[0012] Means used to solve problems

[0013] According to one aspect of the present invention, there is provided a motor abnormality detection device, which is a motor abnormality detection device for detecting an abnormality of a motor for generating steering assist torque in an EPS (Electric Power Steering) system of a vehicle, and comprises: a memory; and a processor, wherein the memory stores an artificial neural network model, the artificial neural network model receives status data related to the status of the vehicle, operation data related to the steering operation of the driver of the vehicle, and estimates a physical quantity related to the output of the motor, the processor inputs the status data and the operation data into the artificial neural network model and performs calculations to output an estimated value of the physical quantity, and the processor detects whether the motor is abnormal by comparing the estimated value with an actual measured value of the physical quantity.

[0014] In the motor abnormality detection device according to one aspect of the present invention, the state data and the operation data may be constituted by signals that can be acquired through a CAN of the vehicle.

[0015] In the motor abnormality detection device according to one aspect of the present invention, the physical quantity may be a steering torque generated when the driver operates a steering wheel of the vehicle.

[0016] In a motor abnormality detection device according to one aspect of the present invention, the status data may include the longitudinal velocity of the vehicle, the lateral acceleration of the vehicle, the yaw rate of the vehicle, and the wheel velocity of each of the multiple wheels of the vehicle.

[0017] In the motor abnormality detection device according to one aspect of the present invention, the operation data may include a steering angle and a steering angular velocity.

[0018] In the motor abnormality detection device according to one aspect of the present invention, the artificial neural network model may include a generative adversarial network (GAN).

[0019] In a motor abnormality detection device according to one aspect of the present invention, the artificial neural network model may include: a generator, which receives the state data and the operation data and generates the estimated value; and a discriminator, which receives measured value related data including the state data, the operation data and the measured value, and outputs a discriminator value for the measured value related data.

[0020] In the motor abnormality detection device according to one aspect of the present invention, the generator may be composed of a multivariate transformer.

[0021] In the motor abnormality detection device according to one aspect of the present invention, the discriminator may further receive estimated value related data including the state data, the operation data, and the estimated value, and further output a discriminant value for the estimated value related data.

[0022] In a motor abnormality detection device according to one aspect of the present invention, the artificial neural network model is a model constructed by alternately learning the generator and the discriminator, and the state data, the operation data and the measured values ​​used in the learning may be data obtained when the vehicle and the motor are in normal conditions.

[0023] In the motor abnormality detection device according to one aspect of the present invention, the estimated value may track an actual measured value of the physical quantity acquired in a state where the vehicle and the motor are normal.

[0024] In a motor abnormality detection device according to one aspect of the present invention, the memory also stores an abnormality detection model, and the processor can input error data related to the difference between the estimated value and the measured value into the abnormality detection model to determine whether the motor is abnormal or not.

[0025] In the motor abnormality detection device according to one aspect of the present invention, the abnormality detection model may include a One-Class Support Vector Machine (OCSVM) algorithm.

[0026] In a motor abnormality detection device according to one aspect of the present invention, there are multiple data sets, the data sets include the state data, the operation data, the measured values ​​and the estimated values, and the error data may include the average error and standard deviation of the errors between the measured values ​​and the estimated values ​​obtained from each of the multiple data sets, the maximum absolute error between the measured values ​​and the estimated values ​​of the multiple data sets, and the discrimination value generated by the discriminator for the measured value related data of the multiple data sets.

[0027] According to another aspect of the present invention, a motor abnormality detection method is provided, which is used to detect an abnormality of a motor for generating steering assist torque in an EPS system of a vehicle, and the motor abnormality detection method includes: a step in which a processor receives status data related to a status of the vehicle, operation data related to a steering operation of a driver of the vehicle, and a measured value of a physical quantity related to the output of the motor; a step in which the processor inputs the status data and the operation data into an artificial neural network model and performs calculations to output an estimated value of the physical quantity; and a step in which the processor detects whether the motor is abnormal by comparing the estimated value with the measured value.

[0028] In the motor abnormality detection method according to another aspect of the present invention, the state data and the operation data may be constituted by signals that can be acquired through a CAN of the vehicle.

[0029] In the motor abnormality detection method according to another aspect of the present invention, the physical quantity may be a steering torque generated when the driver operates a steering wheel of the vehicle.

[0030] In a motor abnormality detection method according to another aspect of the present invention, the status data may include a longitudinal velocity of the vehicle, a lateral acceleration of the vehicle, a yaw rate of the vehicle, and a wheel speed of each of a plurality of wheels of the vehicle.

[0031] In a motor abnormality detection method according to another aspect of the present invention, the operation data may include a steering angle and a steering angular velocity.

[0032] In a motor abnormality detection method according to another aspect of the present invention, the artificial neural network model may include a generative adversarial network (GAN).

[0033] In a motor abnormality detection method according to another aspect of the present invention, the artificial neural network model includes a generator, which receives the state data and the operation data and generates the estimated value, and the step of outputting the estimated value of the physical quantity may include: the processor inputs the state data and the operation data to the generator, and outputs the estimated value generated by the generator.

[0034] In a motor abnormality detection method according to another aspect of the present invention, the artificial neural network model also includes a discriminator, which receives measured value-related data including the state data, the operation data and the measured value, and outputs a discriminant value for the measured value-related data. The step of outputting the estimated value of the physical quantity may also include: the processor inputs the measured value-related data including the state data, the operation data and the measured value of the physical quantity into the discriminator, and outputs a discriminant value for the measured value-related data generated by the discriminator.

