Gas turbine fault diagnosis system based on multiple sensors and control method

Abstract

The invention discloses a gas turbine fault diagnosis system based on multiple sensors and a control method, the gas turbine fault diagnosis system comprises a data acquisition device, a data core service system and a monitoring center, and sensor nodes of the data acquisition device acquire dynamic signals of multiple parts of a gas turbine; a data acquisition module of the data acquisition equipment is connected with a sensor node, receives a dynamic signal and transmits the dynamic signal to a data receiving module; the data receiving module transmits the dynamic signal to the data processing module, the data processing module extracts each characteristic parameter through a characteristic parameter extraction algorithm, performs standardized score calculation on each dynamic signal characteristic data point acquired in real time, and sets a single score threshold value corresponding to each characteristic parameter; the standardized scores of all the features are fused, a comprehensive abnormal scoring formula is obtained, and a comprehensive scoring threshold value is set; when the characteristic data point of the current dynamic signal exceeds the comprehensive score threshold value or the single score threshold value corresponding to each characteristic parameter, the data processing module sends an abnormal signal to the alarm module.

Description

Technical Field

[0001] The present invention relates to the fields of mechanical engineering and power engineering, and in particular to a multi-sensor based gas turbine fault diagnosis system and control method. Background Art

[0002] Vibration monitoring systems are technical systems used to detect, record and analyze the vibration of equipment or structures. They are widely used in industry, construction, aviation, automobiles and other fields to improve the reliability of equipment, extend its service life, prevent failures and optimize maintenance strategies.

[0003] The existing vibration monitoring system uses a velocity sensor to measure the amplitude and frequency of vibration, converts the collected signal from an analog signal to a digital signal and transmits it to a data processing module. The data processing module analyzes the vibration signal and issues an alarm if it exceeds a predetermined threshold.

[0004] The above-mentioned traditional vibration monitoring system mainly has the following deficiencies:

[0005] 1. Single sensor type: Most systems use only a single type of sensor, which cannot fully monitor the vibration of the gas turbine, resulting in deviations in monitoring data, thus affecting the accuracy of fault diagnosis;

[0006] 2. Poor real-time data processing: Traditional systems have complex algorithms and slow processing speeds, and are unable to provide early warning information in a timely manner, which may result in failure to detect and handle faults in a timely manner;

[0007] 3. The traditional data processing module does not extract characteristic parameters from the data, and is unable to comprehensively and accurately evaluate the operating status of the gas turbine. The fault identification accuracy is not high, and there is no standardized evaluation system, so the evaluation accuracy of the test results is not high. Summary of the invention

[0008] To solve the above problems, the present invention provides a gas turbine fault diagnosis system and control method based on multiple sensors with high precision, strong real-time performance and good scalability. The system realizes comprehensive monitoring of the vibration conditions of the gas turbine by introducing various types of sensors, thereby improving the accuracy and reliability of monitoring. The data processing module of the system includes a feature parameter extraction algorithm, which extracts each feature parameter through the feature parameter extraction algorithm, calculates the mean and standard deviation of each feature parameter, realizes comprehensive and detailed analysis and anomaly detection of vibration data, and improves the real-time performance and accuracy of data processing. The control method of the present invention designs an anomaly detection method, which includes single feature detection and comprehensive scoring detection, so as to more accurately evaluate the monitoring results and improve the accuracy of fault identification.

[0009] The technical solution adopted by the present invention is:

[0010] The present invention provides a gas turbine fault diagnosis system based on multiple sensors, comprising a data acquisition device, a data core service system and a monitoring center connected in sequence, wherein the data acquisition device is composed of a sensor node and a data acquisition module; the data core service system is composed of a data receiving module, a data processing module and an alarm module, the data receiving module, the data processing module and the alarm module are connected in sequence, and the alarm module is connected to the data processing module;

[0011] The sensor nodes of the data acquisition equipment include multiple types of sensors, which are installed in different parts of the gas turbine to collect dynamic signals from multiple parts of the gas turbine in real time and monitor the status of the gas turbine;

[0012] The data acquisition module of the data acquisition equipment is used to connect to the sensor nodes, receive dynamic signals and transmit them to the data receiving module;

