DEVICE AND METHOD FOR DETECTING ANOMALIES IN AN ENGINE, AND COMPUTER-READABLE STORAGE MEDIUM FOR STORED A PROGRAM FOR EXECUTING THE METHOD

Abstract

Device for detecting an anomaly (1) of a motor (18) configured to generate a steering assistance torque in an electric power steering system (EPS) (10) of a vehicle, the device comprising: a memory (100) configured to store an artificial neural network model (110) configured to receive state data relating to a state of the vehicle and operational data relating to a steering action by the driver of the vehicle and to estimate a physical quantity relating to an output of the motor (18) using the state data and the operational data, and a processor (200) configured to input the state data relating to the state of the vehicle and the operational data relating to the steering action of the driver into the artificial neural network model (110), an estimated value of the physical quantity,which relates to the output of the motor (18), based on the state data relating to the state of the vehicle and the operating data relating to the driver's steering action, using the artificial neural network model (110), and to detect the anomaly of the motor (18) by comparing the estimated value of the physical quantity relating to the output of the motor (18) with an actual measured value of the physical quantity relating to the output of the motor (18).

Description

CROSS-REFERENCE TO A RELATED APPLICATION

[0001] This application claims priority to and the benefit of Korean Patent Application No. 10-2023-0104199, filed on August 9, 2023, the disclosure of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0002] The present disclosure generally relates to an engine abnormality detection apparatus and method (hereinafter referred to as "engine abnormality detection apparatus and method") and a non-transitory computer-readable storage medium storing a program for performing the engine abnormality detection method, and more particularly, to an engine abnormality detection apparatus and method for detecting an abnormality of a motor configured to generate steering assist torque in an electric power steering (EPS) system of a vehicle, and a non-transitory computer-readable storage medium storing a program for performing the engine abnormality detection method. BACKGROUND

[0003] An electric power steering (EPS) system is a device that uses a motor to assist the steering of a vehicle. The EPS system analyzes the collected data regarding steering wheel input, vehicle speed, etc., and controls the motor to provide the required or appropriate steering assist torque. Unlike conventional hydraulic power steering systems, the EPS system can be used in electric vehicles and is more efficient in terms of space and fuel consumption.

[0004] However, a malfunction of the EPS system while driving, such as loss of assist (LoA), can have a significant impact on the vehicle's steering behavior and lead to a fatal accident. The malfunction of the motor that generates the steering assist torque is the most important factor in the LoA. Therefore, to ensure vehicle safety, the motor that generates the steering torque in the EPS system must be protected against EPS system failure.

[0005] However, the conventional approach to motor resilience in EPS systems focuses on follow-up measures in the event of a failure rather than preventing failure in advance. Meanwhile, some research and development is being conducted on early detection of defects in EPS motors, but these primarily use model-based approaches to simultaneously estimate the system's health and defects. DEMOLITION

[0006] The present disclosure is intended to solve the above-mentioned problems, and the present disclosure provides an engine abnormality detection apparatus and method, and a non-transitory computer-readable storage medium storing a program for performing the engine abnormality detection method, thereby predicting a malfunction of the engine function before an engine failure of an EPS system occurs.

[0007] Another object of the present disclosure is to provide an apparatus and method for detecting engine abnormalities and a non-transitory computer-readable storage medium storing a program for performing the engine abnormality detection method, thereby predicting a decrease in the output of an engine based on an artificial intelligence-based digital twin algorithm before a loss of assist (LoA) or a serious decrease in the output of the engine torque occurs in an EPS system.

[0008] Another object of the present disclosure is to provide an apparatus and method for detecting engine abnormalities, and a non-transitory computer-readable storage medium storing a program for performing the method for detecting engine abnormalities, thereby predicting a decrease in the output of the engine of an EPS system from signals obtained via a controller area network (CAN) of a vehicle.

[0009] The objects of the present disclosure are not limited to the objects described above, and other objects not mentioned will become apparent to one skilled in the art to which the present disclosure belongs from the following description.

[0010] According to one aspect of the present disclosure, there is provided 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, the device including a memory and a processor, the memory storing an artificial neural network model that receives state data related to a state of the vehicle and operation data related to a steering operation of the driver of the vehicle and estimates a physical quantity related to an output of the motor, and the processor inputs the state data and the operation data to the artificial neural network model, performs calculations to output an estimated value of the physical quantity, and compares the estimated value with an actual measured value of the physical quantity to detect whether the motor is abnormal.

[0011] In the engine abnormality detection device according to one aspect of the present disclosure, the state data and the operation data may be composed of signals that can be obtained via a controller area network (CAN) of the vehicle.

[0012] In the engine abnormality detection device according to one aspect of the present disclosure, the physical quantity may be a steering torque generated when the driver operates a steering wheel of the vehicle.

[0013] In the engine abnormality detection device according to one aspect of the present disclosure, the state data may include a longitudinal speed of the vehicle, a lateral acceleration of the vehicle, a yaw rate of the vehicle, and a wheel speed for each wheel of the vehicle.

[0014] In the engine abnormality detection device according to one aspect of the present disclosure, the operation data may include a steering angle and a steering angular velocity.

[0015] In the motor abnormality detection device according to one aspect of the present disclosure, the artificial neural network model may include a generative adversarial network (GAN).

[0016] In the engine abnormality detection device according to one aspect of the present disclosure, the artificial neural network model may include a generator that receives the state data and the operation data and generates the estimated value, and a discriminator that receives the actual measurement-related data including the state data, the operation data, and the actual measurement value and outputs a discrimination value for the actual measurement-related data.

[0017] In the engine abnormality detection device according to one aspect of the present disclosure, the generator may be configured by a multivariate transformer.

[0018] In the engine abnormality detection device according to one aspect of the present disclosure, the discriminator may additionally receive estimated value-related data including the state data, the operation data, and the estimated value, and additionally output a discrimination value for the estimated value-related data.

[0019] In the engine abnormality detection device according to one aspect of the present disclosure, the artificial neural network model is constructed by alternately learning the generator and the discriminator, and the state data, the operation data, and the actual measurement value used during the learning can be obtained in a state where the vehicle and the engine are normal.

[0020] In the engine abnormality detection device according to one aspect of the present disclosure, the estimated value may follow the actual measured value of the physical quantity obtained in a state where the vehicle and the engine are normal.

[0021] In the engine abnormality detection device according to one aspect of the present disclosure, the memory may further store an abnormality detection model, and the processor may determine whether the engine is abnormal by inputting error data related to a difference between the estimated value and the actual measured value to the abnormality detection model.

