Multifunctional automobile engine noise detection system

Abstract

The invention discloses a multifunctional automobile engine noise detection system which comprises a signal acquisition module, a data preprocessing module, a noise analysis module, a multifunctional application module and a result output module. The signal acquisition module synchronously acquires acoustics, vibration, thermal imaging and working condition multi-source signals of an engine, the data preprocessing module performs feature extraction and fusion on the acquired multi-source signals to generate a multi-dimensional feature set, and the noise analysis module analyzes the position, intensity and frequency characteristics of a noise source based on the multi-dimensional feature set. And the multifunctional application module realizes fault diagnosis and predictive maintenance through a machine learning model based on the analysis result and the multi-dimensional feature set, and the result output module performs visual presentation and report output on the analysis and application result. The multifunctional automobile engine noise detection system can solve the problems that a traditional detection method depends on artificial experience, fault positioning is not accurate, the diagnosis dimension is single, and prediction capability is lacked.

Description

Technical Field

[0001] This invention relates to the field of automotive engine testing technology, specifically a multifunctional automotive engine noise detection system. Background Technology

[0002] Automotive engine noise is a direct indicator of engine operating status. Its sources include mechanical vibration, fuel combustion, and airflow disturbance. Accurate noise detection and its characteristics are crucial for engine fault diagnosis and reliability assurance. However, traditional engine noise detection methods suffer from the following drawbacks, affecting detection efficiency and accuracy: First, fault location relies on human experience. In traditional repair scenarios, repair personnel often determine the source of noise by listening to the air. This method is highly dependent on personal experience and is easily affected by environmental noise, which may lead to misjudgment of fault location. For example, the rustling sound of a loose timing chain may be misjudged as an abnormal noise of excessive valve clearance, resulting in incorrect repair direction and increased repair costs.

[0003] Second, the diagnostic dimensions are limited and cannot fully analyze the causes of noise. Traditional testing equipment such as sound pressure meters only focus on noise intensity and cannot combine multi-source signals for comprehensive analysis.

[0004] Third, there is a lack of predictive maintenance capabilities. Engine failures are usually diagnosed after the failure occurs, and the remaining lifespan of components cannot be predicted through noise trend analysis.

[0005] Therefore, there is an urgent need for a multi-functional automotive engine noise detection system that integrates multi-source signal acquisition, multi-dimensional noise feature analysis, intelligent fault diagnosis, and predictive maintenance functions to address the aforementioned pain points in existing technologies and improve detection accuracy and maintenance efficiency. Summary of the Invention

[0006] To address the shortcomings of the aforementioned background technology, this invention integrates multi-source signal acquisition, intelligent analysis, and multi-functional applications to achieve precise location, comprehensive diagnosis, and predictive maintenance of engine fault noise, thereby improving detection accuracy and maintenance efficiency.

[0007] To achieve the above objectives, the present invention provides a multifunctional automotive engine noise detection system, comprising: The signal acquisition module is used to acquire multi-source signals from the engine compartment of a car, including acoustic signals, vibration signals, thermal imaging signals, and OBD data. The data preprocessing module is used to preprocess the acquired multi-source signals, including noise reduction, time synchronization, single-source feature extraction and multi-source feature fusion, and finally generate a multi-dimensional feature set. The noise analysis module is used to perform noise analysis on the multi-dimensional feature set obtained after preprocessing, including noise intensity calculation, frequency feature extraction, noise source localization, and operating condition correlation analysis. A multi-functional application module is used to achieve fault diagnosis, predictive maintenance, and model adaptation based on noise analysis results; The results output module is used to output the detection results, including visualization and report generation.

[0008] Specifically, the signal acquisition module includes: The multi-source sensor array unit includes a microphone array, several accelerometers and an infrared thermal imager, used to collect acoustic signals, vibration signals and thermal imaging signals during engine operation, respectively. The OBD data interface unit connects to the car's OBD-II interface via the CAN bus to obtain engine operating parameters in real time, including engine speed, engine load, coolant temperature, intake pressure, and oil pressure. The signal synchronization unit uses a GPS timing module to add timestamps to multi-source signals.

