Engine sound-vibration multi-information fusion diagnosis device and fault diagnosis method

Abstract

This invention relates to an engine acoustic-vibration multi-information fusion diagnostic device and fault diagnosis method. It removes interference components from the acoustic time-domain data of a faulty engine; through three levels—acoustic-vibration time-domain feature correlation, acoustic order data, and vibration order data—it uses a lookup table method to quickly classify and locate fault features, thereby rapidly determining the possible fault types of the engine; it employs time-domain / frequency-domain multi-layer signal fault information extraction methods to better handle strongly periodic abnormal noise signals and non-steady-state signals, thus facilitating the identification of fault types with non-steady-state noise characteristics. Compared with existing engine on-site abnormal noise diagnostic schemes, this method has good adaptability, and by comparing the features of multi-dimensional information results and performing correlation analysis, it can effectively filter external interference signals during the testing process and locate the source of abnormal noise from a higher-dimensional mechanistic level.

Description

Technical Field

[0001] This invention relates to a vehicle fault diagnosis technology, and more particularly to an engine sound-vibration multi-information fusion diagnostic device and fault diagnosis method for on-site diagnosis. Background Technology

[0002] As the core of a vehicle's powertrain, the engine's primary function is to provide the mechanical energy required for the vehicle's operation. Engine system malfunctions are also a major component of vehicle powertrain system malfunctions. Among these, abnormal engine noises are particularly prevalent, accounting for a significant portion of customer complaints. Due to the complex structure of the engine system, the large number of accessories, and the fact that its noise during operation comprises multiple components including mechanical component noise, hydrodynamic noise, and combustion noise, its malfunctions are diverse in form, have wide-ranging sources, and complex mechanisms.

[0003] Currently, in on-site repair, engine noise malfunctions are typically assessed using either subjective evaluation methods without relying on instruments or objective evaluation methods using data acquisition equipment. The former relies on the technician's auditory judgment of the noise. This method is simple to operate, requires no equipment, and has good identification accuracy for noise malfunctions with obvious characteristics (such as severe engine misfiring or severe mechanical interference noise). However, this method requires a high level of technical skill from the technician. It is difficult to develop a set of objectively quantifiable methods and procedures for identifying engine noise malfunctions, and it is also difficult to analyze the sources of engine noise malfunctions at a theoretical level.

[0004] The latter method uses data acquisition equipment (such as audio recording equipment and acceleration acquisition equipment) to collect the acoustic and vibration characteristics of a faulty engine, thereby objectively analyzing the engine fault data. However, at present, such equipment usually independently collects acoustic and vibration characteristics along with other engine parameters (such as engine speed), and fault determination typically uses threshold determination methods. Therefore, it is only possible to determine whether the engine is faulty from the surface appearance (such as whether the vibration exceeds the limit at a specified order, whether the operating noise exceeds the limit, etc.), making it difficult to comprehensively analyze the multi-dimensional acoustic and vibration data information of the faulty engine and their correlations. The determination of the source of abnormal noise faults is also relatively primitive (usually only through a single acoustic and vibration order lookup table method), lacking the ability to process and distinguish non-steady-state signal faults (such as intermittent faults, pneumatic component faults) and external interference. Summary of the Invention

[0005] To address the existing problems in engine fault diagnosis, a multi-information fusion diagnostic device and fault diagnosis method for engine sound and vibration are proposed.

[0006] The technical solution of the present invention is as follows: an engine sound-vibration multi-information fusion diagnostic device, comprising an acoustic measurement module, a vibration measurement module, a data acquisition module, a data processing and analysis module, and an interactive display module; the acoustic measurement module and the vibration measurement module are installed on the side of the engine body, and the engine noise signal and vibration signal are collected on site and sent to the data acquisition module;

[0007] The data acquisition module acquires real-time engine noise and vibration signals output by the acoustic measurement module and vibration measurement module, as well as real-time speed signals detected by the engine speed sensor. After processing the signals, the data is sent to the data processing and analysis module.

[0008] Data processing and analysis module: Performs calculations and analysis on the data sent by the data acquisition module, outputs fault category and fault information judgment, and sends it to the interactive display module for information interaction with the testing personnel.

[0009] A method for establishing a data processing and analysis module in an engine acoustic-vibration multi-information fusion diagnostic device, specifically including the following steps:

[0010] 1) Background noise measurement: The acoustic measurement module is installed on the side of the engine block. In an environment where the engine is not running, the background noise time domain data is collected, and then the background noise time domain data is converted into the frequency domain using a time-frequency domain conversion method, and its frequency domain features are extracted.

