A Power Machinery Fault Diagnosis System

Abstract

The invention discloses a power machinery fault diagnosis system in the field of mechanical engineering. The system comprises an engine sound collection module, an engine vibration collection module, a data collection module, a data processing module, a fault judgment module and a data output module. By inputting time series data such as automobile service life information, fuel consumption data and total mileage of the automobile into a model, a correction coefficient is obtained, and the sound data and vibration information of the engine are corrected. The corrected sound data and engine vibration information of the automobile engine are processed in two dimensions of time and frequency, and the sound time-frequency two-dimensional signal and the vibration time-frequency two-dimensional signal corresponding to the automobile engine are obtained and input into a trained hybrid neural network. According to the output result of the hybrid neural network, it is judged whether the engine fails and the specific fault location, and the power machinery fault is diagnosed, which helps maintenance personnel to find and handle the power machinery fault as early as possible, thereby improving the reliability and maintenance efficiency of the equipment.

Description

Technical Field

[0001] The invention belongs to the field of mechanical engineering, and in particular relates to a power machinery fault diagnosis system. Background Art

[0002] Power machinery is a mechanical device that converts energy in nature into mechanical energy to perform work. From the perspective of technology, power machinery can be divided into several categories, such as steam turbines, gas turbines, water turbines, and internal combustion engines. Mechanical fault diagnosis of equipment or structures is an important issue related to life and production safety. In order to analyze the status of equipment or structures before failure and reduce the risks and maintenance costs caused by failures, more and more signal acquisition and analysis methods are used to monitor the operating status. Traditional signal analysis methods mainly use Fourier transform methods to obtain frequency components in signals. However, the signals during the use of equipment or structures are mostly time-varying signals, and Fourier transform cannot reflect the law of signal frequency changing with time. Based on this background, time-frequency analysis methods came into being. This method can describe the time-varying frequency characteristics of signals and obtain the status of equipment or structures by analyzing the time-frequency distribution of signals.

[0003] During the use of existing equipment or structures, the main frequency may be generated during its normal use, while the real fault or abnormal information is hidden in other non-main frequency characteristics. In this case, the half-peak full-width energy ratio of the main frequency cannot reflect the actual mechanical fault of the equipment or structure. When the fault condition deteriorates to the point of being presented in the form of the main frequency, it has often reached the dangerous stage of rapid damage, which is extremely detrimental to ensuring safe use.

[0004] In order to solve the above problems, a Chinese patent with a publication number of CN112378633B discloses a mechanical fault diagnosis method, which first collects the vibration signal of the monitored object, performs time-frequency distribution processing on the collected vibration signal, and obtains the time-frequency distribution result of the vibration signal; then, in the frequency domain, the time-frequency distribution result of the vibration signal is divided into multiple frequency domain segments, and in the time domain, according to the running speed of the monitored object, the time-frequency distribution result of the vibration signal is divided into multiple time domain segments, so that the time-frequency distribution result of the vibration signal will form multiple grids; then, the amplitude density in each grid of the time-frequency distribution result of the vibration signal is calculated; if the amplitude density in at least one grid exceeds a preset threshold, it is determined that the monitored object is abnormal.

[0005] The mechanical fault diagnosis method performs mechanical fault diagnosis based on time-frequency analysis, which can timely reflect the real mechanical fault situation of the equipment or structure and ensure the safe use of the equipment or structure. However, the method relies on a preset amplitude density threshold to determine whether there is an abnormality. Selecting a suitable threshold may be a challenge, and different equipment or environments may require different threshold settings. False alarms and missed alarms may occur, resulting in misjudgment or omission of real fault conditions. The engine in the automotive power machinery is an important part of the power machinery fault diagnosis. When the method is used to diagnose the engine fault, the engine vibration intensity increases due to the high load state when the car is driving, which is very likely to exceed the set threshold, resulting in misjudgment or false alarm, and ultimately affecting the use of power machinery equipment. Therefore, this scheme proposes a power machinery fault diagnosis system to diagnose the engine fault, avoid misjudgment or omission of real fault conditions caused by the preset threshold, reduce downtime, improve production efficiency, help maintenance personnel to detect and handle power machinery faults as early as possible, and improve equipment reliability and maintenance efficiency. Summary of the invention

[0006] In order to solve the above problems, the present invention provides a power machinery fault diagnosis system to reduce downtime, improve production efficiency, help maintenance personnel to early detect and handle power machinery faults, and improve equipment reliability and maintenance efficiency.