[0035] In a motor abnormality detection method according to another aspect of the present invention, the step of detecting whether the motor is abnormal or not may include: a step in which the processor calculates error data related to the difference between the estimated value and the measured value; and a step in which the processor inputs the error data into a one-class support vector machine (OCSVM) algorithm to perform calculations.

[0036] In a motor abnormality detection method according to another aspect of the present invention, there are multiple data sets, the data sets include the state data, the operation data, the estimated values ​​and the measured values, and the error data may include the average error and standard deviation of the errors between the measured values ​​and the estimated values ​​obtained from each of the multiple data sets, the maximum absolute error between the measured values ​​and the estimated values ​​of the multiple data sets, and the discrimination value generated by the discriminator for the measured value related data of the multiple data sets.

[0037] In a motor abnormality detection method according to another aspect of the present invention, the artificial neural network model is a model constructed by alternately learning the generator and the discriminator, and the state data, the operation data and the measured values ​​used in the learning may be data obtained when the vehicle and the motor are in normal conditions.

[0038] According to still another aspect of the present invention, there is provided a non-transitory computer-readable storage medium storing a program, the program including at least one instruction for executing the motor abnormality detection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The above description and other objects, features and advantages of the present invention will become more apparent to those skilled in the art by describing in detail exemplary embodiments of the present invention with reference to the accompanying drawings, in which:

[0040] Figure 1 FIG. 1 is a diagram showing a configuration of a motor abnormality detection device according to an embodiment of the present invention.

[0041] Figure 2 It is a diagram showing the EPS system, status data, and operation data of the vehicle.

[0042] Figure 3 is a diagram showing a model stored in a memory of a motor abnormality detection device according to an embodiment of the present invention.

[0043] Figure 4 It is shown Figure 3 Figure 2 shows the detailed structure of the artificial neural network model.

[0044] Figure 5 It is shown Figure 4 A diagram showing the detailed structure of the generator of the artificial neural network model.

[0045] Figure 6 1 is a diagram showing the operation of a motor abnormality detection device according to an embodiment of the present invention.

[0046] Figure 7: is a table showing the effect of the motor abnormality detection device according to an embodiment of the present invention.

[0047] Figure 8 is a flow chart of a motor abnormality detection method according to an embodiment of the present invention.

[0048] Fig. 9 Detailed flowchart of the step of estimating the value of the output physical quantity in the motor abnormality detection method according to an embodiment of the present invention.

[0049] Fig.10 Detailed flowchart of the step of detecting whether a motor is abnormal in a motor abnormality detection method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The embodiments of the present invention will be described in detail below so that those skilled in the art can easily implement the present invention. The present disclosure can be implemented in various forms and is not limited to the embodiments described herein. In order to clearly describe the present disclosure, parts not related to the description are omitted in the accompanying drawings, and the same or similar parts will be given the same reference numerals throughout the specification.

[0051] The words and terms used in the specification and claims are not limited to being interpreted as their ordinary or dictionary meanings, but should be interpreted as meanings and concepts consistent with the technical spirit of the present disclosure based on the principle that inventors can define terms in order to best describe their invention.

[0052] In this specification, it should be understood that terms such as "including" or "having" are intended to specify the existence of the features, numbers, steps, operations, components, parts or combinations thereof described in this specification, and do not exclude the possibility of the existence or addition of one or more other features, numbers, steps, operations, components, parts or combinations thereof.

[0053] Figure 1 is a diagram showing a structure of a motor abnormality detection device according to an embodiment of the present invention.

[0054] According to an embodiment of the present invention, the motor abnormality detection device 1 can detect the abnormality of the motor used to generate the steering assist torque in the EPS system of the vehicle. In more detail, according to an embodiment of the present invention, the motor abnormality detection device 1 can predict (predict) the functional degradation of the motor before the failure of the motor occurs through an artificial intelligence-based digital twin algorithm for the vehicle. Specifically, the digital twin algorithm can virtually construct the EPS system of the vehicle.

[0055] According to an embodiment of the present invention, the motor abnormality detection device 1 detects the functional degradation of the motor before the EPS system of the vehicle loses assistance (LoA) due to the motor failure. In particular, the motor abnormality detection device 1 according to an embodiment of the present invention can also sensitively distinguish a slight degradation (deterioration) of the motor performance.

[0056] Therefore, according to an embodiment of the present invention, a motor failure of the EPS system can be predicted in advance, and as a result, measures can be taken in advance before the motor fails, thereby preventing the motor from failing.

[0057] Reference Figure 1 According to an embodiment of the present invention, a motor abnormality detection device 1 may include a memory 100 and a processor 200 .

[0058] The memory 100 may store data, models, etc. required for detecting an abnormality of the motor. For example, the memory 100 may include a hardware device configured to store and execute program instructions such as ROM, RAM, flash memory, etc. In addition, the memory 100 may include magnetic media such as floppy disks and tapes, optical recording media such as CD-ROM (Compact Disk Read Only Memory) and DVD (Digital Video Disk), magneto-optical media such as floptical disks, etc.

[0059] The processor 200 may input the data into the model and perform calculations (operations). For example, the processor 200 may be formed as a hardware unit that performs calculations and controls on a computer. The processor 200 may include at least one arithmetic logic unit (ALU) and a register.

[0060] The memory 100 stores an artificial neural network model that receives state data related to the state of the vehicle and operation data related to the steering operation of the driver of the vehicle, thereby estimating a physical quantity related to the output of the motor.