[0022] In the engine abnormality detection device according to one aspect of the present disclosure, the abnormality detection model may include a one-class support vector machine (OCSVM) algorithm.

[0023] In the engine abnormality detection device according to one aspect of the present disclosure, there may be a plurality of data sets including the state data, the operation data, the actual measured value, and the estimated value, and the error data includes a mean value and a standard error of the error between the actual measured value and the estimated value obtained from each of the plurality of data sets, a maximum absolute error between the actual measured value and the estimated value of the plurality of data sets, and the discrimination value of the actual measured value-related data discriminator of the plurality of data sets.

[0024] According to another aspect of the present disclosure, there is provided a motor abnormality detection method for detecting an abnormality of a motor that generates steering assist torque in an electric power steering (EPS) system of a vehicle, the method including receiving, with a processor, state data related to a state of the vehicle, operation data related to a steering operation of a driver of the vehicle, and an actual measurement value of a physical quantity related to an output of the motor, inputting the state data and the operation data into an artificial neural network model with the processor, performing calculations and outputting an estimated value of the physical quantity, and comparing the estimated value with the actual measurement value by the processor to detect whether the motor is abnormal.

[0025] In the engine abnormality detection method according to another aspect of the present disclosure, the condition data and the operating data may be configured with signals obtainable via a controller area network (CAN) of the vehicle.

[0026] In the engine abnormality detection method according to another aspect of the present disclosure, the physical quantity may be a steering torque generated when the driver operates a steering wheel of the vehicle.

[0027] In the engine abnormality detection method according to another aspect of the present disclosure, the condition data may include a longitudinal speed 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.

[0028] In the method for detecting engine abnormalities according to another aspect of the present disclosure, the operating data may include a steering angle and a steering angular velocity.

[0029] In the method for detecting motor abnormalities according to another aspect of the present disclosure, the artificial neural network model may comprise a generative adversarial network (GAN).

[0030] In the engine abnormality detection method according to another aspect of the present disclosure, the artificial neural network model may include a generator that receives the state data and the operation data and generates the estimated value, and wherein outputting the estimated value of the physical quantity may include inputting the state data and the operation data to the generator with the processor and outputting the estimated value generated by the generator.

[0031] In the engine abnormality detection method according to another aspect of the present disclosure, the artificial neural network model may further include a discriminator that receives actual measurement-related data, including the state data, the operation data, and the actual measurement value, and outputs a discrimination value for the actual measurement-related data; and wherein outputting the estimated value of the physical quantity may include inputting, with the processor, data related to the actual measurement value, including the state data, the operation data, and the actual measurement value of the physical quantity, to the discriminator and outputting a discrimination value for the actual measurement-related data generated by the discriminator.

[0032] In the engine abnormality detection method according to another aspect of the present disclosure, detecting whether the engine is abnormal may include calculating error data related to a difference between the estimated value and the actual measured value with the processor, and inputting the error data to a one-class support vector machine (OCSVM) algorithm and performing calculations with the processor.

[0033] In the engine abnormality detection method according to another aspect of the present disclosure, there may be a plurality of data sets including the state data, the operation data, the actual measured value, and the estimated value, and the error data may include a mean value and a standard error of the error between the actual measured value and the estimated value obtained from each of the plurality of data sets, a maximum absolute error between the actual measured value and the estimated value of the plurality of data sets, and the discrimination value of the actual measured value-related data discriminator of the plurality of data sets.

[0034] In the engine abnormality detection method according to another aspect of the present disclosure, the artificial neural network model is constructed by alternately learning the generator and the discriminator, and the state data, the operation data, and the actual measurement value used during the learning can be obtained in a state where the vehicle and the engine are normal.

[0035] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium having a program stored thereon including at least one instruction for performing the method for detecting engine abnormalities. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The above and other objects, features and advantages of the present disclosure will become more apparent to those skilled in the art by describing exemplary embodiments thereof with reference to the accompanying drawings. Fig. 1 is a block diagram illustrating the configuration of an engine abnormality detecting apparatus according to an embodiment of the present disclosure. Fig. 2 is a conceptual diagram schematically showing an electric power steering (EPS) system of a vehicle and state data and operational data according to an embodiment of the present disclosure. Fig. 3 is a block diagram showing the models stored in a memory of an engine abnormality detecting device according to an embodiment of the present disclosure. Fig. 4 is a block diagram showing the detailed configuration of an artificial neural network model as shown in Fig. 3, according to one embodiment of the present disclosure. Fig. 5 is a block diagram showing the detailed configuration of a generator of the artificial neural network model of Fig. 4 according to an embodiment of the present disclosure. Fig. 6 is a conceptual diagram showing the operation of an engine abnormality detecting device according to an embodiment of the present disclosure. Fig. 7 is a table showing the technical effects of an engine abnormality detection device according to an embodiment of the present disclosure. Fig. 8 is a flowchart of a method for detecting engine abnormalities according to an embodiment of the present disclosure. Fig. 9 is a flowchart illustrating a step of outputting an estimated value of a physical quantity in a method for detecting engine abnormalities according to an embodiment of the present disclosure. Fig. 10 is a flowchart illustrating a step of detecting whether an engine is abnormal in an engine abnormality detection method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0037] In the following, embodiments of the present disclosure will be described in detail so that a person skilled in the art to which the present disclosure belongs can easily practice the embodiments. The present disclosure may be implemented in many different forms and is not limited to the embodiments described herein. In order to clearly describe the present disclosure, parts that are not related to the description are omitted from the accompanying drawings, and the same or similar components are denoted by the same reference numerals throughout the specification.

[0038] The words and terms used in the specification and claims are not to be construed as having their ordinary or dictionary meanings and should be construed as having meanings and concepts consistent with the technical spirit of the present disclosure, following the principle that inventors may define terms and concepts to best describe their invention.

[0039] Note that in the description, terms such as "comprising" or "having" are intended to specify the presence of features, a respective number, steps, operations, components, parts, or combinations thereof described in the description, but are not intended to exclude the possibility of the presence or addition of one or more other features, a respective number, one or more other steps, operations, components, parts, or combinations thereof.

[0040] Fig. 1 is a block diagram illustrating the configuration of an engine abnormality detection device (hereinafter referred to as “engine abnormality detection device”) according to an embodiment of the present disclosure.

[0041] A motor abnormality detection device 1 according to an embodiment of the present disclosure may be configured to detect an abnormality of a motor configured to generate steering assist torque in an electric power steering (EPS) system of a vehicle. Specifically, the motor abnormality detection device 1 according to an embodiment of the present disclosure may predict the malfunction of the motor before the occurrence of an engine failure through an artificial intelligence-based digital twin algorithm for a vehicle. Specifically, the digital twin algorithm may virtually construct the EPS system of the vehicle.