[0011] 2) Extraction of engine acoustic time domain data: When the generator starts, the acoustic measurement module collects the engine's acoustic time domain data, and combines it with the background noise frequency domain features obtained in step 1) to obtain effective engine acoustic data using a filtering and noise reduction algorithm;

[0012] 3) Engine acoustic-vibration information preprocessing: By using the input speed signal, engine vibration time-domain data, and effective engine acoustic data, the collected engine acoustic-vibration time-domain signal is periodically segmented in the time domain using an adaptive data period segmentation method.

[0013] 4) Multi-layer data feature extraction: For the segmented vibration and acoustic time-domain signals, the data is processed at both the time and frequency domain levels.

[0014] Time-domain data processing: By comparing and analyzing the kurtosis, margin, waveform and peak information of acceleration and acoustic signals, non-strongly periodic signals that are unrelated to the engine mechanical system are analyzed;

[0015] Frequency domain layer data processing: Combine engine speed signal to analyze the order information of vibration and acoustic signals, and distinguish acoustic and vibration signals of different dimensions;

[0016] 5) Multidimensional signal correlation analysis and fault classification:

[0017] Step 4) After separating the data features of the frequency domain layer and the time domain layer, the correlation analysis between acoustic order data, vibration order data and sound-vibration time domain features is performed, and the fault types are classified by looking up tables.

[0018] 6) Fault information determination:

[0019] Step 5) After completing the signal correlation analysis and fault classification, the fault information determination process begins. For faults with significant characteristics, a preset fault threshold method is used for fault determination; for faults without significant characteristics, an artificial intelligence self-learning method based on big data is used for fault information determination.

[0020] Furthermore, the acoustic time-domain data and engine vibration time-domain data of the engine in steps 2) and 3) include normal engine operation data and various engine fault operation data.

[0021] Furthermore, the frequency domain layer data processing involves: using the order frequency domain characteristics related to engine speed order to quickly distinguish and classify mechanical noise generated by different mechanical faults and fluid noise generated by valve body faults in different dimensions; by combining engine speed signals, analyzing the order information of vibration and acoustic signals, extracting and sorting the energy concentration order and the energy of specified orders, and analyzing the vibration and noise introduced by mechanical parts such as engine crankshaft, camshaft, and bearings.

[0022] Furthermore, the engine acoustic-vibration time-domain signal collected in step 3) is periodically segmented in the time domain, and the segmented data is used as training input data for the big data-based artificial intelligence self-learning method.

[0023] A fault diagnosis method for engines based on multi-information fusion of acoustic and vibration data, established by a data processing and analysis module, is proposed. The on-site acoustic measurement module collects background noise and engine operating noise signals during the measurement process, the on-site vibration measurement module collects vibration signals after engine operation, and the speed sensor collects engine speed signals. The data processing and analysis module receives the signals collected by the on-site acoustic measurement module, the on-site vibration measurement module, and the speed sensor through the data acquisition module. Based on the collected data, the data processing and analysis module performs background noise measurement and feature extraction, acoustic information denoising, and acoustic-vibration information preprocessing. After multi-layer data feature extraction, multi-dimensional signal correlation analysis, and fault classification, the fault information is determined, and the fault diagnosis results and corresponding fault characteristics are output.

[0024] The beneficial effects of this invention are as follows: The engine acoustic-vibration multi-information fusion diagnostic device and fault diagnosis method of this invention, due to the establishment of a noise reduction processing flow for the acoustic information of the faulty engine, can remove interference components from the acoustic time-domain data of the faulty engine in non-anechoic environments; the method of this invention rapidly classifies and locates fault features through a lookup table method at three levels: acoustic-vibration time-domain feature correlation, acoustic order data, and vibration order data, thereby quickly determining the possible fault forms of the engine; the use of time-domain / frequency-domain multi-layer signal fault information extraction methods can better handle strongly periodic abnormal noise signals and non-steady-state signals, thus facilitating the identification of fault forms with non-steady-state noise characteristics; for the identification of significant and non-significant fault features, two methods are used respectively: the fault threshold method and artificial intelligence self-learning, thereby ensuring rapid feature recognition while diagnosing and identifying faults that are difficult to identify directly using traditional methods or engine abnormal noise faults under multiple fault coupling states. Attached Figure Description