[0013] The data receiving module of the data core service system receives the dynamic signal and transmits it to the data processing module. The data processing module includes a feature parameter extraction algorithm, which extracts the frequency peak, root mean square value and kurtosis feature parameters through the feature parameter extraction algorithm, calculates the mean and standard deviation of each feature parameter, and performs standardized score calculation for each dynamic signal feature data point collected in real time, and sets the standardized score threshold of each feature parameter, which is called a single scoring threshold; the data processing module also fuses the standardized scores of each feature parameter, obtains the fused comprehensive abnormal scoring formula and sets the comprehensive scoring threshold; if the feature data point of the current dynamic signal exceeds the comprehensive scoring threshold or the single scoring threshold corresponding to each feature parameter, the data processing module sends an abnormal signal to the alarm module;

[0014] The monitoring center is connected to the alarm module, which receives abnormal signals and sends alarm information to the monitoring center for display to remind operation and maintenance personnel.

[0015] Furthermore, the sensor nodes of the data acquisition equipment include multiple vibration sensors and sound sensors, which collect vibration signals and sound signals from multiple parts of the gas turbine in real time. They are installed on the bearings and gearboxes of the gas turbine. Each gas turbine is installed with 6-10 sensor nodes, and each node contains a vibration sensor and a sound sensor to ensure the stability and accuracy of signal transmission.

[0016] Furthermore, the data acquisition module of the data core service system supports multiple device types, including 8 channels and 16 channels. The sampling rate and resolution of each channel are consistent. The data receiving module can support multiple quantities and models of acquisition devices and dynamic IP access; thus, the system supports multiple quantities and models of acquisition devices, ensuring the stability and reliability of data transmission.

[0017] Furthermore, the data core service system also includes a storage module, which is connected to the alarm module and the data processing module. The storage module saves the collected dynamic signal data to a local disk, and saves the original data for a period of time before and after the alarm when an alarm occurs.

[0018] The present invention also provides a control method based on the above-mentioned gas turbine fault diagnosis system, comprising the following steps:

[0019] S1: The data acquisition module of the data acquisition device receives the dynamic signal data collected by the sensor node and transmits it to the data receiving module, and the data receiving module transmits the dynamic signal data to the data processing module;

[0020] S2: The data processing module processes and analyzes the dynamic signal data in real time. By calculating the physical value of the dynamic signal data, the original collected data is converted into a value with actual physical meaning for subsequent signal analysis and processing. By performing fast Fourier transform on the dynamic signal data, the time domain signal is converted into a frequency domain signal for analyzing the frequency components in the dynamic signal.

[0021] S3: The data processing module extracts the characteristic parameters of the dynamic signal through the characteristic parameter extraction algorithm, and evaluates the operating status of the gas turbine by analyzing the extracted characteristic parameters to identify whether an abnormal situation occurs;

[0022] S4: When the data processing module identifies an abnormal situation, it sends an abnormal signal to the alarm module.

[0023] Furthermore, in step S1, the specific communication steps of the data acquisition module, the data receiving module and the data processing module of the data core service system are as follows:

[0024] S11: The data acquisition module receives the dynamic signal of the gas turbine collected in real time by the sensor node, and sends a detection packet 80 to the data receiving module after performing signal processing on the dynamic signal. The detection packet 80 includes the device number and type of the data acquisition module;

[0025] S12: The data receiving module obtains the current device number and type from the detection packet 80 in real time, and obtains the data processing thread corresponding to the current device number from the cache. If the device number exists, the device IP address is updated. If the device number does not exist, a corresponding data processing thread is created according to the current device number and type, and the device number and type as well as the created data processing thread information are stored in the QHASH structure for management, so as to facilitate subsequent data flow;

[0026] S13: The data receiving module sends the detection packet 80 to the data processing thread, and the data processing module sends the configuration parameters 81 of the data acquisition module to the data receiving module. The data processing module sends the configuration parameters 81 of the data acquisition device to the data receiving module according to the parameter configuration. The configuration parameters 81 include the acquisition frequency, sampling points and number of channels, and are forwarded by the data receiving module to the data acquisition module, waiting for the confirmation of the data acquisition module.

[0027] S14: After the data acquisition module receives the device configuration parameters 81, the data acquisition module sends the same confirmation message to the data receiving module. After receiving the message, the data receiving module forwards the data request message to the data acquisition module to inform the data receiving module that it is ready. The data acquisition module starts to collect sensor data, and the sensor data is integrated into raw data 82. The data acquisition module transmits the raw data 82 to the data receiving module, and the data receiving module then forwards the raw data 82 to the data processing module.