[0007] In order to achieve the above object, the technical solution of the present invention is as follows: a power machinery fault diagnosis system, comprising

[0008] The engine sound collection module is installed in the engine compartment and is used to collect the sound data of the car engine;

[0009] The engine vibration collection module is installed at the connection between the engine and the frame to collect the vibration information of the automobile engine;

[0010] Data collection module: used to collect information on the age of the car, fuel consumption data and total mileage of the car;

[0011] Data processing module: input the vehicle age information, fuel consumption data and total mileage of the vehicle into the trained time series model to obtain the correction coefficient;

[0012] Then, the sound data and vibration information of the engine are obtained and the correction coefficient obtained by the time series model is used to correct the sound data and vibration information of the engine; the correction coefficient calculation formula is as follows:

[0013]

[0014] Wherein, T is the correction coefficient; n is the age of the car, in years; m1 is the current fuel consumption data, m2 is the standard fuel consumption data of the car, in L / 100km; v1 is the actual total mileage of the car, v2 is the estimated total mileage of the car corresponding to the age of the car, in kilometers;

[0015] When making corrections, the sound data and the vibration information are respectively multiplied by the correction coefficients to obtain the corrected sound data and vibration information;

[0016] The corrected sound data and engine vibration information of the automobile engine are preprocessed and processed in two dimensions to obtain the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal corresponding to the automobile engine; when performing the two-dimensional time-frequency processing, the short-time Fourier transform method is used for calculation, and the short-time Fourier transform calculation formula is as follows:

[0017] STFT(t,f)=∫w(τ)x(t-τ)e^(-j2πfτ)dτ

[0018] Where t represents time, f represents frequency, w(τ) represents the window function, x(t-τ) represents the signal value at time t minus τ, 2πfτ represents the phase shift, and j is the imaginary unit;

[0019] Fault judgment module: used to obtain the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal of the automobile engine, input the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal into the trained hybrid neural network, and judge whether the engine has a fault and the specific fault location according to the output results of the hybrid neural network;

[0020] Data output module: used to obtain data from the data collection module, data processing module and fault judgment module, and output it to the terminal.

[0021] The principle and beneficial effects of the above scheme:

[0022] Since the automobile engine is part of the automobile power mechanical system, various vibrations and sounds will be generated during the operation of the engine. These vibrations and sounds may be caused by factors such as the mechanical operation inside the engine, the combustion process and the cylinder pressure, or they may be caused by the failure of the power machinery. Therefore, by monitoring the sound data and vibration information of the engine, the operating status and potential failure conditions of the power machinery can be indirectly reflected.

[0023] The system uses a time series model for data processing, which can effectively process time series data and extract useful information from it. By inputting time series data such as the age of the car, fuel consumption data, and total mileage of the car into the model, a correction coefficient can be obtained to correct the sound data and vibration information of the engine. This method can not only improve the data processing efficiency, but also greatly reduce the error rate.

[0024] The two-dimensional time-frequency signals of sound and vibration are used as the input of the hybrid neural network, which retains the information of the two dimensions of time and frequency for the scene data with poor structure, and has better effect than the traditional single-dimensional sound or vibration recognition method. According to the output results of the hybrid neural network, it is judged whether the engine has a fault and the specific fault location. This fault judgment method based on the hybrid neural network has high accuracy and reliability, and can effectively identify various types and locations of engine faults. In addition, this scheme can be used for autonomous judgment of some common engine faults, which can simply, quickly and reliably judge automobile engine faults, and facilitate timely troubleshooting.

[0025] The system can continuously collect and process data, monitor the working status and vibration of the engine in real time, promptly detect fault risks, improve the timeliness of fault diagnosis, and improve the accuracy and reliability of fault diagnosis. The system can diagnose whether the engine will fail based on the continuous and corrected engine vibration data, take repair and maintenance measures in advance, and reduce the impact of the failure on mechanical equipment and production lines.

