Failure prediction method using vehicle data

Abstract

The present invention relates to a method for predicting failures using vehicle data, wherein the method derives a correlation coefficient for failure prediction by performing correlation and regression analysis using vehicle data, sets the product fatigue level based on the derived correlation coefficient, and predicts the time of failure based on the change in the correlation coefficient. The method is implemented by a process in which a vehicle control unit (VCU) sets a failure prediction cycle and a reference correlation coefficient as a failure prediction condition; when the time of failure prediction arrives, the correlation coefficient is derived based on collected vehicle data by element; the change in the correlation coefficient is predicted by comparing the derived correlation coefficient with the reference correlation coefficient; the time of failure is predicted based on the predicted change in the correlation coefficient; and the predicted time of failure is transmitted to the vehicle and user terminal according to an alarm schedule.

Description

Technology Field

[0001] The present invention relates to a method for predicting failures of vehicle parts in advance based on vehicle sensors and driving data. In particular, it relates to a method for predicting failures using vehicle data, wherein correlation analysis and regression analysis are performed using vehicle data to derive a correlation coefficient for failure prediction, and the time of failure is predicted based on the amount of change in the correlation coefficient after setting the product fatigue level to the derived correlation coefficient. Background Technology

[0002] Modern vehicles are becoming more electronically advanced, and consequently, a wide variety of electronic devices are being installed and utilized. While these devices sometimes replace existing functions, they are more frequently installed to implement new capabilities. Consequently, as electronic devices are added alongside conventional mechanical components, the number of parts susceptible to failure is increasing significantly.

[0003] With the increase in components, electronic products for fault diagnosis capable of automatically diagnosing failures in certain functions of electronic or mechanical devices are being released and used.

[0004] For example, if an anomaly is detected while driving, it analyzes the fault code to accurately identify the problem area, provides information on replacing relevant consumables, and offers monthly reports such as vehicle operation data, the presence of abnormalities, and consumable replacement cycles.

[0005] However, this fault diagnosis method was not only capable of diagnosing only some parts, but was also limited to a simple diagnostic technology that notified of a failure only after it had occurred.

[0006] Therefore, there is a need for technology that enables fault prediction for each vehicle element (part) and provides this information in real-time to the vehicle and user terminals upon prediction, thereby allowing for proactive measures to be taken before the vehicle breaks down to ensure safe operation. Prior art literature

[0007] Korean Patent Publication No. 10-2018-0105950 (Vehicle fault diagnosis system and diagnosis method thereof) Korean Patent Publication No. 10-2019-0078705 (System and method for predicting vehicle failure by analyzing component aging patterns from vehicle data) Korean Patent Publication No. 10-2022-0075915 (Deep learning-based device and method for predicting remaining lifespan of automotive components) Korean Patent Publication No. 10-2024-0060018 (Vehicle component failure prediction system) Korean Patent Publication No. 10-2024-0107271 (Device and method for predicting remaining lifespan of vehicle components) Korean Patent Publication No. 10-2025-0037371 (Vehicle component lifespan prediction system) The problem to be solved

[0008] Therefore, the present invention is proposed to solve the problem of being limited to simple fault diagnosis that occurs in the general method of diagnosing vehicle failures as described above. Its purpose is to provide a fault prediction method using vehicle data, which derives a correlation coefficient for fault prediction by performing correlation analysis and regression analysis using vehicle data, sets the product fatigue level based on the derived correlation coefficient, and predicts the time of failure based on the amount of change in the correlation coefficient. means of solving the problem

[0009] In order to achieve the above-mentioned purpose, the "failure prediction method using vehicle data" according to the present invention is,

[0010] (a) A step of collecting vehicle data for fault prediction by element in a vehicle control device equipped with a fault prediction algorithm for predicting vehicle failures;

[0011] (b) a step of setting a fault prediction period for predicting a fault in the vehicle control device;

[0012] (c) A step of setting a reference correlation coefficient, which is a fault prediction condition, in the vehicle control device above;

[0013] (d) A step of confirming the fault prediction time in the vehicle control device (130);

[0014] (e) A step of deriving correlation coefficients based on element-specific vehicle data collected from the above vehicle control device;

[0015] (f) a step of predicting the change in the correlation coefficient by comparing the correlation coefficient derived from the vehicle control unit with the reference correlation coefficient; and

[0016] (g) The method is characterized by including a step of predicting the time of failure based on the predicted change in correlation coefficient above, and transmitting the predicted time of failure to a vehicle and a user terminal according to an alarm schedule.

[0017] Preferably, the above step (d) is,

[0018] It is characterized by determining that the fault prediction point is reached when the set fault prediction cycle, the occurrence of a fault defect, or the set driving distance is exceeded.