[0028] The data receiving module of the present invention adopts multi-threaded data processing technology. Through the above-mentioned communication steps, it can realize dynamic IP access and QHASH structure management, realize dynamic access and efficient data flow of multiple and multi-type acquisition devices. Compared with the prior art, the present invention has obvious advantages in the flexibility and efficiency of data transmission and processing, and can adapt to complex field environments and diverse equipment types.

[0029] Furthermore, in step S3, the dynamic signal includes vibration data and sound data, which are collected by vibration sensors and sound sensors installed at different parts of the gas turbine, and the application of the characteristic parameter extraction algorithm specifically includes the following steps:

[0030] S31: Calculate the peak value of the vibration signal in each data acquisition window of the vibration data. The peak value is the maximum intensity of the signal. The peak value calculation formula of the vibration data is:

[0031] Peak Value = max(|x i振 |)

[0032] Where Peak Value is the peak value of the vibration signal in each data acquisition window, x i振 is the sample value of the vibration signal in each data acquisition window;

[0033] S32: Calculate the root mean square value of the vibration signal in each data acquisition window of the vibration data. The root mean square value reflects the average strength of the signal. The calculation formula of the root mean square value of the vibration signal is:

[0034]

[0035] Where RMS Value is the root mean square value of the vibration signal in each data acquisition window, N is the number of samples in each data acquisition window, and x i振 is the sample value of the vibration signal in each data acquisition window;

[0036] S33: Calculate the kurtosis of the vibration signal in each data acquisition window of the vibration data to identify the impact component in the signal. The calculation formula of the kurtosis of the vibration signal is:

[0037]

[0038] Where Kurtosis is the kurtosis of the vibration signal in each data acquisition window of the vibration data, N is the number of samples in the window, and x i振 is the sample value of the vibration signal in each data acquisition window, μ 振 is the mean of the signal, σ 振 is the standard deviation of the signal;

[0039] in, Where N is the number of samples in each data acquisition window, x i is the i-th vibration signal sample value;

[0040]

[0041] Where N is the number of samples in each data acquisition window, x i is the i-th vibration signal sample value, μ 振 is the average value;

[0042] S34: Calculate the root mean square value of the sound signal in each data collection window of the sound data to reflect the sound intensity characteristics. The calculation formula of the root mean square value of the sound signal is:

[0043]

[0044] Where Loudness is the root mean square value of the sound signal, x i声 is the sample value of the sound signal in each data acquisition window, and N is the number of samples in each data acquisition window;

[0045] S35: Statistical analysis and anomaly detection are performed on the vibration data and sound data, a normal state baseline is established, vibration signal and sound signal data under normal operating conditions are collected, and the mean and standard deviation of each characteristic vibration signal and sound signal are calculated. Each characteristic vibration signal and sound signal includes the peak value, root mean square value, kurtosis of the vibration signal and the root mean square value of the sound signal. A standardized score is calculated for each vibration characteristic data point and sound intensity data point collected in real time. During the calculation, the mean and standard deviation of the peak value of the vibration signal, the mean and standard deviation of the root mean square value, and the mean and standard deviation of the kurtosis are respectively substituted into the formula, and the mean and standard deviation of the root mean square value of the sound signal are substituted into the formula. The standardized score calculation formula for each characteristic is specifically as follows:

[0046]

[0047]

[0048] In the formula, Z vi is the standardized score value of the vibration signal of each feature at present, Z s is the standardized score value of the current sound signal RMS value, x vi is the current vibration characteristic data point, specifically the peak value, root mean square value and kurtosis of the vibration signal collected in real time in each data collection window, which are obtained by the formulas of steps S31, S32 and S33 respectively, x s is the current sound intensity data point, is the root mean square value of the sound signal collected in real time in each data collection window, obtained by the formula of the specific step S34, μ vi It is expressed as the mean of three eigenvalues, namely the peak value, root mean square value and kurtosis of the vibration signal, σ vi It is expressed as the standard deviation of three eigenvalues, namely the peak value, root mean square value and kurtosis of the vibration signal, μ s is the mean of the root mean square value of the sound signal, σ s is the standard deviation of the root mean square value of the sound signal;