[0042] According to one embodiment of the present disclosure, the engine abnormality detection device 1 detects the engine malfunction before a loss of assist (LoA) of the vehicle's EPS system caused by the engine failure occurs. Specifically, the engine abnormality detection device 1 according to one embodiment of the present disclosure can sensitively detect even minor performance deterioration or deterioration of the engine.

[0043] Accordingly, according to one embodiment of the present disclosure, it is possible to predict in advance before the failure of the EPS system motor occurs. This makes it possible to take precautions before a motor malfunction occurs and prevent motor failure.

[0044] Referring to Fig. 1, the engine abnormality detection device 1 according to an embodiment of the present disclosure may include a memory 100 and a processor 200.

[0045] The memory 100 can store data, a model, and the like required to detect an engine abnormality. The memory 100 can, for example, include a device configured to store software or program instructions. The element 100 can, for example, consist of at least one of the following elements: ROM, RAM, and flash memory. In addition, the memory 100 can include magnetic media such as floppy disks and magnetic tapes, optical media such as compact disk read-only memory (CD-ROM), digital video disks (DVD), and magneto-optical media such as floppy disks.

[0046] Processor 200 may be configured to input the data into the model and perform calculations. For example, processor 200 may be a hardware unit capable of performing calculations and controls in a computer. Processor 200 may include at least one arithmetic logic unit (ALU) and one register.

[0047] The memory 100 stores an artificial neural network model. The artificial neural network model receives state data related to a state of the vehicle and operational data related to a driver's steering action, and estimates a physical quantity indicative of or related to an output of the engine.

[0048] The processor 200 inputs the state data and the operation data into the artificial neural network model stored in the memory 100, calculates an estimated value of the physical quantity indicative of or associated with the output of the motor, and compares the estimated value with an actual measured value of the physical quantity to detect whether the motor is abnormal.

[0049] In conjunction with detecting whether the engine is abnormal, the memory 100 may also store an abnormality detection model. Furthermore, the processor 200 may determine whether the engine is abnormal by inputting error data related to a difference between the estimated value and the actual measured value into the abnormality detection model.

[0050] Fig. 2 is a conceptual diagram schematically showing an electric power steering (EPS) system of a vehicle and state data and operational data according to an embodiment of the present disclosure. Fig. 2 shows in particular an EPS system 10 having an engine 18, the abnormality of which is to be detected by an engine abnormality detection device 1 according to an embodiment of the present disclosure, as well as state data (DVehicle) and operating data (DLeering).

[0051] As in Fig. 2, the vehicle's EPS system 10 may include a steering wheel 11, a steering shaft 12, a steering angle sensor 13, a steering torque sensor 14, and a vehicle condition measurement module 15. The steering wheel 11 may be operated by a driver. The steering shaft 12 is connected to the steering wheel 11 to transmit the steering force received by the steering wheel 11 to the wheel of the vehicle. The steering angle sensor 13 is disposed on, near, or in sensing relationship with the steering shaft 12 to measure the steering angle and angular velocity of the steering wheel 11. The steering torque sensor 14 may be disposed on, near, or in sensing relationship with the steering shaft 12 to measure the steering torque of the steering wheel 11. The vehicle condition measurement module 15 may include at least one sensor for measuring the condition of the vehicle.

[0052] The vehicle condition measuring module 15 can, for example, measure the longitudinal speed (V x ), the lateral acceleration (ay ), the yaw rate (θ' z ) and the wheel speed (V Rad ) of each of the vehicle's multiple wheels. Since the vehicle generally has four wheels, the vehicle condition measurement module 15 can generate four wheel speed (VRad) data for each wheel. In other words, the vehicle condition measurement module 15 can measure a total of nine types of data about the vehicle's condition.

[0053] The steering angle (θSteering) and the steering angular velocity (θ'Steering) measured by the steering angle sensor 13, the steering torque (TLeering) measured by the steering torque sensor 14, the vehicle speed (Vx), the lateral acceleration (ay), the yaw rate (θ' z ) and the wheel speed (VRad) of each of the plurality of wheels of the vehicle measured by the vehicle condition measuring module 15 are transmitted to an electronic control unit (ECU) 17 via a Controller Area Network (CAN) bus 16.

[0054] The electronic control unit 17 receives the data transmitted via the CAN bus 16, generates a command to control the motor 18, which generates a steering assist torque by processing or calculating the received data, and transmits the command to the motor 18 to control it. In this case, the command can be given to the motor 18 in the form of a current or in the form of a current supply. The electronic control unit 17 can perform feedback control for the motor 18.

[0055] A torque input by the driver via the steering wheel 11 and a torque generated by the motor 18 are transmitted to a steering mechanism 19 connected to one or more wheels of the vehicle. The direction of the vehicle's wheels can be changed or adjusted by the steering mechanism 19, which is driven by the torque input via the steering wheel 11 and the torque generated by the motor 18, thereby steering the vehicle.

[0056] As described above, the memory 100 stores an artificial neural network model configured to receive state data related to a state of the vehicle and operation data related to a steering operation or manipulation of the driver of the vehicle, and estimate a physical quantity related to an output of the engine based on the state data and the operation data. The processor 200 inputs the state data and the operation data to the artificial neural network model, performs calculations or processes the state data and the operation data to output an estimated value of the physical quantity related to the output of the engine, and compares the estimated value with an actual measured value of the physical quantity related to the output of the engine to detect whether the engine is abnormal.In connection with the detection of whether the engine is abnormal, the memory 100 may also store an abnormality detection model.

[0057] In one embodiment of the present disclosure, the state data may include, for example, but not exclusively, the longitudinal velocity (Vx) of the vehicle, the lateral acceleration (ay), the yaw rate (θ' z ) and the wheel speed (VRad) of each of the vehicle's multiple wheels. Generally, if the vehicle has four wheels, the status data can include seven types of data.

[0058] In addition, the operating data can include the steering angle (θSteering) and the steering angular velocity (θ'Steering). Thus, the operating data can include two types of data.

[0059] As described above, in the engine abnormality detection device 1 according to an embodiment of the present disclosure, the state data and the operation data may be composed of signals obtained via the CAN of the vehicle.