[0009] Specifically, the data preprocessing module includes: The noise reduction unit uses an adaptive filter to eliminate environmental noise for acoustic signals, a low-pass filter to remove high-frequency electronic noise for vibration signals, and a median filter to remove salt-and-pepper noise for thermal imaging signals. The synchronization alignment unit associates multiple source signals through timestamps and adjusts the multiple source signals to the same time axis; The feature extraction unit extracts frequency domain features from acoustic signals and calculates the unweighted 1 / 3 octave spectrum through Fourier transform; it calculates the vibration mean and peak value from vibration signals through time domain analysis and calculates the peak frequency through frequency domain analysis; it extracts temperature data at various locations from thermal imaging signals; and it extracts numerical features of operating parameters from OBD data. The multi-source feature fusion unit fuses the extracted single-source features to generate a multi-dimensional feature set.

[0010] Specifically, the noise analysis module includes: The noise intensity analysis unit, based on the acoustic features in the multidimensional feature set, calculates the A-weighted sound pressure level of each frequency band in the unweighted 1 / 3 octave band spectrum by energy superposition, and performs time series analysis to calculate the A-weighted sound pressure level at each time point, generating a curve of noise intensity changing with time. The noise frequency characteristic analysis unit plots an A-weighted 1 / 3 octave spectrum with the center frequency as the abscissa and the A-weighted sound pressure level as the ordinate, extracts the peak frequency, the second highest peak frequency, and the peak sound pressure level, and determines the noise type based on the frequency components. The noise source localization unit locates the spatial source of noise based on acoustic features, vibration features, and temperature features in a multi-dimensional feature set. The operating condition correlation analysis unit analyzes the variation law of noise with operating conditions based on the acoustic features and OBD operating condition features in the multidimensional feature set, plots the operating condition-noise curve, calculates the correlation between noise features and operating condition parameters using the Pearson correlation coefficient, and compares the current curve with the operating condition-noise model of a normal engine to determine whether it is abnormal.

[0011] Specifically, the working process of the noise source localization unit is as follows: (1) The delay summation beamforming algorithm is used to delay the time of each microphone signal in the microphone array. Based on the microphone position and the speed of sound propagation, the delay time in each direction is calculated, and then the sound pressure level in each direction is summed to obtain the noise source localization heat map. (2) Calculate the cross power spectral density of acoustic signal and vibration signal through cross-spectral analysis to determine whether acoustic signal and vibration signal are from the same source, so as to help confirm that the noise source comes from the mechanical parts of the engine; (3) Use thermal imaging signals to check the highest temperature at the noise source location to determine if there is a combustion fault.

[0012] Specifically, the multi-functional application module includes: The fault diagnosis unit establishes a multi-source feature-fault relationship database. Based on noise analysis results and multi-dimensional feature sets, it uses a rule engine to match candidate faults and then optimizes them through a random forest model to output a structured fault report. The predictive maintenance unit uses an LSTM model to analyze the time series of noise characteristics, combines the relationship between component wear and noise characteristics, predicts the remaining lifespan, and outputs a predictive maintenance report.

[0013] Specifically, the workflow of the fault diagnosis unit is as follows: (1) Rule engine filtering: transform the multi-source feature combination-fault type rules in the multi-source feature-fault relationship database into condition judgment statements and filter out candidate faults that meet the conditions. (2) Use the random forest model to rank the candidate faults output by the rule engine by probability; (3) Output a structured fault report, including fault type, fault location, and verification basis.