[0049] S36: Perform anomaly detection on the calculated standardized score. The anomaly detection includes single feature detection and comprehensive score detection. The single feature detection is to set thresholds for the standardized score values ​​of the vibration signal and the sound signal of each feature, respectively, which are single score thresholds corresponding to each feature parameter. The vibration signal and the sound signal of each feature include the peak value, the root mean square value, the kurtosis of the vibration signal and the root mean square value of the sound signal. If the standardized score value Z of the vibration signal of any feature is vi Or the normalized score Z of the root mean square value of the sound signal s If the corresponding single score threshold is exceeded, the feature is considered abnormal;

[0050] Comprehensive scoring detection is to fuse the standardized scores of the vibration signals and sound signals of each feature, obtain the fused comprehensive abnormality scoring formula and set the comprehensive scoring threshold, such as Z combined If the value exceeds the threshold, it is determined that the machine is abnormal, and the Z obtained in step S35 is vi , that is, the peak value, RMS value and standardized score value of the vibration signal are substituted into the formula to find the average, and the RMS value Z of the sound signal is s Substituting the standardized score into the formula, the calculation formula for the integrated score value is:

[0051]

[0052] In the formula, Z combined is the comprehensive abnormality score after fusion, Z vi is the standardized score value of the vibration signal of each feature at present, that is, the standardized score value of the peak value, root mean square value and kurtosis of the vibration signal, Z s It is the standardized score value of the RMS value of the current sound signal.

[0053] The present invention performs single feature detection and comprehensive score detection by calculating standardized scores to identify abnormal vibration conditions. Compared with the prior art, the accuracy of abnormality detection can be greatly improved.

[0054] Further, in step S36, the abnormal score value Z of the comprehensive score detection is specifically set. combined The steps of threshold θ are:

[0055] S361: Collect multiple comprehensive score values ​​Z of the system in normal operation combined ,These data are from long-term operation records under normal conditions;

[0056] S362: Calculate multiple comprehensive score values ​​Z combined The mean μ normal and standard deviation σ normal , calculate μ normal and σ normal The formula is:

[0057]

[0058] Where N is the comprehensive score Z combined The number of samples, Z combined，i is the comprehensive score value calculated under each normal operating condition, μ normal is the mean of multiple comprehensive ratings, σ normal is the standard deviation of multiple comprehensive score values;

[0059] S363: According to the calculated means and standard deviations of the multiple comprehensive score values, a comprehensive score threshold θ is set. The specific formula for calculating θ is:

[0060] θ=μ normal +k*σ normal

[0061] In the formula, θ is the threshold of comprehensive score detection, μ normal is the mean of multiple comprehensive rating values, σ normal is the standard deviation of multiple comprehensive score values, and k is an empirical coefficient.

[0062] Furthermore, k takes the value of 2 or 3. When the value is 2, the threshold includes more than 95% of normal data; when the value is 3, the threshold includes more than 99% of normal data. By adjusting the coefficient k, the sensitivity of anomaly detection is controlled.

[0063] Furthermore, after receiving the exception information, the alarm module uses the QHttpMultiPart class to encapsulate the parameters.

[0064] Beneficial effects of the present invention:

[0065] 1. The present invention installs multiple sensor nodes at multiple locations of the gas turbine to collect dynamic signals of the gas turbine in real time, thereby comprehensively monitoring the operating status of the gas turbine and improving the accuracy and timeliness of fault detection.

[0066] 2. The data processing module of the present invention processes and analyzes the received vibration and sound data in real time, extracts characteristic parameters such as vibration amplitude, frequency peak, root mean square value, kurtosis, etc. through physical value calculation, fast Fourier transform (FFT) and characteristic parameter extraction algorithm, and calculates the standardized score value of each feature. Compared with the prior art, the feature extraction algorithm of the present invention is more comprehensive and sophisticated, and can more accurately evaluate the operating status of the gas turbine and identify potential faults.