[0061] The processor 200 inputs the state data and the operation data into the artificial neural network model and performs calculation, thereby outputting an estimated value of the physical quantity, and the processor 200 detects whether the motor is abnormal by comparing the estimated value with the measured value of the physical quantity.

[0062] In relation to detecting whether the motor is abnormal, the memory 100 may further store an abnormality detection model. In addition, the processor 200 may input error data related to the difference between the estimated value and the measured value into the abnormality detection model, thereby determining whether the motor is abnormal.

[0063] Figure 2 FIG. 1 is a diagram showing the EPS system, status data, and operation data of a vehicle. Figure 2 The EPS system 10 including the motor 18 and the state data D are shown as an example. vehicle and operation data D steering The motor 18 forms a detection target of the motor abnormality detection device 1 according to an embodiment of the present invention.

[0064] Reference Figure 2 The EPS system 10 of the vehicle may include: a steering wheel 11 operated by a driver; a steering shaft 12 connected to the steering wheel 11 and transmitting steering force to the wheels of the vehicle; a steering angle sensor 13 arranged on the steering shaft 12 and used to measure the steering angle and steering angular velocity of the steering wheel 11; a steering torque sensor 14 arranged on the steering shaft 12 and used to measure the steering torque of the steering wheel 11; and a vehicle state measurement module 15 including one or more sensors for measuring information related to the vehicle state.

[0065] The vehicle state measurement module 15 can measure the longitudinal velocity V x , lateral acceleration y 、Yaw rate θ′ z and the wheel velocity V of each wheel of the vehicle wheel Typically, a vehicle has four wheels, so the wheel speed V wheel There can be four types of data. In other words, the vehicle state measurement module 15 can measure nine types of data.

[0066] The steering angle θ measured by the steering angle sensor 13 steering and the steering angular velocity θ′ steering , the steering torque T measured by the steering torque sensor 14 steering , the vehicle speed V measured by the vehicle state measurement module 15 x , lateral acceleration a y , yaw angular velocity θ′ z , the wheel speed V of each wheel of the vehicle wheelThe signals are transmitted to an electronic control unit (ECU) 17 via a CAN (Controller Area Network) bus 16 .

[0067] The electronic control unit 17 receives a plurality of data transmitted via the CAN bus 16 and performs calculations, thereby generating a command for the motor 18 for generating the steering assist torque, and transmits it to the motor 18. At this time, the command may be provided in the form of current to the motor 18. The electronic control unit 17 may perform feedback control for the motor 18.

[0068] The torque inputted through the steering wheel 11 and the torque generated by the motor 18 are transmitted to the steering mechanism 19 connected to the wheels of the vehicle. The steering mechanism 19 driven by the torque inputted through the steering wheel 11 and the torque generated by the motor 18 adjusts the direction of the wheels of the vehicle, thereby steering the vehicle.

[0069] As described above, an artificial neural network model is stored in the memory 100, and the artificial neural network model receives state data related to the state of the vehicle, operation data related to the steering operation of the driver of the vehicle, and estimates the physical quantity related to the output of the motor. In addition, the processor 200 inputs the state data and the operation data into the artificial neural network model and performs calculations, thereby outputting an estimated value of the physical quantity, and the processor 200 detects whether the motor is abnormal by comparing the estimated value with the measured value of the physical quantity. In connection with detecting whether the motor is abnormal, an abnormality detection model may be additionally stored in the memory 100.

[0070] In one embodiment of the present invention, the state data may include the longitudinal velocity V of the vehicle. x , lateral acceleration a y , yaw angular velocity θ′ z and the wheel speed V of each of the plurality of wheels of the vehicle wheel When the vehicle has four wheels, the status data may include 7 types of data.

[0071] In addition, the operation data may include a steering angle θ steering and the steering angular velocity θ′ steering As mentioned above, the operation data may include 2 types of data.

[0072] As described above, in the motor abnormality detection device 1 according to an embodiment of the present invention, the state data and the operation data may be constituted by signals that can be obtained through a CAN (Controller Area Network) of the vehicle.

[0073] On the one hand, the physical quantity may be a steering wheel torque T generated when the driver of the vehicle operates the steering wheel 11. steering In other words, in one embodiment of the present invention, the processor 200 outputs an estimated value of the steering wheel torque through calculation, and can detect whether the motor 18 is abnormal by comparing the estimated value with the measured value of the steering wheel torque. In this case, the measured value can be the steering wheel torque T measured by the steering torque sensor 14. steering .

[0074] On the other hand, the Dugoff tire model can be applied to the estimation of the steering wheel torque. Since the Dugoff tire model is a well-known model in the field of automobile engineering, its detailed description is omitted.

[0075] Figure 3 FIG. 2 is a diagram showing a model stored in a memory of a motor abnormality detection device according to an embodiment of the present invention. Figure 3 , an artificial neural network model 110 and an anomaly detection model 120 may be stored in the memory 100 .

[0076] Hereinafter, an artificial neural network model 110 for receiving the state data and the operation data and outputting the estimated value of the steering wheel torque, an abnormality detection model 120 for detecting whether the motor 18 is abnormal by comparing the estimated value of the steering wheel torque with the measured value, and calculations of the processor 200 using the model will be described in detail.

[0077] The artificial neural network model 110 receives the state data and the operation data, and outputs the estimated value of the steering torque. In one embodiment of the present invention, the artificial neural network model 110 may include a generative adversarial network (GAN).