[0019] Preferably, the above step (e) is,

[0020] It is characterized by calculating correlation coefficients by processing collected vehicle data by element using correlation analysis and regression analysis.

[0021] Preferably, the above step (g) is,

[0022] It is characterized by determining the condition of a part by comparing the predicted change in correlation coefficient with the change in reference correlation coefficient set for each driving distance, predicting the time of failure by applying fatigue based on driving distance to the change in correlation coefficient according to the determined condition of the part, and transmitting the time of failure to the vehicle and user terminal according to an alarm schedule. Effects of the invention

[0023] According to the present invention, correlation and regression analysis are performed using vehicle data to derive a correlation coefficient for failure prediction, and after setting the product fatigue level based on the derived correlation coefficient, the failure time can be accurately predicted based on the amount of change in the correlation coefficient.

[0024] In addition, by transmitting the predicted failure time in real time to the in-vehicle system and user terminal, it enables proactive maintenance measures, which has the advantage of enhancing the safety and efficiency of vehicle operation. Brief explanation of the drawing

[0025] FIG. 1 is a schematic diagram of a vehicle system to which a fault prediction method using vehicle data according to the present invention is applied, and FIG. 2 is a flowchart showing a fault prediction method using vehicle data according to the present invention, and FIG. 3 is a flowchart showing the process of predicting the change in correlation coefficients in the present invention, and FIG. 4 is an example of a process for generating an early component failure prediction indicator in the present invention, and FIG. 5 is a graph predicting the change in correlation coefficient in the present invention, and FIG. 6 is an example of predicting the failure time in the present invention, and Figure 7a is an example of correlation coefficient analysis of LDC components, and Figure 7b is an example of correlation coefficient analysis of battery components, and FIG. 8 is a block diagram of a system for diagnosing noise-based abnormalities when the present invention is applied to a power electronic device. Specific details for implementing the invention

[0026] A method for predicting failures using vehicle data according to a preferred embodiment of the present invention will be described in detail below with reference to the attached drawings.

[0027] The terms or words used in the present invention described below should not be interpreted as being limited to their ordinary or dictionary meanings, but should be interpreted in a meaning and concept consistent with the technical spirit of the present invention, based on the principle that the inventor can appropriately define the concept of the terms to best describe his invention.

[0028] Therefore, the embodiments described in this specification and the configurations illustrated in the drawings are merely preferred embodiments of the present invention and do not represent all technical concepts of the present invention; thus, it should be understood that various equivalents and modifications that can replace them may exist at the time of filing this application.

[0029] FIG. 1 is a vehicle system (100) to which a fault prediction method using vehicle data according to a preferred embodiment of the present invention is applied, and may include an LDC (110) that converts a high DC voltage to a low voltage, an inverter (120) that converts DC power to AC power, a vehicle control unit (VCU) (130) that diagnoses vehicle faults and controls the overall operation of the vehicle, a battery management system (BMS) (140) for battery management, an audio / video / navigation (AVN) (170) for displaying information, a gateway (CGW) (150), and an integrated control unit (CCU) (160) that manages the network.

[0030] Reference numeral 200 represents a vehicle management server, and reference numeral 300 can be implemented as a user terminal, such as a mobile device like a smartphone.

[0031] FIG. 2 is a flowchart showing a method for predicting a failure using vehicle data according to a preferred embodiment of the present invention, comprising: (a) a step of collecting vehicle data for predicting a failure by element in a vehicle control unit (VCU) (130) equipped with a failure prediction algorithm for predicting a vehicle failure (S101); (b) a step of setting a failure prediction cycle for predicting a failure in the vehicle control unit (130) (S102); (c) a step of setting a reference correlation coefficient, which is a failure prediction condition, by part (by element) in the vehicle control unit (130) (S103); (d) a step of confirming a failure prediction time in the vehicle control unit (130) (S104 - S106); (e) a step of deriving a correlation coefficient based on the vehicle data by element collected in the vehicle control unit (130) (S107); (f) a step of predicting a change in the correlation coefficient by comparing the correlation coefficient derived in the vehicle control unit (130) with the reference correlation coefficient (S108); and (g) the predicted correlation coefficient It may include a step (S109 - S110) of predicting the time of failure based on the amount of change and transmitting the predicted time of failure to the vehicle and user terminal according to the alarm schedule.

[0032] The above step (d) can be determined as the point of failure prediction when the set failure prediction cycle, the occurrence of a failure defect, or the failure prediction distance is reached after the set driving distance.

[0033] In addition, the above step (e) can calculate correlation coefficients by processing the collected vehicle data by element through correlation analysis and regression analysis.