[0067] 3. The present invention designs single feature detection and comprehensive score detection. The single feature detection is to set thresholds for the standardized score values ​​of the vibration signal and sound signal of each feature respectively, which is a single score threshold corresponding to each feature parameter. The comprehensive score detection is to fuse the standardized scores of the vibration signal and sound signal of each feature, obtain the fused comprehensive abnormality score formula and set the comprehensive score threshold. The evaluation criteria of the present invention are comprehensive and can greatly improve the evaluation accuracy of the monitoring data. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 The structure block diagram of the gas turbine fault diagnosis system based on multiple sensors of the present invention.

[0069] Figure 2The present invention is a communication process diagram of a data acquisition module, a data receiving module and a data processing module of a multi-sensor based gas turbine fault diagnosis system. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solution and advantages of the present invention more clear, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and a preferred implementation manner.

[0071] Example 1

[0072] The present invention provides a gas turbine vibration monitoring system, which monitors the gas turbine specifically for vibration and sound monitoring. The system structure is as follows: Figure 1 As shown, it includes data acquisition equipment, a data core service system and a monitoring center, wherein the data acquisition equipment is composed of sensor nodes and data acquisition modules connected in sequence; the data core service system is deployed on the server, and the data core service system is composed of a data receiving module, a data processing module and an alarm module, which are connected in sequence, and the data receiving module and the data acquisition module are connected by wire to realize two-way transmission from the acquisition module to the receiving module. On the one hand, the data receiving module receives the collected data from the data acquisition module, and on the other hand, the data receiving module sends configuration parameters to the data acquisition module, such as sampling frequency, number of samplings, etc. The monitoring center is another monitoring system, and the monitoring center is connected to the alarm module to receive alarm information and display it to remind the operation and maintenance personnel.

[0073] The sensor nodes of the data acquisition equipment include multiple vibration sensors and sound sensors, which are installed in key parts of the gas turbine to collect vibration and sound signals from multiple parts of the gas turbine in real time. The sensor nodes are installed in key parts of the gas turbine (such as bearings and gearboxes, etc.), and each gas turbine is installed with 6-10 sensor nodes, each of which contains vibration sensors and sound sensors to ensure the stability and accuracy of signal transmission.

[0074] The data acquisition module is used to connect the sensor nodes, receive vibration and sound signals and transmit them to the data receiving module through wires. The data acquisition module supports a variety of device types, such as 8 channels, 16 channels, etc. The sampling rate of each channel is 20kHz and the resolution is 16 bits. It is responsible for collecting the vibration and sound signals collected by the sensor nodes and performing signal processing (including amplification, filtering and analog-to-digital conversion) on the signals.

[0075] The data receiving module can support multiple quantities and models of acquisition devices. The data receiving module is connected to the data acquisition module through wires to ensure real-time transmission of data. The data receiving module of the data core service system processes the dynamic signal and transmits it to the data processing module. The processing includes data reception and message delivery to the processing thread.

[0076] The data processing module includes a feature parameter extraction algorithm, through which the various feature parameters of the vibration and sound signals are extracted. The feature parameters are specifically the frequency peak, root mean square value and kurtosis of the vibration signal, and the root mean square value of the sound signal. The mean and standard deviation of each feature parameter are calculated, and a standardized score is calculated for each dynamic signal feature data point collected in real time, and a standardized score threshold for each feature parameter is set respectively, which is called a single scoring threshold; the data processing module also fuses the standardized scores of each feature, obtains a fused comprehensive abnormality scoring formula and sets a comprehensive scoring threshold, which is called a comprehensive scoring threshold; if the feature data point of the current dynamic signal exceeds the comprehensive scoring threshold or the single scoring threshold corresponding to each feature parameter, the data processing module sends an abnormal signal to the alarm module;

[0077] The monitoring center is connected to the alarm module, which receives abnormal signals and sends alarm information to the monitoring center for display to remind operation and maintenance personnel.

[0078] Furthermore, the data core service system also includes a storage module, which is connected to the alarm module and the data processing module. The storage module saves the collected dynamic signal data to a local disk, and saves the original data for a period of time before and after the alarm when an alarm occurs.

[0079] Example 2

[0080] Based on the above embodiment 1, the present invention further provides a control method of the gas turbine fault diagnosis system based on the above embodiment 1, comprising the following steps:

[0081] S1: The data acquisition module of the data acquisition device receives the dynamic signal data collected by the sensor node and transmits it to the data receiving module, and the data receiving module transmits the dynamic signal data to the data processing module;

[0082] S2: The data processing module processes and analyzes the dynamic signal data in real time. By calculating the physical value of the dynamic signal data, the original collected data is converted into a value with actual physical meaning for subsequent signal analysis and processing. By performing fast Fourier transform on the dynamic signal data, the time domain signal is converted into a frequency domain signal for analyzing the frequency components in the dynamic signal.