[0014] Specifically, the workflow of the predictive maintenance unit is as follows: (1) Time series feature extraction: Obtain monthly noise feature data of the engine under test over the past n months, and extract trend features from them, including acoustic signal frequency peak, vibration peak, and maximum A-weighted sound pressure value; (2) Predicting nonlinear trends using LSTM models; (3) Based on the relationship between component wear and noise characteristics, calculate the remaining life according to the output results of the LSTM model, and output a predictive maintenance report, including the remaining life and the verification basis.

[0015] Specifically, the result output module includes: The visualization unit supports display on a screen, and the content includes noise intensity changing curve over time, A-weighted 1 / 3 octave spectrum, and noise source location heat map. The report generation unit generates standardized reports, including model compatibility results, noise analysis results, structured fault reports, and predictive maintenance reports.

[0016] In summary, the beneficial technical effects of the present invention are as follows: By acquiring and intelligently fusioning acoustic, vibration, thermal imaging, and OBD multi-source signals, and comprehensively utilizing multi-dimensional features for analysis, the problem of single diagnostic dimensions in existing technologies is solved; the collaborative application of multiple modules solves the problem of reliance on human experience in existing technologies; and predictive maintenance is achieved through multi-functional application modules. Attached Figure Description

[0017] Figure 1 This is a system structure diagram of the present invention. Detailed Implementation

[0018] To make the technical means, creative features, objectives and effects of this invention clearer and easier to understand, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0019] Example like Figure 1 As shown, the multifunctional automotive engine noise detection system provided by the present invention includes a signal acquisition module, a data preprocessing module, a noise analysis module, a multifunctional application module, and a result output module.

[0020] The signal acquisition module is responsible for acquiring acoustic signals, vibration signals, thermal imaging signals and operating parameters during engine operation, providing raw data for subsequent processing. It includes a multi-source sensor array unit, an OBD data interface unit and a signal synchronization unit. The multi-source sensor array unit includes a microphone array, several accelerometers, and an infrared thermal imager. The microphone array is mounted on the front crossbeam of the engine compartment, pointing towards the center of the engine, and is used to collect acoustic signals of engine operation, including mechanical noises such as piston knocking and excessive valve clearance. The accelerometers are multiple magnetic piezoelectric accelerometers, which are installed in the engine cylinder head, oil pan, timing chain cover, etc., to capture vibration signals of various mechanical components such as valves, pistons, crankshaft, connecting rod shaft, and timing system. The infrared thermal imager is a miniaturized infrared thermal imager, which is installed inside the engine hood facing the engine surface, and is used to monitor the engine temperature distribution. The OBD data interface unit connects to the car's OBD-II interface via a CAN bus to acquire engine operating parameters in real time, including engine speed, engine load, coolant temperature, intake pressure, ignition advance angle, and oil pressure. Among these parameters, engine speed reflects the engine's operating speed and is directly related to the frequency of mechanical noise; engine load reflects the engine's current power output and is positively correlated with the intensity of combustion noise; coolant temperature reflects the engine's thermal state, and abnormal coolant temperature may lead to increased combustion noise; intake pressure reflects the state of the intake system, and poor intake may lead to increased combustion noise; and oil pressure reflects the lubrication status, and low oil pressure may lead to bearing wear and abnormal vibration. The signal synchronization unit uses a GPS timing module to add precise timestamps to multi-source signals.