[0078] The generative adversarial network is a network constructed by alternately learning a generator and a discriminator, and the state data, the operation data, and the measured value used in the learning may be data obtained when the vehicle and the motor 18 are in a normal state. Therefore, the estimated value can track the measured value of the physical quantity obtained when the vehicle and the motor 18 are in a normal state.

[0079] In other words, the artificial neural network model 110 may be a virtual twin model in the EPS system of the vehicle. In more detail, the artificial neural network model 1110 may be a deep learning-based neural twin model for the EPS system.

[0080] In one embodiment of the present invention, the artificial neural network model 110 estimates the steering torque by receiving 9 types of data including the state of the steering wheel and the vehicle. At this time, the steering torque estimated by the artificial neural network model 110 can reflect the normal operating conditions.

[0081] Figure 4 It is shown Figure 3 The detailed structure of the artificial neural network model is shown in Figure 2. Figure 4 , the artificial neural network model 110 may include a generator 111 and a discriminator 112 .

[0082] The generator 111 generates the estimated value by receiving the state data and the operation data. The generator 111 may include a neural network. In one embodiment of the present invention, the generator 111 functions as a virtual twin model in the EPS system of the vehicle.

[0083] For example, the generator 111 may be composed of a multivariate transformer. That is, the artificial neural network model 110 may be a GAN based on a multivariate transformer.

[0084] In one embodiment of the present invention, the generator 111 may receive a total of 9 types of data, namely, the longitudinal speed V of the vehicle x , lateral acceleration a y , yaw angular velocity θ′ z , the wheel speed V of each of the plurality of wheels of the vehicle wheel (When the vehicle has four wheels, obtain 4), steering angle θ steering and the steering angular velocity θ′ steering Furthermore, the generator 111 outputs an estimated value of the steering wheel torque by calculating the input data.

[0085] Figure 5 It is shown Figure 4 The detailed structure of the generator of the artificial neural network model shown in FIG. Figure 5 , the generator 111 may include an encoder 111a and a decoder 111b.

[0086] The encoder 111a and the decoder 111b constitute a network for estimating the steering torque. In other words, the generator 111 may be composed of an encoder 111a-decoder 111b structure network trained in an end-to-end manner to estimate (calculate) the steering torque.

[0087] The encoder 111a may be designed to extract high-level features from input data using a transformer block and a linear operation. In this case, the transformer block may include an embedding layer, a self-attention layer, and a feed-forward layer.

[0088] First, in the Transformer module, the embedding layer is able to encode spatiotemporal information into the input sequences using a trainable embedding matrix.

[0089] Next, deep features can be extracted through a self-attention layer and refined using a feed-forward layer.

[0090] At this point, the self-attention layer can apply a scaled dot-product attention mechanism. The self-attention layer can be the most important building block in the Transformer-based Deep Neural Network (DNN) architecture to allow the model to process and understand dependencies in the input sequence.

[0091] The scaled dot product attention first applies layer normalization to the input sequence and processes its result into three intrinsic vectors of query Q, key K and value V through linear operation.

[0092] Next, the attention score AS for each position of the input sequence can be calculated based on the derived vector. The calculation formula of the attention score AS can be shown in the following formula 1:

[0093] [Formula 1]

[0094]

[0095] Among them, σ(·) represents the softmax (function) calculation, ∥x input ∥ dimrepresents the feature level of the input, and (·) represents the inner product.

[0096] The attention score quantifies the relative importance of various positions in the input sequence with respect to a given query. That is, higher scores are assigned to more appropriate positions.

[0097] The output of the transformer module can be provided to two different linear operations to calculate the query, key vector and initial input value of decoder 111b

[0098] The decoder 111 b may use the acquired values ​​and estimate the steering torque through an autoregressive method similar to Sequence-to-Sequence Recurrent Neural Networks.

[0099] Similar to the encoder 111a, the decoder 111b may include a transformer module and a linear operation. Additionally, the decoder 111b may further include a cross-attention layer in the transformer module.

[0100] As described above, the cross-attention layer may perform scaled inner-product attention using the query and key vectors of encoder 111a to learn the dynamic correlation between the input signal and the inferred signal.

[0101] The discriminator 112 receives the measured value related data including the state data, the operation data and the measured value, and outputs a discriminant value for the measured value related data. In addition, the discriminator 112 additionally receives the estimated value related data including the state data, the operation data and the estimated value, and can additionally output a discriminant value for the estimated value related data.

[0102] For example, the discriminator 112 may be configured as a discriminant network for use in adversarial training of a GAN. In other words, the generator 111 and the discriminator 112 may be trained in an adversarial training manner to improve the inference performance and set the GAN model. In more detail, the discriminator 112 may perform optimization alternately with the generator 111 to solve the min-max problem of Wasserstein as shown in the following formula 2.

[0103] [Formula 2]

[0104]

[0105] Among them, x D,Real refers to the actual data set consisting of input data and measured values ​​of steering torque, x D,FakeIt refers to a virtual data sample consisting of input data and estimated value of steering torque.

[0106] In the artificial neural network model 110, the generator 111 is trained to deceive the D,Real and x D,Fake Therefore, the generator 111 can generate an estimated value of the steering wheel torque for tracking the actual value of the steering torque in the EPS system of the vehicle in a normal state.

[0107] In this regard, the loss function for GAN training can be defined as the following formula 3.

[0108] [Formula 3]

[0109]

[0110]

[0111] Where N is the batch size, L G is the loss function of the generator, L D is the loss function of the discriminator.