[0060] The physical quantity related to the output of the motor may be a steering wheel torque (Tsteering) generated when the driver of the vehicle operates the steering wheel 11. In other words, in one embodiment of the present disclosure, the processor 200 may output the estimated value of the steering wheel torque through calculations and compare the estimated value with the actual measured value of the steering wheel torque to determine whether the motor 18 is abnormal. The actual measured value may, for example, be the steering torque (T Lenkung ) be.

[0061] The estimated steering torque can be calculated, for example, but not exclusively, using a Dugoff tire model. Since the Dugoff tire model is a well-known model in the field of automotive engineering, a detailed description will be omitted.

[0062] Fig. 3 is a block diagram showing the models stored in a memory of an engine abnormality detection device according to an embodiment of the present disclosure. As shown in Fig. 3, the memory 100 can store an artificial neural network model 110 and an anomaly detection model 120.

[0063] The following describes in detail the artificial neural network model 110 configured to receive the state data and the operating data and output the estimated value of the steering wheel torque, the anomaly detection model 120 configured to detect whether the motor 18 is abnormal based on a comparison between the estimated value of the steering wheel torque and the actual measured value of the steering wheel torque, and the calculations of the processor 200 using the models stored in the memory 100.

[0064] The artificial neural network model 110 receives the state data related to the state of the vehicle and the operational data related to a steering operation or manipulation by the vehicle driver and outputs the estimated steering torque. In one embodiment of the present disclosure, the artificial neural network model 110 may include a generative adversarial network (GAN).

[0065] The generative adversarial network is constructed by alternately learning a generator and a discriminator. The state data related to the state of the vehicle, the operational data related to a steering operation or manipulation of the vehicle driver, and the actual measured value of the steering wheel torque used during learning can be obtained in a state where the vehicle and the engine 18 are normal. Accordingly, the estimated value of the steering wheel torque can be equal to or substantially identical to the actual measured value of the physical quantity of the steering wheel torque obtained in a state where the vehicle and the engine 18 are normal.

[0066] In other words, the artificial neural network model 110 may be a virtual twin model that is virtually duplicated with the vehicle's EPS system. More specifically, the artificial neural network model 110 may be a deep learning-based neural twin model for the EPS system.

[0067] In one embodiment of the present disclosure, the artificial neural network model 110 receives nine types of data, including the state of the steering wheel and the vehicle, and estimates the steering torque. In this case, the steering torque estimated by the artificial neural network model 110 may reflect a normal operating state.

[0068] Fig. 4 is a block diagram showing the detailed configuration of an artificial neural network model as shown in Fig. 3, according to an embodiment of the present disclosure. Referring to Fig. 4, the artificial neural network model 110 may include a generator 111 and a discriminator 112.

[0069] The generator 111 receives the state data relating to the state of the vehicle and the operational data relating to a steering action or manipulation by the driver of the vehicle and generates the estimated steering wheel torque. The generator 111 may include a neural network. In one embodiment of the present disclosure, the generator 111 may have a virtual duplicate of the virtual twin model with the vehicle's EPS system.

[0070] The generator 111 may, for example, comprise a multivariate transformer. In other words, the artificial neural network model 110 may be a GAN based on the multivariate transformer.

[0071] In one embodiment of the present disclosure, generator 111 may receive the longitudinal velocity (Vx) of the vehicle, the lateral acceleration (ay), the yaw rate (θ'z), the wheel speed (Vwheel) of each of the vehicle's multiple wheels (e.g., four wheel speeds are obtained if the vehicle has four wheels), the steering angle (θsteer), and the steering angular velocity (θ'steer). Furthermore, generator 111 calculates or processes the input data and outputs the estimated steering wheel torque.

[0072] Fig. 5 is a block diagram showing the detailed configuration of a generator of the artificial neural network model of Fig. 4 according to an embodiment of the present disclosure. As shown in Fig. 5, the generator 111 may include an encoder 111a and a decoder 111b.

[0073] Encoder 111a and decoder 111b may form a steering torque estimation network. In other words, generator 111 may consist of a structural encoder-decoder network that is continuously trained to estimate steering torque.

[0074] Encoder 111a may be configured to extract high-quality features from the input data using a transformer block and a linear operation. The transformer block may include, for example, an embedding layer, a self-attention layer, and a feedforward layer.

[0075] First, the embedding layer in the transformer block can encode spatio-temporal information into input sequences using a trainable embedding matrix.

[0076] Next, deep features can be extracted by the self-attention layer of the transformer block, and refinement can be performed with the feedforward layer of the transformer block.

[0077] At this point, the self-attention layer can use a scaled dot-product attention mechanism. The self-attention layer is perhaps the most important component in the transformer-based deep neural network (DNN) architecture, allowing the model to process and understand dependencies in the input sequence.

[0078] Scaled Dot-Product Attention can first apply level normalization to the input sequence and process the results through a linear operation into three unique vectors: query (Q), key (K), and value (V).

[0079] An attention score (AS) can then be calculated for each position of the input sequence based on the derived vectors. The formula for calculating the attention score (AS) can be represented as equation (1) as follows: AS=σ(Q⋅K‖Xinput‖dim)⋅V where σ(-) are softmax calculations, ||x input || dim is a feature dimension of an input and (-) means the inner product.

[0080] The attention score (AS) quantifies the relative importance of each position in the input sequence associated with a particular query. This means that higher scores are assigned to more appropriate positions.

[0081] The output of the transformer block can be fed to two different linear calculations to calculate the queries, key vectors and initial input values ​​(ŷ1) of the decoder 111b.

[0082] The decoder 111b may estimate the steering torque from the obtained values ​​in an autoregressive method similar to recurrent neural networks (RNNs) that operate from sequence to sequence.

[0083] Similar to encoder 111a, decoder 111b may include the transformer block and a linear operation. Furthermore, decoder 111b may include a cross-attention layer in the transformer block.

[0084] The cross-attention layer may perform scaled dot-product attention using the queries and key vectors of the encoder 111a to learn dynamic correlations between the input signal and the estimated signal, as mentioned above.

[0085] The discriminator 112 receives data related to the actual measured value, including state data related to the state of the vehicle, operational data related to a steering operation or manipulation by the vehicle driver, and the actual measured value of the steering wheel torque, and outputs a discrimination value for the data related to the actual measured value. Furthermore, the discriminator 112 may additionally receive estimated value-related data, including state data related to the state of the vehicle, operational data related to a steering operation or manipulation by the vehicle driver, and the estimated value of the steering wheel torque, and may additionally output a discrimination value for the estimated value-related data.