[0034] In addition, step (g) determines the state of a part by comparing the predicted change in correlation coefficient with the pre-set standard change in correlation coefficient for each driving distance, predicts the time of failure by applying fatigue according to driving distance to the change in correlation coefficient based on the determined state of the part, and can transmit the time of failure to the vehicle and user terminal according to the alarm schedule.

[0035] A method for predicting failures using vehicle data according to a preferred embodiment of the present invention configured as described above will be explained in detail below with reference to the attached drawings, FIGS. 1 to 7b.

[0036] First, the vehicle control unit (VCU) (130) is equipped with a fault prediction algorithm to predict vehicle failures and collects actual vehicle driving data (S101).

[0037] For example, the vehicle control unit (130) collects vehicle data (VDC; Vehicle Data Collection) for fault prediction by element. That is, as shown in FIG. 4, when the fault prediction target is an LDC (110), the LDC output voltage, LDC output current, LDC consumption current, LDC internal temperature, driving distance, and speed can be collected as vehicle data.

[0038] Next, the vehicle control unit (130) connects to the vehicle management server (200) to receive a fault prediction period (e.g., 5,000 km, 10,000 km, or 1,000 km in units of 1,000 km when the set driving distance is 18,000 km or more) for predicting failures by part (element) of the vehicle, stores it in internal memory, and sets the fault prediction period (S102).

[0039] Next, the vehicle control device (130) sets a reference correlation coefficient, which is a fault prediction condition for each component (element), and stores it in the internal memory (S103).

[0040] Here, the standard correlation coefficient for each element can be obtained through the vehicle management server (200).

[0041] Afterwards, the vehicle control unit (130) determines whether the time for a fault prediction has arrived based on the fault prediction period and current driving distance stored in the internal memory (S104 - S106).

[0042] For example, when the current driving distance reaches the set failure prediction cycle (e.g., 5,000 km), when a failure occurs, or when the failure prediction distance (1,000 km) is reached after the set driving distance (18,000 km), it is determined that it is the failure prediction time.

[0043] Subsequently, if it is determined that the failure prediction cycle has arrived, the correlation coefficient ( ) derives (S107).

[0044]

[0045] Here, correlation coefficients can be derived by processing the collected vehicle data by element using correlation and regression analysis. Specifically, they can be derived by using correlation analysis between LDC output voltage and temperature, LDC output current and temperature, LDC consumption current and temperature, LDC internal temperature, mileage and SOC, or speed and consumption current.

[0046] Next, the correlation coefficient derived from the vehicle control device (130) is compared with the reference correlation coefficient to predict the change in the correlation coefficient (S108).

[0047] That is, as illustrated in Fig. 5, the difference can be extracted by comparing the derived correlation coefficient with the reference correlation coefficient of the corresponding part for each driving distance, and the amount of change in the correlation coefficient can be predicted based on the extracted difference, which can be expressed as a formula as follows.

[0048]

[0049] Next, the failure time is predicted based on the change in the correlation coefficient predicted above.

[0050] That is, as shown in Fig. 3, the weighted sum of the changes in the correlation coefficients of various variables is converted into a "failure score" (S121 - S122). This can be expressed as a formula as follows.

[0051]

[0052] Next, the condition of the part is determined by comparing the change in the correlation coefficient (Δr) with the change in the reference correlation coefficient set for each driving distance (S123).

[0053] Here, if the change in correlation coefficient (Δr) is 0.1, it is determined that it is within the range of the standard value set for each mileage and the part is determined to be normal; if the change in correlation coefficient (Δr) is 0.2, it is determined that it is in a state requiring maintenance; if the change in correlation coefficient (Δr) is 0.3, it is determined that the LDC cooling system needs maintenance; and if the change in correlation coefficient (Δr) is 0.4, it is determined that it is in an abnormal state.

[0054]

[0055]

[0056] By fitting this function, it is possible to predict the point at which the end of life (e.g., efficiency below 85%, ripple more than twice) occurs in the trend of 10km to 200km.

[0057]

[0058] Since the rate of degradation varies depending on the temperature, it is corrected using the following formula.

[0059] for example,

[0060] E a = 0.7eV (based on electrolytic capacitor)

[0061] T1 = 298K (25℃)

[0062] T2 = 328K (55℃) -> AF When the temperature rises from 6 to 30℃, the lifespan is shortened to 1 / 6, reflecting the typical trend of "a 10℃ increase in temperature leads to a halving of the lifespan."

[0063] Figure 7a is an example of correlation analysis of LDC components, and Figure 7b is an example of correlation analysis of battery components.