[0025] Figure 1 This is a structural block diagram of the engine acoustic-vibration multi-information fusion diagnostic equipment for on-site diagnostics according to the present invention;

[0026] Figure 2 This is a flowchart of the data processing and fault diagnosis process of this invention. Detailed Implementation

[0027] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0028] like Figure 1 The diagram shows the structural block of an engine acoustic-vibration multi-information fusion diagnostic device used for on-site diagnostics. The device includes an on-site acoustic measurement module, an on-site vibration measurement module, a data acquisition module, a data processing and analysis module, and an interactive display module. The on-site acoustic measurement module collects background noise and engine operating noise signals during the measurement process, while the on-site vibration measurement module collects vibration signals after engine operation. The data acquisition module collects and processes the signals and information data from each module during the engine measurement process. The data processing and analysis module performs background noise measurement and feature extraction, acoustic noise reduction, and acoustic-vibration information preprocessing based on the collected data, ultimately determining the fault information. The interactive display module then allows for information exchange with the testing personnel.

[0029] As the core component of the equipment, such as Figure 2The data processing and fault diagnosis flowchart shown shows that the basic workflow of the data operation and analysis module includes background noise measurement and feature extraction, acoustic information noise reduction, sound-vibration information preprocessing, multi-layer data feature extraction, multi-dimensional signal correlation analysis, fault classification, and final fault information determination.

[0030] In the background noise measurement and feature extraction process, the background noise time-domain data is first collected, and then the background noise time-domain data is converted into the frequency domain using a time-frequency domain conversion method, and its frequency domain features are extracted.

[0031] In the noise reduction process of acoustic information of faulty engines, effective acoustic fault data is obtained by collecting acoustic time-domain data of the faulty engine and combining it with the background noise features obtained in the previous background noise measurement and feature extraction stages, and then using a filtering noise reduction algorithm.

[0032] The preprocessing flow for the acoustic-vibration information of a faulty engine involves using the input engine speed signal, the time-domain vibration data of the faulty engine, and the effective acoustic fault time-domain data after noise reduction processing. An adaptive data periodic segmentation method is employed to periodically segment the acquired engine acoustic-vibration time-domain signal. This facilitates data processing of the acoustic-vibration information with the minimum possible period in subsequent process analysis.

[0033] The multi-layer data feature analysis process extracts information from the preprocessed acoustic-vibration information at both the time and frequency domains. Time-domain features include kurtosis, margin, waveform, and peak value of the acoustic-vibration signal. Frequency-domain features primarily consist of order-related frequency characteristics related to engine speed (such as acoustic-vibration energy at a specific order, order energy ranking, etc.). This facilitates rapid differentiation and classification of different fault characteristics (such as mechanical noise from mechanical faults and fluid noise from valve body faults) across different dimensions.

[0034] The multidimensional signal correlation analysis and fault classification process classifies faults from three levels: acoustic order data, vibration order data, and correlation comparison of sound-vibration time-domain characteristics. It uses a pre-defined fault lookup table classification method to categorize faults, facilitating the process of fault information determination and result output.

[0035] The fault information determination and result output process includes a fault feature data threshold determination method and a determination method based on big data and artificial intelligence. The former is used to quickly determine fault types with distinctive features, while the latter is used to identify fault types with less obvious features and more complex mechanisms, involving multiple coupled structures. Finally, through an interactive display module, the fault diagnosis results and corresponding fault features are displayed and communicated to the diagnostic personnel, thereby completing the fault diagnosis work.

[0036] The specific implementation of engine acoustic-vibration multi-information fusion diagnostics applied to on-site diagnostics includes:

[0037] 1. On-site Engine Abnormal Noise Fault Collection and Diagnosis Process

[0038] 1.1 Installation of equipment measuring points

[0039] After confirming an abnormal noise problem in the vehicle's engine, the troubleshooting personnel installed the on-site acoustic measurement module and the on-site vibration measurement module on the side of the engine block of the faulty vehicle. These measurement modules were then connected to the data acquisition module via data cables. Simultaneously, the engine speed sensor signal was separated using a splitter and then connected to the data acquisition module via a data cable. This allowed for the separate acquisition of acoustic, vibration, and engine speed signals.

[0040] Depending on the vehicle model and engine type, the acoustic measurement module and vibration measurement module can be arranged using single-point or multi-point methods. Furthermore, other signals (such as coolant temperature and engine intake / exhaust temperatures) can be acquired through optional expansion modules.