[0086] For example, discriminator 112 may be configured as a discriminator network used for adversarial training of GANs. In other words, generator 111 and discriminator 112 may perform training using the adversarial training method to improve estimation performance and establish the GAN model. For example, discriminator 112 may alternately perform optimization with generator 111 to solve the Wasserstein minimum-maximum problem using equation (2). minGmaxDExD,Real[D(xD,Real)]−ExD,Fake[D(xD,Fake)] where x D,Real represents an actual data set consisting of input data and an actual measured value of a steering torque, and x D,Fake represents a virtual data sample containing input data and the steering torque estimate.

[0087] In the artificial neural network model 110, the generator 111 is trained to simulate the discriminator 112, which uses the spatial-temporal features of x D,Real and x D,Fake Accordingly, the generator 111 can generate the estimated value of the steering wheel torque that follows the actual measured value of the steering torque in the EPS system of the vehicle in a steady state.

[0088] In this context, the loss function for GAN training can be defined as in the following equation (3). LG=1N∑|yG−y^G|+0.001×1N∑[−D(xD,Fake)],LD=1N∑[D(xD,Fake)−D(xD,Real)]+10×1N∑[(‖∇xD,FakeD(xD,Fake)‖2−1)2] where N is the stack size, L G the loss function of a generator and L D is the loss function of a discriminator.

[0089] The first point of the loss of generator 111 corresponds to the mean absolute error loss during map training (LMAE). The remaining part of the loss of generator 111, together with the loss of discriminator 112, forms a GAN loss function (LGAN).

[0090] Meanwhile, the degree of deterioration of the engine can be predicted based on the estimated value of the physical quantity indicative of or associated with the output of the engine and generated by the generator 111 and the actual measured value of the physical quantity indicative of or associated with the output of the engine and actually measured.

[0091] In one embodiment of the present disclosure, the degree of deterioration of the engine may be predicted by the anomaly detection model 120.

[0092] The anomaly detection model 120 may receive error data related to a difference between the estimated value and the actual measured value, indicating or associated with the output of the engine, and perform calculations or process the received error data. The anomaly detection model 120 may determine whether the engine 18 is abnormal. In one embodiment of the present disclosure, the anomaly detection model 120 may include a one-class support vector machine (OCSVM) algorithm.

[0093] In addition, there are a plurality of data sets, including the state data related to the state of the vehicle, the operation data related to a steering operation or manipulation of the driver of the vehicle, the actual measured value of the steering wheel torque, and the estimated value of the steering wheel torque, and the error data may include a mean value and a standard error between the actual measured value and the estimated value of the steering wheel torque obtained from each of the plurality of data sets, a maximum absolute error between the actual measured value and the estimated value of the plurality of data sets, and the discrimination value of the discriminator 112 for actual measured value-related data of the plurality of data sets.

[0094] In other words, a plurality of data sets, including the status data related to the vehicle's condition, the operational data related to a steering operation or driver manipulation, the actual measured value of the steering wheel torque, and the estimated value of the steering wheel torque, can be treated as a batch. For example, 100 to 150 data sets (e.g., 128 data sets) can be treated or processed as a batch, and the error data can be determined for each batch.

[0095] The fault data obtained for a batch may be input to the anomaly detection model 120. Furthermore, an anomaly level of the engine may be detected based on the output (feature) of the anomaly detection model 120 that receives the fault data.

[0096] Regarding the anomaly detection metric, the F1 score given in equation (4) can be considered. F1=2×TP2×TP+FP+FN where TP stands for true positive, FP for false positive, FN for false negative and positive means deterioration of the engine.

[0097] Meanwhile, the discriminator 112 of the artificial neural network model 110 can calculate not only the discrimination value for the actual measured data of the dataset, but also the discrimination value for the estimated data of the dataset. The discrimination value of the discriminator 112 for the estimated data of the dataset can be fed back to the generator 111 of the artificial neural network model 110 and used for training.

[0098] The configuration of the engine abnormality detection device 1 according to an embodiment of the present disclosure has been described in detail above. The operation of the engine abnormality detection device 1 according to an embodiment of the present disclosure will be described in detail below.

[0099] Fig. 6 is a conceptual diagram showing the operation of an engine abnormality detection device according to an embodiment of the present disclosure. According to Fig. 6, the engine abnormality detection device 1 according to an embodiment of the present disclosure may function as follows.

[0100] First, the memory 100 stores the status data (D Fahrzeug ) relating to the condition of the vehicle, the operating data (D Lenkung ) relating to a steering action or steering manipulation by the driver of the vehicle and the actual measured value (T Lenkung) of the steering torque received from the vehicle's EPS system. As described above, the status data (D Fahrzeug ) the longitudinal speed (V x ) of the vehicle, the lateral acceleration (a y ), the yaw rate (θ' z ) and the wheel speed (V Rad ) each of the vehicle's multiple wheels. In addition, the operating data (D Lenkung ) the steering angles (θ Lenkwinkel ) and steering angle velocities (θ Lenkung ) include.

[0101] Meanwhile, the status data (D Fahrzeug ), the steering data (D Lenkung ) and the actual measured value (TLeering) of the steering torque may be treated or processed as a batch. Furthermore, in one embodiment of the present disclosure, the processing and calculation of data may be performed as a unit of a data batch.

[0102] Next, the processor 200 outputs the status data (D Fahrzeug) relating to the condition of the vehicle and the operating data (D Lenkung ) relating to a steering operation or steering manipulation of the vehicle driver into the artificial neural network model 110, performs the calculation based on the state data (DVehicle) and the operating data (DLeering) to estimate the steering wheel torque, and outputs the estimated value (T̂ steering ) of the steering wheel torque.

[0103] Processor 200 then outputs the error data for each batch of data. As described above, the error data may include a mean and standard error between the actual measurement value and the estimated value obtained from each of the multiple data sets, a maximum absolute error between the actual measurement value and the estimated value of the multiple data sets, and the discrimination value of discriminator 112 for actual measurement-related data of the multiple data sets.

[0104] Finally, the processor 200 inputs the fault data to the anomaly detection model 120 and outputs the result calculated by the anomaly detection model 120. As described above, the anomaly detection model 120 may include, for example, but not exclusively, a one-class support vector machine (OCSVM) algorithm. The anomaly level of the engine may be detected based on the output (feature) of the anomaly detection model 120 receiving the fault data.

[0105] Fig. 7 is a table showing the technical effects of an engine abnormality detection device according to an embodiment of the present disclosure.

[0106] In Fig. 7, the engine deterioration level is predicted by different techniques and models, and an F1 score (anomalies detection metric) is calculated for each technique and model when the vehicle speeds are 20 km / h, 30 km / h and 40 km / h and the engine deterioration levels of the vehicle's EPS system are 10%, 25% and 40%, respectively.