[0112] The first term of the loss of generator 111 is equivalent to the mean absolute error loss L which guides the training MAE The remaining items of the loss of the generator 111 together with the loss of the discriminator 112 constitute the GAN loss function L GAN .

[0113] Meanwhile, the degradation level of the motor may be predicted based on the estimated value of the physical quantity generated by the generator 111 and the measured value of the physical quantity actually measured.

[0114] In one embodiment of the present invention, the degradation level of the motor may be predicted by the abnormality detection model 120 .

[0115] The abnormality detection model 120 may receive the difference-related error data between the estimated value and the measured value and perform calculations. The abnormality detection model 120 may be used to determine whether the motor 18 is abnormal. In one embodiment of the present invention, the abnormality detection model 120 may include an OCSVM (One-Class Support Vector Machine) algorithm.

[0116] In addition, there are multiple data sets, which include the status data, the operation data, the measured values ​​and the estimated values, and the error data may include the average error and standard error of the errors between the measured values ​​and the estimated values ​​obtained from each of the multiple data sets, the maximum absolute error between the measured values ​​and the estimated values ​​in the multiple data sets, and the discrimination value of the discriminator 112 for the measured value related data of the multiple data sets.

[0117] In other words, a plurality of data sets including the state data, the operation data, the measured values, and the estimated values ​​may be considered as a batch. For example, 100 to 150 data sets (as a specific example, 128 data sets) may be considered as a batch, and the error data may be obtained for each batch.

[0118] The error data obtained for each batch may be input to the abnormality detection model 120. Furthermore, the abnormality level of the motor (performance degradation of the motor) may be detected based on the output (feature) obtained from the abnormality detection model 120 that received the error data.

[0119] Regarding the anomaly detection metric, the F1 score given as shown in the following formula 4 can be considered.

[0120] [Formula 4]

[0121]

[0122] Among them, TP stands for true positive, FP stands for false positive, FN stands for false negative, and positive means the performance of the motor is poor.

[0123] On the other hand, the discriminator 112 of the artificial neural network model 110 can calculate not only the discriminant value for the measured value related data in the data set, but also the discriminant value for the estimated value related data in the data set. The discriminant value of the discriminator 112 for the estimated value related data in the data set can be provided as feedback to the generator 111 of the artificial neural network model 110, thereby being used in training.

[0124] So far, the structure of the motor abnormality detection device 1 according to an embodiment of the present invention has been described in detail. Hereinafter, the operation of the motor abnormality detection device 1 according to an embodiment of the present invention will be described in detail.

[0125] Figure 6FIG. 2 is a diagram showing the operation of a motor abnormality detection device according to an embodiment of the present invention. Figure 6 , the motor abnormality detection device 1 according to an embodiment of the present invention can operate in the following manner.

[0126] First, the memory 100 stores the state data D acquired from the EPS system of the vehicle. vehicle , Operation data D steering And the measured value of steering torque T steering As mentioned above, the state data D vehicle The longitudinal velocity V of the vehicle may be included x , lateral acceleration a y , yaw angular velocity θ′ z , and the wheel speed V of each of the plurality of wheels of the vehicle wheel In addition, the operation data D steering The steering angle θ may be included steering and the steering angular velocity θ′ steering .

[0127] At the same time, multiple state data D vehicle , Operation data D steering And the measured value of steering torque T steering It can be regarded as a batch. In addition, in one embodiment of the present invention, data processing and calculation (operation) can be performed in batch units of data.

[0128] Next, the processor 200 converts the state data D vehicle and operation data D steering The artificial neural network model 110 performs calculations to output an estimated value of the steering wheel torque.

[0129] Next, the processor 200 outputs the error data for each data batch. As described above, the error data may include the average error and standard error of the error between the measured value and the estimated value obtained from each of the plurality of data sets, the maximum absolute error between the measured value and the estimated value of the plurality of data sets, and the discriminant value of the discriminator 112 for the measured value related data of the plurality of data sets.

[0130] Finally, the processor 200 inputs the error data into the abnormality detection model 120 and outputs the result. As described above, the abnormality detection model 120 may include an OCSVM (One-Class Support Vector Machine) algorithm. The abnormality level of the motor (the performance degradation of the motor) may be detected based on the output (feature) obtained from the abnormality detection model 120 that receives the error data.

[0131] Figure 7 : is a table showing the effect of the motor abnormality detection device according to an embodiment of the present invention.

[0132] Figure 7 It shows the case where the vehicle speed is 20 km / h and the degradation level of the motor of the EPS system of the vehicle is 10%, 25% and 40% respectively, the case where the vehicle speed is 30 km / h and the degradation level of the motor of the EPS system of the vehicle is 10%, 25% and 40% respectively, and the case where the vehicle speed is 40 km / h and the degradation level of the motor of the EPS system of the vehicle is 10%, 25% and 40% respectively. The degradation level of the motor is predicted by various technologies (techniques) and models, and the F1 score (abnormality detection index) is calculated according to each technology and model.

[0133] exist Figure 7 In the text, “feature-based” technology (technique) refers to the case where the neural twin structure is not used, and “neural twin-based” refers to the case where a neural twin structure based on deep learning is used.

[0134] Here, “feature-based” techniques include the empirical cumulative distribution-based outlier detection (ECOD) algorithm, the deep one-class classification (DOCC) model, the beta-variational autoencoder (β-VAE) model, and Generative Adversarial Active Learning (GAAL).