[0026] In general, this system infers possible fault conditions based on current vibration data, sound signals and existing fault modes, avoiding misjudgment or missing real fault conditions due to preset thresholds. Possible fault contents include fault types, such as insufficient engine cylinder pressure and cylinder misfire. At the same time, this power machinery fault diagnosis system has multiple advantages such as real-time, accuracy, cost-effectiveness and automation, which reduces downtime, improves production efficiency and helps to achieve a safer working environment. It helps maintenance personnel to detect and deal with power machinery faults as early as possible, improving the reliability and maintenance efficiency of equipment.

[0027] Furthermore, in the fault judgment module, when training the hybrid neural network, several groups of sound time-frequency two-dimensional signals and vibration time-frequency two-dimensional signals and corresponding engine fault types are obtained; each sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal and engine fault type are input into AlexNet for training to obtain a trained hybrid neural network.

[0028] Beneficial effects: Using deep learning models such as AlexNet for training can greatly improve training efficiency. These models can automatically learn more complex features when processing large amounts of data, thereby reducing reliance on manual feature engineering. The hybrid neural network combines the sound time-frequency two-dimensional signal and the vibration time-frequency two-dimensional signal and the corresponding engine fault type, which can more comprehensively consider the engine's fault characteristics, thereby improving the accuracy of fault judgment.

[0029] Furthermore, the preprocessing includes but is not limited to filtering operations and denoising operations. The filtering operation is used to remove high-frequency noise or unnecessary frequency components in the vibration signal, and the denoising operation is used to further reduce random interference in the vibration data.

[0030] Benefits: Filtering and denoising operations help improve the quality, reliability, and interpretability of vibration data, making it more suitable for use in engineering, scientific, and machine learning applications, thereby increasing the accuracy and feasibility of data analysis.

[0031] Furthermore, when collecting the total mileage of the car, the total mileage of the car is obtained through the car OBD system.

[0032] Beneficial effects: The OBD system can monitor and record the vehicle's driving data in real time, including the total mileage. This means that users can understand the vehicle's driving status at any time without any additional operation or maintenance.

[0033] Furthermore, it also includes a user interface module, which is used to provide a user interface so that the operator can monitor the status of the system, view diagnostic results, access historical data and perform necessary operations; it also includes a real-time notification module, which is used to provide real-time notifications through communication methods including but not limited to text messages and emails, and immediately notify relevant personnel or maintenance teams when abnormal conditions or faults are detected.

[0034] Beneficial effects: The real-time notification module can enhance the practicality and effectiveness of the power machinery fault diagnosis system, ensure that problems are handled in a timely manner, improve equipment availability and production efficiency, and reduce potential losses and safety risks. The user interface module allows operators to monitor the status of the system in a visual way. Operators can view the system status, diagnostic results and access historical data in real time, making it easier to understand the performance status of the equipment. It provides operators with an intuitive way to monitor, control and analyze the system, which helps improve equipment management efficiency, performance visualization and decision support, thereby enhancing the practicality of the system and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 Schematic diagram of a power machinery fault diagnosis system according to an embodiment of the present invention.

[0036] Figure 2 This is a module connection diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following is further described in detail through specific implementation methods:

[0038] Embodiment 1

[0039] The embodiment is basically as shown in the attached Figure 1 and attached Figure 2 As shown:

[0040] A power machinery fault diagnosis system, comprising:

[0041] The engine sound collection module is installed in the engine compartment and is used to collect the sound data of the car engine.

[0042] The engine vibration collection module is located at the connection between the engine and the frame and is used to collect vibration information of the vehicle engine.

[0043] Data collection module: used to collect information on the age of the car, fuel consumption data and the total mileage of the car; when collecting the total mileage of the car, the total mileage of the car is obtained through the car's OBD system.

[0044] Data processing module: used to obtain the vehicle's age information, fuel consumption data and total mileage, and input the vehicle's age information, fuel consumption data and total mileage into the trained time series model to obtain the correction coefficient.

[0045] Then, the sound data and vibration information of the engine are obtained and the correction coefficient obtained by the time series model is used to correct the sound data and vibration information of the engine; the correction coefficient calculation formula is as follows:

[0046]

[0047] Among them, T is the correction coefficient; n is the age of the car, in years; m1 is the current fuel consumption data, m2 is the standard fuel consumption data of the car, in L / 100km; v1 is the actual total mileage of the car, v2 is the estimated total mileage of the car corresponding to the age of the car, in kilometers.