[0107] In Fig. 7, the “feature-based” technique refers to a case where no neural twin structure is used, and the “neural twin structure” refers to a case where a deep learning-based neural twin structure is used.

[0108] Here, the "feature-based" technique includes an empirical cumulative distribution-based outlier detection algorithm (ECOD), a deep one-class classification model (DOCC), a beta-variational auto-encoder model (β-VAE), and generative adversarial active learning (GAAL).

[0109] Furthermore, the “neural twin-based” technique includes a gated recurrent unit (GRU) model, a long short memory (LSTM) model, their attention base versions (A-GRU, A-LSTM), a simple multivariate transformer (MVT) model without GAN training, and a combination (MVT-GAN) of the artificial neural network model 110 and the anomaly detection model 120 stored in the motor anomaly detection device 1 according to an embodiment of the present disclosure.

[0110] Meanwhile, the Fig. The technique and model presented in Figure 7 were applied to data acquired via the CAN bus during in-vehicle experiments. The data can be acquired, for example, at a sampling rate of 100 Hz.

[0111] Fig. Figure 7 can show that the combination (MVT-GAN) of the artificial neural network model 110 and the anomaly detection model 120 stored in the engine anomaly detection device 1 according to an embodiment of the present disclosure has the highest F1 score in most cases. Specifically, according to an exemplary embodiment of the present disclosure, at an F1 score of 0.85 or higher, performance can be reduced by up to 10% compared to the maximum output of the engine.

[0112] In other words, Fig. 7 confirm that the engine abnormality detection device 1 according to an embodiment of the present disclosure can accurately predict the degree of engine deterioration in the vehicle's EPS system before the engine completely fails. Specifically, according to an embodiment of the present disclosure, even a slight decrease in engine output can be accurately predicted.

[0113] In this way, some embodiments of the present disclosure may enable estimation of the system's future operation and remaining useful life (RUL), as well as predictive diagnostics suitable for predictive maintenance (PdM) applications. This allows for effective intervention and maintenance before the vehicle's EPS system motor fails.

[0114] The engine abnormality detection device according to some embodiments of the present disclosure has been described in detail above. A method for detecting an abnormality of an engine according to an embodiment of the present disclosure will be described below.

[0115] Fig. 8 is a flowchart of a method for detecting an abnormality of an engine (hereinafter referred to as “engine abnormality detection method”) according to an embodiment of the present disclosure.

[0116] The motor abnormality detection method (S100) according to an embodiment of the present disclosure is a method for detecting abnormalities of a motor configured to generate steering assist torque in an electric power steering (EPS) system of a vehicle, and may be performed by the motor abnormality detection device 1 according to some embodiments of the present disclosure described above.

[0117] With reference to Fig. 8, the engine abnormality detection method S100 according to an embodiment of the present disclosure may be performed as follows.

[0118] First, the processor 200 receives state data related to the state of the vehicle, operation data related to the steering operation of a vehicle driver, and an actual measurement value of a physical quantity related to the output of the engine (step S110).

[0119] In one embodiment of the present disclosure, the state data may include the longitudinal velocity of the vehicle (Vx), the lateral acceleration (ay), the yaw rate (θ' z ) and the wheel speed (VRad) for each of the vehicle's multiple wheels. As described above, the status data can include seven types of data if the vehicle has four wheels.

[0120] In addition, the operating data may include a steering angle (θSteering) and a steering angular velocity (θ'Steering). This means that the operating data may include two types of data.

[0121] The data listed as examples of status data and operational data can be implemented as signals accessible via the vehicle's Controller Area Network (CAN). If the status data and operational data are implemented as signals accessible via the vehicle's Controller Area Network (CAN), it is not necessary to install a separate sensor in the vehicle to collect data, thus improving data acquisition efficiency.

[0122] When external data transmission is required, the data transmission can be performed efficiently. For example, if the engine abnormality detection device 1 according to an embodiment of the present disclosure is arranged outside the vehicle at a location remote from the vehicle, and the engine abnormality detection method according to an embodiment of the present disclosure can be performed at the location remote from the vehicle, the status data and the operation data can be efficiently received via telecommunications while the vehicle is being driven.

[0123] Meanwhile, the physical quantity may be a steering wheel torque (TLeering) generated when the vehicle driver moves or operates the steering wheel 11. In one embodiment of the present disclosure, the processor 200 may output the estimated value of the steering wheel torque through calculations and compare the estimated value with the actual measured value of the steering wheel torque to determine whether the motor 18 is abnormal. At this time, the actual measured value may be, for example, but not exclusively, the steering torque (TLeering) measured by the steering torque sensor 14.

[0124] In this case, a plurality of data sets, including the state data, the operating data, the actual measured value, and the estimated value, can be processed or handled as a batch. In other words, the processor 200 can receive a plurality of data sets, configure a batch in units of a predetermined number, and perform the calculations described below in batch units.

[0125] Next, the processor 200 inputs the state data and the operating data into the artificial neural network model, performs calculations using the artificial neural network model, and outputs an estimated value of the physical quantity (step S120). As described above, the artificial neural network model 110 may include a generative adversarial network (GAN).

[0126] The artificial neural network model 110 receives the state data and the operating data and outputs the estimated steering torque. In one embodiment of the present disclosure, the artificial neural network model 110 may include the generative adversarial network (GAN). Furthermore, the artificial neural network model 110 may include the generator 111 and the discriminator 112 as described above.

[0127] Fig. 9 is a flowchart illustrating a step of outputting an estimated value of a physical quantity in a method for detecting engine abnormalities according to an embodiment of the present disclosure.

[0128] Referring to Fig. 9, the step S120 of outputting the estimated value of the physical quantity in the engine abnormality detection method (S100) may be performed as follows.

[0129] First, processor 200 inputs the state data related to the vehicle's state and the operational data related to a steering operation or manipulation by the vehicle driver to generator 111 and outputs the estimated value of the steering wheel torque generated by generator 111 (step S121). Here, generator 111 receives the state data and operational data to generate the estimated value of the steering wheel torque (step S121).

[0130] The generator 111 can be configured, for example, as a multivariate transformer. In other words, the artificial neural network model 110 can be a GAN based on a multivariate transformer. In one embodiment of the present disclosure, the generator 111 serves as a virtual twin model that is virtually duplicated with the vehicle's EPS system.