[0135] In addition, the "neural network based" technology includes a gated recurrent unit (GRU) model, a long short-term memory (LSTM) model, its attention-based versions (A-GRU, A-LSTM), a simple multivariate transformer model (MVT) without GAN training, and a combination of an artificial neural network model 110 and an anomaly detection model 120 stored in a motor anomaly detection device 1 according to an embodiment of the present invention (MVT-GAN).

[0136] At the same time, Figure 7The techniques and models shown use data obtained from the CAN bus through in-vehicle experiments. For example, the data can be obtained at a sampling rate of 100 Hz.

[0137] Reference Figure 7 It can be confirmed that the combination of the artificial neural network model 110 and the anomaly detection model 120 (MVT-GAN) stored in the motor anomaly detection device 1 according to an embodiment of the present invention shows the highest F1 score in all cases. Specifically, according to an embodiment of the present invention, a 10% decrease in the maximum output of the motor can be predicted as an F1 score of 0.85 or more (that is, the model can very accurately detect the state where the output of the motor decreases by 10%, and the closer the F1 score is to 1, the better its performance).

[0138] In other words, it can be confirmed that the motor abnormality detection device 1 according to one embodiment of the present invention can accurately predict the deterioration level of the motor in the EPS system of the vehicle before the motor completely fails. In particular, it can be confirmed that according to one embodiment of the present invention, even a slight decrease in the output of the motor can be accurately predicted.

[0139] As described above, according to the present invention, it is possible to estimate the future behavior of the system and the remaining useful life (RUL), and predictive diagnosis suitable for predictive maintenance (PdM) applications. Therefore, advance measures and maintenance can be effectively guided before the motor of the vehicle's EPS system fails.

[0140] The above is a detailed description of a motor abnormality detection device according to an embodiment of the present invention. Hereinafter, a motor abnormality detection method according to an embodiment of the present invention will be described.

[0141] Figure 8 is a flow chart of a motor abnormality detection method according to an embodiment of the present invention.

[0142] The motor abnormality detection method (step S100) according to an embodiment of the present invention is a motor abnormality detection method for detecting abnormality of a motor for generating steering assist torque in an EPS system of a vehicle, which can be performed by a motor abnormality detection device 1 according to an embodiment of the present invention.

[0143] Reference Figure 8 , the motor abnormality detection method (step S100) according to an embodiment of the present invention can be performed as follows.

[0144] First, the processor 200 receives state data related to the state of the vehicle, operation data related to the steering operation of the driver of the vehicle, and measured values ​​of physical quantities related to the output of the motor (step S110).

[0145] In one embodiment of the present invention, the state data may include the longitudinal velocity V of the vehicle. x , lateral acceleration ay, yaw angular velocity θ′ z and the wheel speed V of each of the plurality of wheels of the vehicle wheel As described above, when the vehicle has four wheels, the status data may include seven types of data.

[0146] In addition, the operation data may include a steering angle θ steering and the steering angular velocity θ′ steering That is, the operation data may include 2 types of data.

[0147] The data listed as examples of the status data and the operation data are signals that can be obtained through the CAN of the vehicle. As described above, when the status data and the operation data are composed of signals that can be obtained through the CAN of the vehicle, there is no need to configure additional sensors on the vehicle for collecting data, so the efficiency of data collection can be improved.

[0148] In addition, when data needs to be transmitted to the outside of the vehicle, the data transmission can also be effectively performed. For example, when the motor abnormality detection device 1 according to an embodiment of the present invention is configured at a remote location outside the vehicle, and when the motor abnormality detection method according to an embodiment of the present invention is executed at the remote location, the state data and the operation data can be effectively received by communicating during the driving of the vehicle.

[0149] On the other hand, the physical quantity may be a steering wheel torque T generated when the driver of the vehicle operates the steering wheel 11. steering In other words, in one embodiment of the present invention, the processor 200 outputs an estimated value of the steering wheel torque through calculation, and can detect whether the motor 18 is abnormal by comparing the estimated value with the measured value of the steering wheel torque. In this case, the measured value can be the steering torque T measured by the steering torque sensor 14. steering .

[0150] At this time, the plurality of data sets including the state data, the operation data, the measured values ​​and the estimated values ​​can be regarded as a batch. In other words, the processor 200 receives a plurality of data sets and forms a batch in units of a predetermined number, and can also perform the following calculations in units of batches.

[0151] Next, the processor 200 inputs the state data and the operation data into the artificial neural network model and performs calculation, thereby outputting the estimated value of the physical quantity (step S120). As described above, the artificial neural network model 110 may include a generative adversarial network (GAN).

[0152] The artificial neural network model 110 receives the state data and the operation data, and outputs the estimated value of the steering torque. In one embodiment of the present invention, the artificial neural network model 110 may include a generative adversarial network (GAN). In addition, the artificial neural network model 110 may include a generator 111 and a discriminator 112.

[0153] Fig. 9 Detailed flowchart of the step of estimating the value of the output physical quantity in the motor abnormality detection method according to an embodiment of the present invention.

[0154] Reference Fig. 9 , the step of outputting the estimated value of the physical quantity can be performed as follows.

[0155] First, the processor 200 inputs the state data and the operation data to the generator 111, and outputs the estimated value generated by the generator 111 (step S121). The generator 111 receives the state data and the operation data, thereby generating the estimated value (step S121).

[0156] For example, the generator 111 may be composed of a multivariate transformer. In other words, the artificial neural network model 110 may be a GAN based on a multivariate transformer. In one embodiment of the present invention, the generator 111 functions as a virtual twin model in the EPS system of the vehicle.