[0048] When making corrections, the sound data and vibration information are respectively multiplied by the correction coefficients to obtain the corrected sound data and vibration information.

[0049] The corrected sound data and engine vibration information of the automobile engine are preprocessed and processed in two dimensions to obtain the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal corresponding to the automobile engine; during the preprocessing, the filtering operation is used to remove high-frequency noise or unnecessary frequency components in the vibration signal, and the denoising operation is used to further reduce random interference in the vibration data.

[0050] When performing time-frequency two-dimensional processing, the short-time Fourier transform method is used for calculation. The short-time Fourier transform calculation formula is as follows:

[0051] STFT(t,f)=∫w(τ)x(t-τ)e^(-j2πfτ)dτ

[0052] Wherein, t represents time, f represents frequency, w(τ) represents window function, x(t-τ) represents the signal value at time t minus τ, 2πfτ represents phase shift, and j is an imaginary unit.

[0053] Fault judgment module: used to obtain the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal of the automobile engine, input the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal into the trained hybrid neural network, and judge whether the engine has a fault and the specific fault location according to the output results of the hybrid neural network.

[0054] When training the hybrid neural network, several groups of sound time-frequency two-dimensional signals and vibration time-frequency two-dimensional signals and corresponding engine fault types are obtained; each sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal and engine fault type are input into AlexNet for training to obtain a trained hybrid neural network.

[0055] Data output module: used to obtain data from the data collection module, data processing module and fault judgment module, and output it to the terminal.

[0056] The specific implementation process is as follows:

[0057] Step 1: Data collection:

[0058] First, the engine sound collection module is installed in the engine compartment to collect the sound data of the car engine. At the same time, the engine vibration collection module is installed at the connection between the engine and the frame to collect the vibration information of the car engine. These modules can be connected to the data collection module by wire or wirelessly.

[0059] Step 2: Data integration:

[0060] The data collection module not only collects the sound data and vibration information of the engine, but also obtains the total mileage of the car through the car's OBD system. It also collects information such as the car's age and fuel consumption data, integrates these data and uses them for subsequent data processing and analysis.

[0061] Step 3: Data correction:

[0062] The data processing module inputs the vehicle age information, fuel consumption data and total mileage data into the trained time series model to obtain the correction coefficient. Then, the data processing module obtains the engine sound data and vibration information, uses the time series model to obtain the correction coefficient, and uses this correction coefficient to correct the sound data and vibration information.

[0063] The correction coefficient calculation formula is as follows:

[0064]

[0065] Among them, T is the correction coefficient; n is the age of the car, in years; m1 is the current fuel consumption data, m2 is the standard fuel consumption data of the car, in L / 100km; v1 is the actual total mileage of the car, v2 is the estimated total mileage of the car corresponding to the age of the car, in kilometers.

[0066] When making corrections, the sound data and vibration information are respectively multiplied by the correction coefficients to obtain the corrected sound data and vibration information.

[0067] Step 4: Data preprocessing and time-frequency two-dimensional processing:

[0068] The corrected sound data and vibration information of the automobile engine are preprocessed and processed in two dimensions. The preprocessing includes filtering and denoising operations to remove unnecessary frequency components and random interference. The two-dimensional time-frequency processing uses the short-time Fourier transform method to calculate to obtain the two-dimensional time-frequency signal of sound and the two-dimensional time-frequency signal of vibration. These signals will be used for subsequent fault judgment and analysis.

[0069] In the two-dimensional time-frequency processing, the window function selection in the time-frequency analysis includes but is not limited to the Hamming window, the Hamming window and the Hanning window. The selection of the window function will balance the leakage of signal energy and resolution, so that the window function can effectively suppress frequency aliasing and retain the detailed characteristics of the signal. For example, for applications that need to reduce spectrum leakage, you can choose the Hamming window or the Hamming window; for applications that need to retain signal spectrum information, you can choose the Hamming window or the Hanning window.

[0070] When performing time-frequency two-dimensional processing, the short-time Fourier transform method is used for calculation. The short-time Fourier transform calculation formula is as follows:

[0071] STFT(t,f)=∫w(τ)x(t-τ)e^(-j2πfτ)dτ

[0072] Wherein, t represents time, f represents frequency, w(τ) represents window function, x(t-τ) represents the signal value at time t minus τ, 2πfτ represents phase shift, and j is an imaginary unit.