[0131] The generator 111 can receive the vehicle's longitudinal velocity (Vx), lateral acceleration (ay), yaw rate (θ'z), wheel speed (Vwheel) of each of the vehicle's multiple wheels (e.g., four wheel speeds are obtained if the vehicle has four wheels), steering angle (θsteer), and steering angular velocity (θ'steer). Furthermore, the generator 111 calculates the input data and outputs the estimated steering torque.

[0132] Next, the processor 200 inputs actual measurement-related data including the state data, the operation data, and the actual measurement value of the physical quantity of the steering wheel torque to the discriminator 112, and outputs a discrimination value for the actual measurement-related data generated by the discriminator 112 (step S122).

[0133] Meanwhile, the processor 200 inputs estimated value-related data including the state data related to the state of the vehicle, the operation data related to a steering operation or manipulation of the driver, and the estimated value of the physical quantity of the steering wheel torque to the discriminator 112, and outputs a discrimination value for the estimated value-related data generated by the discriminator 112 (step S123).

[0134] In step S122 of outputting the discrimination value for the actual measured value-related data and in step S123 of outputting the discrimination value for the estimated value-related data, the discriminator 112 receives the actual measured value-related data, including the state data, the operating data, and the actual measured value, and outputs the discrimination value for the actual measured value-related data. The discrimination value for the actual measured value generated by the discriminator 112 may represent error data that is input to the anomaly detection model 120, as explained later.

[0135] Furthermore, the discriminator 112 can additionally receive the estimated value-related data, including the state data related to the state of the vehicle, the operational data related to a steering operation or manipulation by the driver, and the estimated value of the steering wheel torque, and output a discriminatory value for the estimated value-related data. The discriminatory value for the estimated value-related data generated by the discriminator 112 can be used for training the generator 111 of the artificial neural network model 110.

[0136] Step S122 of outputting the discrimination value for the actual measurement-related data and step S123 of outputting the discrimination value for the estimated-related data may be performed simultaneously. Alternatively, one of steps S122 of outputting the discrimination value for the actual measurement-related data and S123 of outputting the discrimination value for the estimated-related data may be performed first, and the other may be performed later.

[0137] In the engine abnormality detection method according to an embodiment of the present disclosure, the artificial neural network model 110 may include, for example, but not exclusively, a generative adversary neural network. In this example, the generative adversary network may be constructed by alternately learning the generator 111 and the discriminator 112, and the state data related to the state of the vehicle, the operational data related to a steering operation or manipulation by the driver of the vehicle, and the actual measured value of the steering wheel torque used during learning may be obtained in a state where the vehicle and the engine are normal.

[0138] In other words, the artificial neural network model 110 may be a virtual twin model that is virtually duplicated with the vehicle's EPS system. More specifically, the artificial neural network model 110 may be a deep learning-based neural twin model for the EPS system.

[0139] In this way, the artificial neural network model 110 obtains nine types of data, including the state of the steering wheel and the vehicle, and estimates the steering torque. In this case, the steering torque estimated by the artificial neural network model 110 may reflect a normal operating state.

[0140] After the step S120 of outputting the estimated value of the physical quantity of the steering wheel torque, the processor 200 compares the estimated value with the actual measured value of the steering wheel torque to determine whether the motor is abnormal (step S130).

[0141] Fig. 10 is a flowchart illustrating a step of detecting whether an engine is abnormal in an engine abnormality detection method according to an embodiment of the present disclosure.

[0142] Referring to Fig. 10, step S130 for detecting whether the engine is abnormal in the engine abnormality detection process (S100) may be performed as follows.

[0143] First, the processor 200 calculates error data relating to a difference between the estimated value and the actual measured value of the steering wheel torque (step S131).

[0144] As described above, there are a plurality of data sets including the state data related to the state of the vehicle, the operation data related to a steering operation or manipulation by the driver of the vehicle, the actual measured value of the steering wheel torque, and the estimated value of the steering wheel torque, and the error data may include a mean value and a standard error between the actual measured value and the estimated value of the steering wheel torque from each of the plurality of data sets, a maximum absolute error between the actual measured value and the estimated value of the plurality of data sets, and the discrimination value of the discriminator 112 for actual measured value-related data of the plurality of data sets.

[0145] The multitude of data sets, including status data, operating data, actual measured values, and estimated values, can be treated or processed as a batch. For example, 100 to 150 data sets (e.g., 128 data sets) can be treated or processed as a batch, and the error data can be determined for each batch.

[0146] Next, the processor 200 inputs the error data to the anomaly detection model 120 to perform calculations (step S132). The anomaly detection model 120 may include, for example, a one-class support vector machine (OCSVM) algorithm. In other words, the processor 200 may input the error data to the one-class support vector machine (OCSVM) algorithm to perform calculations.

[0147] More specifically, the fault data obtained for a batch may be input to the anomaly detection model 120. Furthermore, an anomaly level of the engine may be detected based on the output (feature) of the anomaly detection model 120 that receives the fault data.

[0148] Meanwhile, some embodiments of the present disclosure may be implemented in a non-transitory computer-readable storage medium storing a program for performing the engine abnormality detection method. In particular, a non-transitory computer-readable storage medium having a program stored thereon containing at least one instruction for performing the engine abnormality detection method may be provided according to an embodiment of the present disclosure.

[0149] At this point, the instruction may include not only machine code generated by a compiler, but also high-level language code executable by a computer.

[0150] Recording media include hard disks, magnetic media such as floppy disks and magnetic tapes, optical media such as Compact Disk Read Only Memory (CD-ROMs) and Digital Video Disks (DVDs), magneto-optical media such as a floppy disk, hardware devices configured to store and execute program instructions such as ROM, RAM, flash memory, etc.

[0151] It should be understood that the effects of the present disclosure are not limited to the effects described above, and include all effects that can be derived from a configuration of the invention described in detailed descriptions or claims of the present disclosure.