[0021] The data preprocessing module is responsible for transforming the raw signal into structured features that can be used for noise analysis, including a denoising unit, a synchronization alignment unit, a feature extraction unit, and a multi-source feature fusion unit; The noise reduction unit uses an NLMS adaptive filter for acoustic signals, taking the signal collected by the background noise microphone outside the engine compartment as a reference to cancel environmental noise in real time. It uses a low-pass filter to remove the high-frequency electronic noise of the sensor itself for vibration signals, and a 3×3 median filter to remove salt-and-pepper noise in thermal imaging signals. The synchronization alignment unit reads the timestamp of each signal, uses GPS time as a reference, interpolates and aligns the sampling data from different devices to obtain a time-synchronized multi-source signal; The feature extraction unit extracts frequency domain features from acoustic signals, calculates the unweighted 1 / 3 octave band spectrum using Fast Fourier Transform, with a center frequency ranging from 25Hz to 16kHz, totaling 31 frequency bands. The sound pressure level of each frequency band is calculated using the sound pressure level calculation formula. For vibration signals, the unit calculates the vibration mean and peak value through time domain analysis, and calculates the vibration spectrum through frequency domain analysis, extracting the peak frequency. For thermal imaging signals, the unit extracts temperature data at various locations. For OBD data, the unit directly extracts numerical features of operating condition parameters. The multi-source feature fusion unit concatenates the above single-source features according to time points to generate a multi-dimensional feature set, with each time point corresponding to a feature vector.

[0022] The noise analysis module provides key information for subsequent fault diagnosis and predictive maintenance of multi-functional application modules through multi-dimensional noise characteristic analysis, including a noise intensity analysis unit, a noise frequency characteristic analysis unit, a noise source location unit, and an operating condition correlation analysis unit. The noise intensity analysis unit quantifies the noise perception level. Its input is the acoustic features in the multi-dimensional feature set. First, the sound pressure level of each frequency band in the unweighted 1 / 3 octave band spectrum is obtained by energy superposition to obtain the total A-weighted sound pressure level. Then, time series analysis is performed to statistically analyze the total A-weighted sound pressure level at each time point. The output results can be used as the time series features of the predictive maintenance unit to generate a noise intensity change curve over time and transmit it to the result output module for visualization. The noise frequency feature analysis unit is used to analyze the causes of noise, extract the frequency distribution characteristics of noise, and determine the noise type. Its input is the acoustic features in a multi-dimensional feature set. The specific workflow is as follows: plotting the A-weighted 1 / 3 octave spectrum with the center frequency of the 1 / 3 octave spectrum as the x-axis and the A-weighted sound pressure level as the y-axis, identifying and extracting the peak frequency, the second highest peak frequency, and the peak sound pressure level in the spectrum, and classifying the noise type according to the range of the peak frequency: mechanical noise: 500-3000Hz, including piston knocking, excessive valve clearance, etc.; combustion noise: 100-500Hz, including detonation, incomplete combustion, etc.; aerodynamic noise: above 3000Hz, including intake / exhaust noise. The output of the noise frequency feature analysis unit includes the A-weighted 1 / 3 octave spectrum, peak frequency, and noise type. On the one hand, it is sent to the result output module for visual display of the spectrum, and on the other hand, it serves as a key feature of the fault diagnosis unit. The noise source localization unit is used to locate the spatial source of noise based on acoustic, vibration, and temperature characteristics in a multi-dimensional feature set. Its workflow is as follows: (1) The time delay of each microphone signal in the microphone array is calculated using the delay-sum beamforming algorithm. Based on the microphone position and the speed of sound propagation, the delay time in each direction is calculated, and then the sound pressure level in each direction is obtained by summing the delay times. The formula is as follows: in It is the direction angle. For microphone weights, Given the signal from the i-th microphone, a noise source localization heatmap is generated by calculating the sound pressure level in all directions. (2) Perform cross-spectral analysis to calculate the cross-power spectral density of the acoustic signal and the vibration signal, and determine whether the acoustic signal and the vibration signal originate from the same source. The formula is as follows: in, For the Fourier transform of the acoustic signal, For the Fourier transform of the vibration signal, As the expectation operator, if the CPSD amplitude is ≥0.8, it is determined that the acoustic signal and the vibration signal have the same origin, confirming that the noise comes from the engine mechanical parts; (3) Check whether the highest temperature extracted by the thermal imaging signal is abnormal. If the temperature is abnormal, it is determined that the noise may come from the combustion system. If the temperature is normal, it is further confirmed that the noise comes from mechanical parts. The operating condition correlation analysis unit is used to analyze the variation of noise characteristics with engine operating conditions and determine whether there are any abnormalities. Its inputs are acoustic characteristics and OBD operating condition characteristics from a multi-dimensional feature set. The process includes: 1. Plotting an operating condition-noise curve with engine speed or load as the x-axis and A-weighted sound pressure level or peak frequency as the y-axis; 2. Calculating the correlation between noise characteristics and operating condition parameters using the Pearson correlation coefficient, the formula is as follows: ,in, Noise characteristics, These are the operating parameters. For covariance, Let be the standard deviation, if If the two are strongly positively correlated, it indicates that the noise increases with the increase of operating conditions; 3. Compare the current operating condition-noise curve with the operating condition-noise model of a normal engine. If the curve deviates from the threshold, it is determined that the noise is abnormal. For example, if the A-weighted sound pressure level is too high at idle, it may be a problem with the idle speed motor. The output is sent to the result output module to visualize the curve and serve as the basis for the fault diagnosis unit.