[0073] Step 5: Fault diagnosis and output:

[0074] The fault judgment module is used to obtain the sound time-frequency two-dimensional signal and the vibration time-frequency two-dimensional signal, and they are input into the trained hybrid neural network. During the training process, the hybrid neural network model learns the relationship between different fault types and sound and vibration signals, so it can automatically determine whether the engine has a fault and identify the specific fault type. The hybrid neural network consists of multiple subnetworks, each of which corresponds to a specific fault type. These subnetworks include different types of neural networks, such as multilayer perceptrons, radial basis function networks, long short-term memory networks (LSTM), etc. Specifically, the hybrid neural network learns the characteristics of various engine sound and vibration signals during the training process and associates these characteristics with the corresponding fault types. Once a new sound and vibration signal is received, the network can determine whether a fault exists and determine the type of fault based on its learned knowledge.

[0075] The specific values ​​of the parameters in the hybrid neural network depend on the structure of the neural network and the training method. These parameters include but are not limited to the learning rate, number of iterations, regularization strength, neural network structure, optimizer, and batch size. For example, for the diagnosis of engine overheating faults, the following parameter values ​​can be set:

[0076] Learning rate: 0.01; number of iterations: 100; regularization strength: 0.01; neural network structure: one hidden layer with 20 nodes; optimizer: Adam and batch size: 10.

[0077] Finally, the data output module obtains all collected, processed and judged data and outputs them to the terminal for users or engineers to view and analyze.

[0078] For example, when the engine is overheated, the temperature of the cylinder head, engine body and other parts will rise, the structural deformation will increase, resulting in unstable engine operation, abnormal vibration and noise. The system obtains the vibration and noise generated by the engine, which usually increases with the increase of engine temperature. The power machinery fault diagnosis system can diagnose the engine overheating fault, output the diagnostic information, and remind relevant personnel to conduct timely inspection or maintenance, thereby improving the reliability and safety of the engine.

[0079] For example, when a bearing burns, it will emit a "squeaking" metal friction sound. Abnormal valve noise occurs at the top of the engine due to excessive valve clearance. In addition, poor valve lubrication, abnormal wear of the valve rocker arm, etc. can also cause abnormal valve noise, which will emit a continuous and rhythmic "tick-tick" sound. The system can obtain the "squeaking" metal friction sound and "tick-tick" sound of the engine. The power machinery fault diagnosis system can diagnose power machinery faults such as bearing burnout and excessive valve clearance.

[0080] Embodiment 2

[0081] The present embodiment differs from the above-mentioned embodiments in that: it also includes a user interface module, which is used to provide a user interface so that the operator can monitor the status of the system, view diagnostic results, access historical data and perform necessary operations; it also includes a real-time notification module, which is used to provide real-time notifications through communication methods including but not limited to text messages and emails, and immediately notify relevant personnel or maintenance teams when abnormal conditions or faults are detected.

[0082] The specific implementation process is as follows: Work closely with end users or operators to collect their needs and expectations. Understand the information and functions they want to obtain from the user interface in order to meet their work needs. Based on the needs analysis, design the appearance and layout of the user interface. Consider using intuitive graphic elements, charts, tables, buttons, menus, etc. to enable operators to easily navigate and understand the interface.

[0083] Develop the front-end portion of the user interface, typically using web technologies (e.g., HTML, CSS, JavaScript) or application development frameworks (e.g., React, Angular, Vue.js). Ensure that the interface works responsively on different devices, including computers, tablets, and phones. Establish connections between the user interface and the back-end of the system to obtain real-time data and perform actions. This may involve using technologies such as APIs or WebSockets to communicate with the back-end.

[0084] Display real-time data on the interface, including vibration information and oil volume estimates. Present data in a visual form using charts, real-time images, or data tables. Implement the ability to query historical data, allowing operators to select a specific time range to view historical vibration and oil volume data. Provide search and filter options to facilitate data export or analysis. Present the system's diagnostic results to the operator in an easy-to-understand manner. This can include warnings, error messages, or graphical fault indications.

[0085] If required, implement operational and control functions that operators can perform. For example, remotely rebooting equipment, setting alarm thresholds, or initiating maintenance procedures. Ensure that the user interface module has appropriate security measures, including authentication and access control, to prevent unauthorized access and potential security risks.