[0152] Although embodiments of the present disclosure have been described, the spirit of the present disclosure is not limited to the embodiments set forth in the description. A person skilled in the art who understands the spirit of the present disclosure can easily propose other embodiments by supplementing, changing, removing, or adding components within the scope of this spirit, but these embodiments are also within the spirit of the present disclosure. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] KR 10-2023-0104199

[0001]

Claims

[1] Device for detecting an anomaly (1) of a motor (18) arranged to generate a steering assistance torque in an electric power steering system (EPS) (10) of a vehicle, the device comprising: a memory (100) configured to store an artificial neural network model (110) configured to receive state data relating to a state of the vehicle and operating data relating to a steering operation of a driver of the vehicle, and to estimate a physical quantity relating to an output of the engine (18) using the state data and the operating data, and a processor (200) configured to input the state data relating to the state of the vehicle and the operation data relating to the driver's steering operation into the artificial neural network model (110), calculate an estimated value of the physical quantity relating to the output of the engine (18) based on the state data relating to the state of the vehicle and the operation data relating to the driver's steering operation using the artificial neural network model (110), and detect the abnormality of the engine (18) by comparing the estimated value of the physical quantity relating to the output of the engine (18) with an actual measured value of the physical quantity relating to the output of the engine (18). [2] Device (1) according to claim 1, wherein the status data and the operating data are contained in signals received via a Controller Area Network (CAN) of the vehicle. [3] The device (1) according to claim 1 or 2, wherein the physical quantity related to the output of the motor (18) is a steering torque generated by the driver's operation of a steering wheel (11) of the vehicle. [4] Device (1) according to one of claims 1 to 3, wherein the state data relating to the state of the vehicle comprises: a longitudinal speed 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] Device (1) according to one of claims 1 to 4, wherein the operating data relating to the driver's steering operation comprises: a steering angle and a steering angle speed. [6] Device (1) according to one of claims 1 to 5, wherein the artificial neural network model (110) comprises: a generator (111) configured to receive the state data relating to the state of the vehicle and the operating data relating to the driver's steering operation and to generate the estimated value of the physical quantity relating to the output of the motor (18); and a discriminator (112) configured to receive data related to the actual measured value, including the state data related to the state of the vehicle, the operation data related to the driver's steering operation, and the actual measured value of the physical quantity related to the output of the motor (18), and to output a discrimination value for the data related to the actual measured value. [7] The device (1) according to claim 6, wherein the discriminator (112) is arranged to receive estimated value-related data including the state data relating to the state of the vehicle, the operation data relating to the driver's steering operation, and the estimated value of the physical quantity relating to the output of the motor (18), and to output a discrimination value for the estimated value-related data. [8] Device (1) according to claim 7, wherein: the artificial neural network model (110) is constructed by alternately learning the generator (111) and the discriminator (112), and the state data, the operating data and the actual measured value used during the learning of the generator (111) and the discriminator (112) are obtained in a state in which the vehicle and the engine (18) are normal. [9] The apparatus (1) according to claim 8, wherein, in the state where the vehicle and the engine (18) are normal, the estimated value of the physical quantity related to the output of the engine (18) follows the actual measured value of the physical quantity related to the output of the engine (18). [10] Device (1) according to one of claims 6 to 9, wherein: the memory (100) is configured to store an anomaly detection model (120), and the processor (200) is configured to detect the abnormality of the motor (18) by inputting error data relating to a difference between the estimated value of the physical quantity related to the output of the motor (18) and the actual measured value related to the output of the motor (18) into the anomaly detection model (120). [11] The device (1) according to claim 10, wherein the error data input to the anomaly detection model (120) comprises: a mean and standard error of the difference between the estimated value and the actual measured value obtained from each of a plurality of data sets, including the state data relating to the state of the vehicle, the operating data relating to the driver's steering operation, the actual measured value of the physical quantity relating to the output of the motor (18), and the estimated value of the physical quantity relating to the output of the motor (18), a maximum absolute error between the actual measured value and the estimated value of the plurality of data sets, and the discrimination value of the discriminator (112) for actual measured value-related data from the plurality of data sets. [12] A method for detecting an anomaly of a motor (18) configured to generate a steering assist torque in an electric power steering system (EPS) (10) of a vehicle, the method comprising: Inputting (S120), by a processor (200), state data relating to a state of the vehicle and operating data relating to a driver's steering operation into an artificial neural network model (110), and calculating, by the processor (200), an estimated value of a physical quantity relating to an output of the engine (18) using the state data and the operating data; and Detecting (S130) the abnormality of the motor (18) by the processor (200) by comparing the estimated value of the physical quantity related to the output of the motor (18) with an actual measured value related to the output of the motor (18). [13] The method according to claim 12, wherein the physical quantity related to the output of the motor (18) is a steering torque generated by the driver's operation of a steering wheel (11) of the vehicle. [14] A method according to claim 12 or 13, wherein the condition data relating to the condition of the vehicle comprises: a longitudinal speed 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] A method according to any one of claims 12 to 14, wherein the operating data relating to the driver’s steering operation comprises: a steering angle and a steering angle speed. [16] Method according to one of claims 12 to 15, wherein the artificial neural network model (110) comprises: a generator (111) configured to receive the state data relating to the state of the vehicle and the operating data relating to the driver's steering operation and to generate the estimated value of the physical quantity relating to the output of the motor (18); and wherein calculating the estimated value of the physical quantity related to the output of the engine (18) comprises inputting, by the processor (200), the state data related to the state of the vehicle and the operation data related to the driver's steering operation to the generator (111) and outputting the estimated value generated by the generator (111). [17] The method of claim 16, wherein the artificial neural network model (110) further comprises: a discriminator (112) configured to receive data related to the actual measured value, including the state data related to the state of the vehicle, the operation data related to the driver's steering operation, and the actual measured value of the physical quantity related to the output of the motor (18), and to output a discrimination value for the data related to the actual measured value; and wherein calculating the estimated value of the physical quantity related to the output of the engine (18) comprises inputting, by the processor (200), actual measurement-related data, including the state data related to the state of the vehicle, the operation data related to the driver's steering operation, and the actual measurement value of the physical quantity related to the output of the engine (18), into the discriminator (112) and outputting, by the processor (200), a discrimination value for the data related to the actual measurement value generated by the discriminator (112). [18] The method of claim 17, wherein detecting the abnormality of the engine (18) comprises: Calculating (S131), by the processor (200), error data relating to a difference between the estimated value of the physical quantity relating to the output of the motor (18) and the actual measured value of the physical quantity relating to the output of the motor (18); and Inputting (132) the error data by the processor (200) into a one-class support vector machine (OCSVM) algorithm to detect the anomaly of the engine (18). [19] The engine abnormality detection method of claim 18, wherein the fault data input to the one-class support vector machine comprises: a mean and standard error of the difference between the estimated value and the actual measured value obtained from each of a plurality of data sets, including the state data relating to the state of the vehicle, the operating data relating to the driver's steering operation, the actual measured value of the physical quantity relating to the output of the motor (18), and the estimated value of the physical quantity relating to the output of the motor (18), a maximum absolute error between the actual measured value and the estimated value of the plurality of data sets, and the discrimination value of the discriminator (112) for actual measured value-related data from the plurality of data sets. [20] Non-transitory computer-readable storage medium having a program stored thereon which contains at least one instruction for carrying out the method according to one of claims 12 to 19.

Images