[0157] The generator 111 can receive a total of 9 types of data, namely, the longitudinal speed V of the vehicle x , lateral acceleration a y , yaw angular velocity θ′ z , the wheel speed V of each of the plurality of wheels of the vehicle wheel (When the vehicle has four wheels, obtain 4), steering angle θ steering and the steering angular velocity θ′ steering Furthermore, the generator 111 outputs an estimated value of the steering wheel torque by calculating the input data.

[0158] Next, the processor 200 inputs the measured value related data including the state data, the operation data, and the measured value of the physical quantity to the discriminator 112, and outputs a discriminant value for the measured value related data generated by the discriminator 112 (step S122).

[0159] At the same time, the processor 200 inputs the estimated value related data including the state data, the operation data and the estimated value of the physical quantity to the discriminator 112, and outputs the discriminant value for the estimated value related data generated by the discriminator 112 (step S123).

[0160] In the step of outputting the discriminant value for the measured value related data (step S122) and the step of outputting the discriminant value for the estimated value related data (step S123), the discriminator 112 receives the measured value related data including the state data, the operation data and the measured value, and generates and outputs the discriminant value for the measured value related data. As described later, the discriminant value for the measured value related data generated by the discriminator 112 can constitute the error data input to the abnormality detection model 120.

[0161] In addition, the discriminator 112 may additionally receive estimated value-related data including the state data, the operation data, and the estimated value, and generate and output a discriminant value for the estimated value-related data. The discriminant value for the estimated value-related data generated by the discriminator 112 may be applied to the training of the generator 111 of the artificial neural network model 110.

[0162] The step of outputting the discrimination value for the measured value related data (step S122) and the step of outputting the discrimination value for the estimated value related data (step S123) may be performed simultaneously. In addition, either the step of outputting the discrimination value for the measured value related data (step S122) or the step of outputting the discrimination value for the estimated value related data (step S123) may be performed first, and the other may be performed last.

[0163] In the motor abnormality detection method according to an embodiment of the present invention, the artificial neural network model 110 may include a generative adversarial network. In this case, the generative adversarial network is constructed by alternately learning the generator 111 and the discriminator 112, and the state data, the operation data and the measured value used in the learning may be values ​​obtained when the vehicle and the motor are in a normal state.

[0164] In other words, the artificial neural network model 110 may be a virtual twin model in the EPS system of the vehicle. In more detail, the artificial neural network model 1110 may be a deep learning-based neural twin model for the EPS system.

[0165] The artificial neural network model 110 estimates the steering torque by receiving 9 types of data including the state of the steering wheel and the vehicle. At this time, the steering torque estimated by the artificial neural network model 110 can reflect the normal working condition.

[0166] Finally, after the step of outputting the estimated value of the physical quantity (step S120), the processor 200 compares the estimated value with the measured value to detect whether the motor is abnormal (step S130).

[0167] Fig.10 Detailed flowchart of the step of detecting whether a motor is abnormal in a motor abnormality detection method according to an embodiment of the present invention.

[0168] Reference Fig.10 , the step of detecting whether the motor is abnormal or not can be performed as follows.

[0169] First, the processor 200 calculates error data related to the difference between the estimated value and the measured value (step S131).

[0170] As described above, there are multiple data sets, the data sets include the status data, the operation data, the measured values ​​and the estimated values, and the error data may include the average error and standard error of the errors between the measured values ​​and the estimated values ​​obtained from each of the multiple data sets, the maximum absolute error between the measured values ​​and the estimated values ​​of the multiple data sets, and the discrimination value of the discriminator 112 for the measured value related data in the multiple data sets.

[0171] A plurality of data sets including the state data, the operation data, the measured values, and the estimated values ​​may be considered as a batch. For example, 100 to 150 data sets (as a specific example, 128 data sets) may be considered as a batch, and the error data may be obtained for each batch.

[0172] Next, the processor 200 inputs the error data into the anomaly detection model 120 and performs calculations (step S132). At this time, the anomaly detection model 120 may include a one-class support vector machine (OCSVM) algorithm. In other words, the processor 200 may input the error data into the OCSVM (One-Class Support Vector Machine) algorithm and perform calculations.

[0173] In more detail, the error data obtained for one batch may be input to the abnormality detection model 120. In addition, the abnormality level of the motor (performance degradation of the motor) may be detected based on the output (feature) obtained from the abnormality detection model 120 that received the error data.

[0174] At the same time, the present invention additionally provides a non-transitory computer-readable storage medium storing a program for executing the working method of the motor abnormality detection method. Specifically, the present invention may provide a non-transitory computer-readable storage medium storing a program, wherein the program includes at least one instruction for executing the working method of the motor abnormality detection method according to an embodiment of the present invention.

[0175] At this time, the instruction includes not only a machine code generated by a compiler but also a high-level language code executable by a computer.

[0176] The recording medium may include a hardware device configured to store and execute program instructions such as magnetic media (Magnetic Media) such as hard disks, floppy disks and magnetic tapes, optical recording media (Optical Media) such as compact disk read only memory (CD-ROM), high-density digital video disk (Digital Video Disk, DVD), magneto-optical media (Magneto-Optical Media) such as floppy disks (Floptical Disk), ROM, RAM, flash memory, etc.

[0177] It should be understood that the effects of the present invention are not limited to the above-mentioned effects, and include all effects inferred from the configuration of the present invention described in the detailed description of the present invention or the claims.