[0086] You need to choose a suitable communication method, such as SMS, email, instant messaging, mobile app notification, etc., to notify relevant personnel in real time. The communication method should be determined based on the recipient's preference and availability. Based on the selected communication method, configure the notification interface or module so that the system can send notifications through this interface.

[0087] Define rules for when notifications are triggered. Usually, notification rules are related to anomalies or failures detected by the system. Create a list of recipients with contact information of relevant persons who need to be notified when anomalies occur. These persons can be maintenance team members, engineers, managers, or other relevant parties. The real-time notification module should constantly monitor the status and events of the system so that anomalies or failures can be detected immediately. This can be achieved by periodically querying the system status or subscribing to events.

[0088] When the system detects an abnormal situation or failure according to the notification rules, the notification module will trigger a notification and send a notification message to the relevant persons in the recipient list. The notification message should include a description of the problem, a timestamp, and possible recommended actions.

[0089] The above is only an embodiment of the present invention, and the common knowledge such as the known specific structure and / or characteristics in the scheme is not described in detail here. It should be pointed out that for those skilled in the art, several deformations and improvements can be made without departing from the structure of the present invention, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A power machinery fault diagnosis system, characterized in that: include The engine sound collection module is installed in the engine compartment and is used to collect the sound data of the car engine; The engine vibration collection module is installed at the connection between the engine and the frame to collect the vibration information of the automobile engine; Data collection module: used to collect information on the age of the car, fuel consumption data and total mileage of the car; Data processing module: input the vehicle age information, fuel consumption data and total mileage of the vehicle into the trained time series model to obtain the correction coefficient; Then, the sound data and vibration information of the engine are obtained and the correction coefficient obtained by the time series model is used to correct the sound data and vibration information of the engine; the correction coefficient calculation formula is as follows: Wherein, T is the correction coefficient; n is the age of the car, in years; m1 is the current fuel consumption data, m2 is the standard fuel consumption data of the car, in L / 100km; v1 is the actual total mileage of the car, v2 is the estimated total mileage of the car corresponding to the age of the car, in kilometers; When making corrections, the sound data and the vibration information are respectively multiplied by the correction coefficients to obtain the corrected sound data and vibration information; The corrected sound data and engine vibration information of the automobile engine are preprocessed and processed in two dimensions to obtain the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal corresponding to the automobile engine; when performing the two-dimensional time-frequency processing, the short-time Fourier transform method is used for calculation, and the short-time Fourier transform calculation formula is as follows: STFT(t,f)=∫w(τ)x(t-τ)e^(-j2πfτ)dτ Where t represents time, f represents frequency, w(τ) represents the window function, x(t-τ) represents the signal value at time t minus τ, 2πfτ represents the phase shift, and j is the imaginary unit; Fault judgment module: used to obtain the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal of the automobile engine, input the sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal into the trained hybrid neural network, and judge whether the engine has a fault and the specific fault location according to the output results of the hybrid neural network; Data output module: used to obtain data from the data collection module, data processing module and fault judgment module, and output it to the terminal.

2. The power machinery fault diagnosis system according to claim 1, characterized in that: In the fault judgment module, when training the hybrid neural network, several groups of sound time-frequency two-dimensional signals and vibration time-frequency two-dimensional signals and corresponding engine fault types are obtained; each sound time-frequency two-dimensional signal and vibration time-frequency two-dimensional signal and engine fault type are input into AlexNet for training to obtain a trained hybrid neural network.

3. The power machinery fault diagnosis system according to claim 2, characterized in that: The preprocessing includes but is not limited to filtering operations and denoising operations. The filtering operation is used to remove high-frequency noise or unnecessary frequency components in the vibration signal, and the denoising operation is used to further reduce random interference in the vibration data.

4. The power machinery fault diagnosis system according to claim 3, characterized in that: When collecting the total mileage of a car, the total mileage of the car is obtained through the car's OBD system.

5. The power machinery fault diagnosis system according to claim 4, characterized in that: It also includes a user interface module, which is used to provide a user interface so that the operator can monitor the status of the system, view diagnostic results, access historical data and perform necessary operations; it also includes a real-time notification module, which is used to provide real-time notifications through communication methods including but not limited to text messages and emails, and immediately notify relevant personnel or maintenance teams when abnormal conditions or faults are detected.

Images