[0023] The multi-functional application module is used to implement fault diagnosis, predictive maintenance, and model adaptation functions based on the output of the noise analysis module and the multi-dimensional feature set of the data preprocessing module through machine learning models. It includes a fault diagnosis unit, a predictive maintenance unit, and a model adaptation unit. The inputs to the fault diagnosis unit include the noise source localization results, frequency characteristics, and operating condition correlation results output by the noise analysis module, as well as the multi-dimensional feature set output by the data preprocessing module. Its workflow is as follows: 1. Construct a multi-source feature-fault relationship database: Based on a large number of engine fault cases, organize the correspondence rules between multi-source feature combinations and fault types, for example: Rule 1: Mechanical noise (peak frequency 500-1000Hz) + noise source located in cylinder head + vibration peak frequency and acoustic peak frequency from the same source (cross-spectral density ≥0.8) + idling condition (engine speed 1000±100rpm) → piston knocking; Rule 2: Mechanical noise (peak frequency 2000-3000Hz) + noise source located in valve chamber + normal thermal imaging temperature (cylinder head temperature 70-90℃) + high load conditions (engine load ≥80%) → excessive valve clearance; Rule 3: Aerodynamic noise (peak frequency ≥ 3000 Hz) + noise source location: exhaust manifold + abnormal thermal imaging temperature (exhaust manifold temperature ≥ 120 ℃) ​​+ abnormal intake pressure (intake pressure ≤ 20 kPa) → exhaust blockage.

[0024] 2. Filter out candidate faults through the rule engine: Input the multi-dimensional feature set into the rule engine, match the combination of multi-source features, and filter out candidate faults that meet the conditions. For example, the features of the engine being tested are: peak frequency 600Hz (mechanical noise), noise source is located in the cylinder head, vibration peak frequency 600Hz (same source), engine speed 1000rpm (idle speed) → match rule 1, the candidate fault is piston knocking.

[0025] 3. Random Forest Model Optimization: For multiple candidate faults output by the rule engine, the fault probability is calculated by the random forest model. The training of the random forest model is based on the historical fault-feature training set, which comes from a large amount of engine fault data. Each data point includes a set of multi-source features and a fault label.

[0026] 4. Output structured fault report: Integrate rule matching results with random forest probability to output structured fault report, including fault type, fault location and verification basis. The format of verification basis is as follows: "peak frequency 600Hz, noise source located in cylinder head, vibration peak frequency from the same source, idling condition, matching rule 1, random forest probability 95%".

[0027] The predictive maintenance unit is used to analyze the time series trend of noise characteristics and predict the remaining life of components. Its workflow is as follows: 1. Time series feature extraction: Extract trend features from the historical multidimensional feature set of the detected engine, including the monthly change of peak frequency and the monthly change of maximum A-weighted sound pressure level in acoustic features, and the monthly change of vibration peak value in vibration features; 2. LSTM model predicts nonlinear trends: Input time series features into the LSTM model. The model is trained based on historical fault data. The input is the trend features of the past n months, and the output is the trend prediction for the next m months.