[0041] 1.2 Background noise measurement

[0042] Troubleshooting personnel use the equipment's interactive display module and follow the system prompts to use the acoustic measurement module to collect background noise data from the test environment. This allows them to obtain the frequency domain characteristics of the background noise. The environment should be kept as quiet as possible during the test.

[0043] 1.3 Abnormal Noise Fault Acquisition

[0044] After completing the measurement point setup and background noise measurement, the fault diagnosis personnel started the engine and brought it to the required operating conditions according to the specific fault diagnosis requirements. Subsequently, following the instructions of the human-machine interface, they collected the noise and vibration signals of the faulty engine.

[0045] 1.4 Fault Data Display and Fault Diagnosis

[0046] After collecting data on abnormal noise, the data processing and analysis module will automatically analyze the fault characteristics and display the results (including frequency domain fault characteristics, fault mode determination, etc.) on the human-machine interface. This completes the diagnosis of engine abnormal noise faults. Simultaneously, fault information can be saved and exported as needed.

[0047] 2. Equipment data processing and fault diagnosis process:

[0048] During the diagnostic process, data processing and fault diagnosis are both carried out within the data operation and analysis module. Its basic workflow includes background noise measurement and feature extraction, information noise reduction, sound-vibration information preprocessing, multi-layer data feature extraction, multi-dimensional signal correlation analysis and fault classification, and final fault information determination.

[0049] 2.1 Background noise measurement, feature extraction, and noise reduction

[0050] In section 1.2, after the fault repair personnel complete the measurement point setup and background noise collection, the equipment automatically records the time-domain data of the background noise and converts it into a frequency-domain signal to extract its frequency-domain features. In section 1.3, after completing the acoustic characteristics of the engine with abnormal noise faults, the effective fault acoustic data is separated by combining the background noise frequency-domain features and using noise reduction filtering.

[0051] 2.2 Engine acoustic and vibration information preprocessing and multi-layer data feature stripping

[0052] Since most faults during engine operation exhibit periodicity (the data shows significant periodicity in the time domain), it is necessary to segment the data in the time domain using certain methods (such as segmenting by specific periodic time or frequency) to make the representative periodic characteristics more significant. The time-domain segmentation interval can be determined by the time-domain periodic characteristics of engine speed or the original vibration signal. Furthermore, the segmented data can also serve as input data for training decision-making methods based on big data and artificial intelligence.

[0053] After completing the above work, the segmented vibration and acoustic time-domain signals were processed at both the time and frequency domain levels.

[0054] The frequency domain layer primarily analyzes the order information of vibration and acoustic signals by combining engine speed signals. It extracts and sorts the energy of orders with high energy concentration and specified orders. This facilitates the analysis of vibrations and noise introduced by engine mechanical parts (such as crankshafts, camshafts, and bearings).

[0055] The time-domain layer mainly analyzes non-periodic signals (such as belt slip noise, water pump noise, etc.) that are unrelated to the engine mechanical system by comparing and analyzing information such as kurtosis, margin, waveform and peak value of acceleration and acoustic signals.

[0056] 2.3 Multidimensional Signal Correlation Analysis and Fault Classification

[0057] After separating the data features from the frequency domain and time domain, correlation analysis was performed on acoustic order data, vibration order data and acoustic-vibration time domain features respectively. Possible fault types were classified by looking up tables (indicators such as acoustic energy order sorting and vibration energy order sorting).

[0058] 2.4 Fault Information Judgment and Result Output

[0059] After completing the signal correlation analysis and fault classification, the system enters the fault information determination process. This determination involves two methods: a preset fault threshold method and an AI self-learning method based on big data. The former is used for faults with significant characteristics, while the latter is used for faults without significant characteristics. After determining the fault information, the system displays the relevant results (including possible fault types, fault determination criteria, original frequency-time domain signals, etc.) on the human-computer interaction interface and completes the diagnostic work.