[0178] Although the embodiments of the present invention have been described, the spirit of the present invention is not limited to the embodiments presented in the specification. Those skilled in the art who understand the spirit of the present invention will be able to easily propose other embodiments within the scope of the same spirit by adding, changing, deleting or adding components, but this will also be included in the spirit of the present invention.

Claims

1. A motor abnormality detection device for detecting abnormality of a motor generating a steering assist torque in an electric power steering system of a vehicle, characterized in that: include: Memory, and processor; the memory stores an artificial neural network model that receives state data related to a state of the vehicle, operation data related to a steering operation of a driver of the vehicle, and estimates a physical quantity related to an output of the motor, The processor inputs the state data and the operation data into the artificial neural network model and performs calculation to output an estimated value of the physical quantity, and the processor detects whether the motor is abnormal by comparing the estimated value with an actual measured value of the physical quantity.

2. The motor abnormality detection device according to claim 1, characterized in that: The status data and the operation data are constituted by signals that can be acquired via a controller area network of the vehicle.

3. The motor abnormality detection device according to claim 1, characterized in that: The physical quantity is a steering torque generated when the driver operates a steering wheel of the vehicle.

4. The motor abnormality detection device according to claim 1, characterized in that: The status data includes a longitudinal velocity of the vehicle, a lateral acceleration of the vehicle, a yaw rate of the vehicle, and a wheel speed of each wheel of the vehicle.

5. The motor abnormality detection device according to claim 1, characterized in that: The operation data includes a steering angle and a steering angular velocity.

6. The motor abnormality detection device according to claim 1, characterized in that: The artificial neural network model comprises: a generator that receives the state data and the operation data and generates the estimated value; and A discriminator receives measured value related data including the state data, the operation data, and the measured value, and outputs a discriminant value for the measured value related data.

7. The motor abnormality detection device according to claim 6, characterized in that: The discriminator also receives estimated value-related data including the state data, the operation data, and the estimated value, and also outputs a discriminant value for the estimated value-related data.

8. The motor abnormality detection device according to claim 7, characterized in that: The artificial neural network model is constructed by alternately learning the generator and the discriminator. The state data, the operation data, and the actual measurement values ​​used in the learning are acquired when the vehicle and the motor are in a normal state.

9. The motor abnormality detection device according to claim 8, characterized in that: The estimated value tracks an actual measured value of the physical quantity acquired when the vehicle and the motor are in a normal state.

10. The motor abnormality detection device according to claim 6, characterized in that: The memory also stores an anomaly detection model, The processor inputs error data related to the difference between the estimated value and the measured value into the abnormality detection model and determines whether the motor has an abnormality.

11. The motor abnormality detection device according to claim 10, characterized in that: There are a plurality of data sets, the data sets including the state data, the operation data, the measured values, and the estimated values, The error data includes an average error and a standard error of errors between measured values ​​and estimated values ​​obtained from each of the plurality of data sets, a maximum absolute error between the measured values ​​and estimated values ​​of the plurality of data sets, and a discriminant value of the discriminator for the measured value-related data of the plurality of data sets.

12. A motor abnormality detection method for detecting abnormality of a motor generating steering assist torque in an electric power steering system of a vehicle, characterized in that: include: The processor receives state data related to the state of the vehicle, operation data related to the steering operation of the driver of the vehicle, and measured values ​​of physical quantities related to the output of the motor; The processor inputs the state data and the operation data into an artificial neural network model and performs calculations to output an estimated value of the physical quantity; as well as The processor detects whether the motor has an abnormality by comparing the estimated value with the actual measurement value.

13. The motor abnormality detection method according to claim 12, characterized in that: The physical quantity is a steering torque generated when the driver operates a steering wheel of the vehicle.

14. The motor abnormality detection method according to claim 12, characterized in that: The status data includes a longitudinal velocity of the vehicle, a lateral acceleration of the vehicle, a yaw rate of the vehicle, and a wheel speed of each wheel of the vehicle.

15. The motor abnormality detection method according to claim 12, characterized in that: The operation data includes a steering angle and a steering angular velocity.

16. The motor abnormality detection method according to claim 12, characterized in that: The artificial neural network model includes a generator, which receives the state data and the operation data and generates the estimated value. The step of outputting the estimated value of the physical quantity includes the steps of the processor inputting the state data and the operation data into the generator, and outputting the estimated value generated by the generator.

17. The motor abnormality detection method according to claim 16, characterized in that: The artificial neural network model further includes a discriminator, which receives measured value related data including the state data, the operation data and the measured value, and outputs a discriminant value for the measured value related data. The step of outputting the estimated value of the physical quantity further includes the step of the processor inputting measured value related data including the state data, the operation data and the measured value of the physical quantity into the discriminator, and outputting a discriminant value of the discriminator for the measured value related data.

18. The motor abnormality detection method according to claim 17, characterized in that: The step of detecting whether the motor is abnormal comprises: The processor calculates error data associated with a difference between the estimated value and the measured value; and The processor inputs the error data into a one-class support vector machine algorithm and performs calculations.

19. The motor abnormality detection method according to claim 18, characterized in that: There are a plurality of data sets, the data sets including the state data, the operation data, the estimated values, and the measured values, The error data includes an average error and a standard error of errors between measured values ​​and estimated values ​​obtained from each of the plurality of data sets, a maximum absolute error between the measured values ​​and estimated values ​​of the plurality of data sets, and a discriminant value of the discriminator for the measured value-related data of the plurality of data sets.

20. A non-transitory computer-readable storage medium storing a program, wherein: The program includes at least one instruction for executing the motor abnormality detection method according to claim 12 .

Images