[0028] 3. Calculate remaining life by combining component wear models: If the fault is determined to be mechanical, combine relevant component wear models, such as the relationship model between piston clearance and peak frequency, to convert the trend characteristics predicted by LSTM into the component wear state, and then calculate the remaining life. 4. Output Predictive Maintenance Report: Integrate LSTM prediction results with component wear model calculations to output a predictive maintenance report, which includes remaining life and verification basis. The verification basis is formatted as "The peak frequency is predicted to rise to 660Hz in the next 3 months, corresponding to a piston clearance of 1.6mm and an average monthly wear rate of 0.05mm / month".

[0029] The results output module includes: The visualization unit supports display on a screen, and the displayed content includes noise intensity changing curve over time, A-weighted 1 / 3 octave spectrum, and noise source location heat map. The report generation unit generates standardized reports, including model compatibility results, noise analysis results, structured fault reports, and predictive maintenance reports.

[0030] Therefore, the multifunctional automotive engine noise detection system provided by this invention integrates functional modules such as multi-source signal acquisition, intelligent data preprocessing, multi-dimensional noise feature analysis, and multifunctional applications. It realizes the entire process of engine "multi-source signal fusion - noise feature analysis - intelligent decision-making" detection, solving the pain points of existing technologies that rely on human experience, which can easily lead to misjudgment of fault location, single diagnostic dimensions that cannot fully analyze the causes of noise, and lack of predictive maintenance capabilities. It improves the accuracy and maintenance efficiency of engine noise detection, is easy to operate, and can provide early warning of the remaining lifespan of components, ensuring the reliability and safety of engine operation.

[0031] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-functional automobile engine noise detection system, characterized by, The application relates to a vehicle engine noise analysis system. The system comprises: a signal acquisition module for acquiring multi-source signals in an engine compartment of a vehicle, including acoustic signals, vibration signals, thermal imaging signals and OBD data; a data preprocessing module for preprocessing the acquired multi-source signals, including denoising, time synchronization, single-source feature extraction and multi-source feature fusion, and finally generating a multi-dimensional feature set; a noise analysis module for analyzing the multi-dimensional feature set obtained after preprocessing, including noise intensity calculation, frequency feature extraction, noise source positioning and working condition correlation analysis; a multi-functional application module for realizing fault diagnosis and predictive maintenance based on the noise analysis results; 2. The multi-functional automobile engine noise detection system according to claim 1, wherein a result output module for outputting detection results, including visual display and report generation. The signal acquisition module comprises: a multi-source sensor array unit including a microphone array, a plurality of acceleration sensors and an infrared thermal imager, for respectively acquiring acoustic signals, vibration signals and thermal imaging signals during engine operation; an OBD data interface unit connected to an automobile OBD-II interface through a CAN bus, for acquiring working condition parameters of the engine in real time, including engine speed, engine load, cooling water temperature, intake pressure and oil pressure; 3. The multi-functional automobile engine noise detection system of claim 2, wherein, a signal synchronization unit adding time stamps to the multi-source signals by using a GPS timing module. The data preprocessing module comprises: a denoising unit using an adaptive filter to eliminate environmental noise for the acoustic signals, using a low-pass filter to remove high-frequency electronic noise for the vibration signals, and using a median filter to remove salt and pepper noise for the thermal imaging signals; a synchronization alignment unit associating the multi-source signals by using the time stamps and adjusting the multi-source signals to the same time axis; a feature extraction unit extracting frequency domain features for the acoustic signals, calculating unweighted 1 / 3 octave spectrum by Fourier transform; calculating vibration mean and vibration peak by time domain analysis and calculating vibration peak frequency by frequency domain analysis for the vibration signals; extracting temperature data at each position for the thermal imaging signals; and extracting numerical features of working condition parameters for the OBD data; 4. The multi-functional automobile engine noise detection system of claim 3, wherein, a multi-source feature fusion unit fusing the extracted single-source features to generate a multi-dimensional feature set. The noise analysis module comprises: a noise intensity analysis unit calculating A-weighted sound pressure level of each frequency band in the unweighted 1 / 3 octave spectrum by energy superposition based on the acoustic features in the multi-dimensional feature set, and performing time series analysis to calculate A-weighted sound pressure level at each time point, and generating a curve of noise intensity changing with time; a noise frequency feature analysis unit drawing an A-weighted 1 / 3 octave spectrum with center frequency as the horizontal coordinate and A-weighted sound pressure level as the vertical coordinate, extracting peak frequency, secondary peak frequency and peak sound pressure level, and judging noise types according to frequency components; a noise source positioning unit positioning the spatial source of noise based on the acoustic features, vibration features and temperature features in the multi-dimensional feature set; a working condition correlation analysis unit analyzing the change rule of noise with working conditions based on the acoustic features and OBD working condition features in the multi-dimensional feature set, drawing a working condition-noise curve, calculating the correlation between noise features and working condition parameters by using a Pearson correlation coefficient, comparing the current curve with a working condition-noise model of a normal engine, and judging whether it is abnormal.