[0064] Based on the condition of the part determined in this way, the fatigue level according to the driving distance is applied to the change in correlation amount to predict the time of failure as shown in Fig. 6, and the time of failure is transmitted to the vehicle and user terminal according to the alarm schedule (S109 - S110).

[0065] By notifying the user (driver) of the predicted failure time through the in-vehicle system and real-time user terminal, the user can take precautionary measures before a failure occurs, thereby promoting safe driving.

[0066] FIG. 8 is a block diagram showing how to predict failures due to noise and vibration by applying the failure prediction of the present invention to power electronic devices (LDC / OBC / battery pack), etc., of a vehicle.

[0067] Noise in power electronic devices (LDC / OBC / battery pack) is measured using a hybrid acoustic-vibration sensor (MEMS microphone + piezoelectric vibration sensor), and the measured noise signal is preprocessed through a preprocessing module. Specifically, the noise measurement signal is preprocessed using adaptive band / notch filters, and the signal is processed through FFT / STFT / Wavelet analysis.

[0068] Next, the processed noise signal is processed by an artificial intelligence (AI)-based diagnostic engine to extract soundness. Specifically, the diagnostic engine is trained using sound signature learning, and the noise signal processed by the trained engine is analyzed to calculate an anomaly score and perform self-correction.

[0069] Next, the extracted health is evaluated, and a 3-stage warning (preliminary, warning / urgent) is performed based on the evaluation results, and the health evaluation results are transmitted to a user terminal in real time through the vehicle control unit (ECU) and cloud integration.

[0070] According to the present invention described above, correlation analysis and regression analysis are performed using vehicle data to derive a correlation coefficient for failure prediction, and after setting the product fatigue level based on the derived correlation coefficient, the time of failure can be accurately predicted based on the amount of change in the correlation coefficient.

[0071] In addition, the safety of vehicle operation can be ensured by transmitting the predicted failure time to the vehicle and user terminal to allow for preemptive measures to be taken.

[0072] Although the invention made by the inventors has been specifically described according to the above embodiments, it is obvious to those skilled in the art that the invention is not limited to the above embodiments and can be modified in various ways without departing from the gist thereof. Explanation of the symbols

[0073] 100: Vehicle System 110: LDC 120: Inverter 130: Vehicle Control Unit (VCU) 140: Battery Management System (BMS) 150: Gateway (CGW) 160: Integrated Control Unit (CCU) 170: Audio / Video / Navigation (AVN) 200: Vehicle Management Server 300: User Terminal

Claims

Claim 1 (a) a step of collecting vehicle data for fault prediction by element in a vehicle control unit (VCU) equipped with a fault prediction algorithm for predicting vehicle failures; (b) a step of setting a fault prediction cycle for predicting failures in the vehicle control unit (VCU); (c) a step of setting a reference correlation coefficient as a fault prediction condition in the vehicle control unit; (d) a step of confirming the fault prediction time in the vehicle control unit; (e) a step of deriving a correlation coefficient based on the vehicle data collected by element in the vehicle control unit; (f) a step of predicting the amount of change in the correlation coefficient by comparing the correlation coefficient derived in the vehicle control unit with the reference correlation coefficient; and (g) a step of predicting the time of failure based on the predicted change in correlation coefficient and transmitting the predicted time of failure to a vehicle and a user terminal according to an alarm schedule, wherein the correlation coefficient in step (e) is derived by using correlation analysis between LDC output voltage and temperature, by using correlation analysis between LDC output current and temperature, by using correlation analysis between LDC consumption current and temperature, by using analysis of LDC internal temperature, by using correlation analysis between driving distance and SOC, or by using correlation analysis between speed and consumption current, when the component for which the time of failure is to be predicted is an LDC; the correlation coefficient in step (f) extracts the difference by comparing the derived correlation coefficient with the reference correlation coefficient of the corresponding component according to driving distance and predicts the change in correlation coefficient based on the extracted difference; the state of the component is determined by comparing the predicted change in correlation coefficient (Δr) with the pre-set reference change in correlation coefficient according to driving distance, and the time of failure is determined by applying fatigue according to driving distance to the change in correlation coefficient according to the determined state of the component. A method for predicting failures using vehicle data, characterized by predicting the time of failure and transmitting the time of failure to the vehicle and user terminal according to an alarm schedule. Claim 2 A method for predicting a failure using vehicle data in claim 1, wherein step (d) determines that the failure prediction point is reached when a set failure prediction period or a failure defect occurs or a set driving distance is reached. Claim 3 Claim 1, wherein step (e) is characterized by calculating a correlation coefficient by processing collected element-specific vehicle data through correlation analysis and regression analysis. Claim 4 delete

Images