[0060] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. An engine acoustic-vibration multi-information fusion diagnostic device, characterized in that, It includes an acoustic measurement module, a vibration measurement module, a data acquisition module, a data processing and analysis module, and an interactive display module; The acoustic measurement module and vibration measurement module are installed on the side of the engine block to collect engine noise and vibration signals on site and send them to the data acquisition module. The data acquisition module acquires real-time engine noise and vibration signals output by the acoustic measurement module and vibration measurement module, as well as real-time speed signals detected by the engine speed sensor. After processing the signals, the data is sent to the data processing and analysis module. Data processing and analysis module: Performs calculations and analysis on the data sent by the data acquisition module, outputs fault category and fault information judgment, and sends it to the interactive display module for information interaction with the testing personnel; The data processing and analysis module performs the following steps: 1) Background noise measurement: The acoustic measurement module is installed on the side of the engine block. In an environment where the engine is not running, the background noise time domain data is collected, and then the background noise time domain data is converted into the frequency domain using a time-frequency domain conversion method, and its frequency domain features are extracted. 2) Extraction of engine acoustic time domain data: When the generator starts, the acoustic measurement module collects the engine's acoustic time domain data, and combines it with the background noise frequency domain features obtained in step 1), and uses a filtering and noise reduction algorithm to obtain effective engine acoustic data; 3) Engine acoustic-vibration information preprocessing: By using the input speed signal, engine vibration time-domain data, and effective engine acoustic data, the collected engine acoustic-vibration time-domain signal is periodically segmented in the time domain using an adaptive data period segmentation method. 4) Multi-layer data feature extraction: For the segmented vibration and acoustic time-domain signals, the data is processed at both the time and frequency domain levels. Time-domain data processing: By comparing and analyzing the kurtosis, margin, waveform and peak information of acceleration and acoustic signals, non-strongly periodic signals that are unrelated to the engine mechanical system are analyzed; Frequency domain layer data processing: Combine engine speed signal to analyze the order information of vibration and acoustic signals, and distinguish acoustic and vibration signals of different dimensions; 5) Multidimensional signal correlation analysis and fault classification: Correlation analysis was conducted on acoustic order data, vibration order data, and acoustic-vibration time-domain characteristics, and fault types were classified using a lookup table method. 6) Fault information determination: Step 5) After completing the signal correlation analysis and fault classification, the fault information determination process begins. For faults with significant characteristics, a preset fault threshold method is used for fault determination; for faults without significant characteristics, an artificial intelligence self-learning method based on big data is used for fault information determination. The frequency domain layer data processing involves: using the order frequency domain characteristics related to engine speed order to quickly distinguish and classify mechanical noise generated by different mechanical faults and fluid noise generated by valve body faults in different dimensions; by combining engine speed signals, analyzing the order information of vibration and acoustic signals, extracting and sorting the energy concentration order and the energy of specified orders, and analyzing the vibration and noise introduced by mechanical parts such as engine crankshaft, camshaft, and bearings; The engine acoustic-vibration time-domain signal collected in step 3) is periodically segmented in the time domain, and the segmented data is used as the training input data for the big data-based artificial intelligence self-learning method. In the adaptive data period segmentation method, the time-domain segmentation interval of the data is determined by the time-domain periodic characteristics of the original vibration signal in the engine speed signal or the engine vibration time-domain data, so as to process the sound-vibration information with the minimum period; the lookup table method classifies possible fault types by acoustic energy order sorting and vibration energy order sorting index.

2. The method for establishing a data processing and analysis module in the engine acoustic-vibration multi-information fusion diagnostic equipment according to claim 1, characterized in that, The acoustic time-domain data and engine vibration time-domain data of the engine mentioned in steps 2) and 3) include normal engine operation data and various engine fault operation data.

3. A fault diagnosis method based on the engine acoustic-vibration multi-information fusion diagnostic equipment as described in claim 1, characterized in that, The on-site acoustic measurement module collects background noise and engine operating noise signals during the measurement process, the on-site vibration measurement module collects vibration signals after engine operation, and the speed sensor collects engine speed signals. The data processing and analysis module receives signals collected by the on-site acoustic measurement module, the on-site vibration measurement module, and the speed sensor through the data acquisition module. The data processing and analysis module performs background noise measurement and feature extraction, acoustic information noise reduction, and sound-vibration information preprocessing based on the collected data. After multi-layer data feature extraction, multi-dimensional signal correlation analysis, and fault classification, the module determines the fault information and outputs the fault diagnosis results and corresponding fault characteristics. In the acoustic-vibration information preprocessing, the time-domain segmentation interval of the data in the adaptive data period segmentation method is determined by the time-domain periodic characteristics of the original vibration signal in the engine speed signal or the engine vibration time-domain data, so as to process the acoustic-vibration information with the minimum period. In the multidimensional signal correlation analysis and fault classification, when the fault type is classified by the lookup table method, the possible fault types are classified by the acoustic energy order ranking and vibration energy order ranking index.

Images