5. The multi-functional automobile engine noise detection system of claim 4, wherein: The working process of the noise source positioning unit is as follows: (1) The time delay of each microphone signal of the microphone array is calculated by using the delay-and-sum beamforming algorithm, the delay time of each direction is calculated according to the microphone position and the sound propagation speed, and then the sound pressure level in each direction is summed to obtain a noise source positioning heat map; (2) The cross-power spectral density of the acoustic signal and the vibration signal is calculated by cross-spectrum analysis to determine whether the acoustic signal and the vibration signal are homologous, so as to assist in confirming that the noise source comes from the engine mechanical parts; (3) The highest temperature of the noise source position is checked by using the thermal imaging signal to assist in judging whether it is a combustion failure.

6. The multi-functional automobile engine noise detection system of claim 4, wherein: The multifunctional application module comprises: A fault diagnosis unit establishes a multi-source feature-fault relationship library, and according to the noise analysis result and the multi-dimensional feature set, a rule engine is used to match candidate faults, and then a random forest model is used for optimization to output a structured fault report; A predictive maintenance unit uses an LSTM model to analyze the time series of noise characteristics, combines the relationship between component wear and noise characteristics, predicts the remaining life, and outputs a predictive maintenance report.

7. The multi-functional automobile engine noise detection system of claim 6, wherein: The working process of the fault diagnosis unit is as follows: (1) The rule engine filters the multi-source feature combination-fault type rules in the multi-source feature-fault relationship library into conditional judgment statements, and filters out the candidate faults that meet the conditions; (2) The random forest model is used to sort the candidate faults output by the rule engine by probability; (3) A structured fault report is output, including fault type, fault location and verification basis.

8. The multi-functional automobile engine noise detection system of claim 6, wherein: The working process of the predictive maintenance unit is as follows: (1) Time series feature extraction: obtain the monthly noise feature data of the detected engine in the past n months, and extract trend features including acoustic signal frequency peak, vibration peak and maximum A-weighted sound pressure value; (2) Use the LSTM model to predict the nonlinear trend; (3) Based on the relationship model between component wear and noise characteristics, the remaining life is calculated according to the output result of the LSTM model, and a predictive maintenance report is output, including the remaining life and verification basis.

9. The multi-functional automobile engine noise detection system of claim 6, wherein: The result output module comprises: A visual display unit supports display screen display, including noise intensity-time curve, A-weighted 1 / 3 octave spectrum and noise source positioning heat map; A report generation unit generates a standardized report, including model adaptation result, noise analysis result, structured fault report and predictive maintenance report.

Images