[0083] S3: The data processing module extracts the characteristic parameters of the dynamic signal through the characteristic parameter extraction algorithm, and evaluates the operating status of the gas turbine by analyzing the extracted characteristic parameters to identify whether an abnormal situation occurs;

[0084] S4: When the data processing module identifies an abnormal situation, it sends an abnormal signal to the alarm module.

[0085] In step S1, the communication process of the data acquisition module, the data receiving module and the data processing module of the data core service system is as follows: Figure 2 As shown, the specific communication steps are as follows:

[0086] S11: The data acquisition module receives the dynamic signal of the gas turbine collected in real time by the sensor node, and sends a detection packet 80 to the data receiving module after performing signal processing on the dynamic signal. The detection packet 80 includes the device number and type of the data acquisition module;

[0087] S12: The data receiving module obtains the current device number and type from the detection packet 80 in real time, and obtains the data processing thread corresponding to the current device number from the cache. If the device number exists, the device IP address is updated. If the device number does not exist, a corresponding data processing thread is created according to the current device number and type, and the device number and type as well as the created data processing thread information are stored in the QHASH structure for management, so as to facilitate subsequent data flow;

[0088] S13: The data receiving module sends the detection packet 80 to the data processing thread, and the data processing module sends the configuration parameters 81 of the data acquisition module to the data receiving module. The data processing module sends the configuration parameters 81 of the data acquisition device to the data receiving module according to the parameter configuration. The configuration parameters 81 include the acquisition frequency, sampling points and number of channels, and are forwarded by the data receiving module to the data acquisition module, waiting for the confirmation of the data acquisition module.

[0089] S14: After the data acquisition module receives the device configuration parameters 81, the data acquisition module sends the same confirmation message to the data receiving module. After receiving the message, the data receiving module forwards the data request message to the data acquisition module to inform the data receiving module that it is ready. The data acquisition module starts to collect sensor data, and the sensor data is integrated into raw data 82. The data acquisition module transmits the raw data 82 to the data receiving module, and the data receiving module then forwards the raw data 82 to the data processing module.

[0090] In step S2, the physical value calculation mainly calculates the physical value in each data acquisition window of the vibration data and the sound data, and the calculation formula of the physical value is:

[0091]

[0092] In the formula, nSingleValue is the original value collected from the sensor, 32767.0 is the maximum full-scale value of the vibration sensor and the sound sensor, 25000 is the maximum value of the physical quantity, when nSingleValue is equal to 32767, the physical value will reach 25000 units, fSens is the sensitivity of the channel, and fAlarm is the zero calibration value;

[0093] Through physical value calculation, the original collected data can be converted into values ​​with actual physical meaning for subsequent signal analysis and processing.

[0094] Fast Fourier Transform In order to analyze the frequency characteristics of the vibration signal, the time domain signal is converted into a frequency domain signal using the Fast Fourier Transform (FFT). The calculation formula of FFT is:

[0095]

[0096] Where X(f) is the frequency domain signal, x n is the nth sample point in the time domain signal, and N is the number of samples in the acquisition window. FFT can help analyze the main frequency components in the vibration signal, thereby identifying abnormal conditions in equipment operation, such as resonance and impact.

[0097] In step S3, the dynamic signal includes vibration data and sound data, which are collected by vibration sensors and sound sensors installed at different parts of the gas turbine, and the characteristic parameter extraction algorithm is applied, which specifically includes the following steps:

[0098] S31: Calculate the peak value of the vibration signal in each data acquisition window of the vibration data. The peak value is the maximum intensity of the signal. The peak value calculation formula of the vibration data is:

[0099] Peak Value = max(|x i振 |)

[0100] Where Peak Value is the peak value of the vibration signal in each data acquisition window, x i振 is the sample value of the vibration signal in each data acquisition window;

[0101] S32: Calculate the root mean square value of the vibration signal in each data acquisition window of the vibration data. The root mean square value reflects the average strength of the signal. The calculation formula of the root mean square value of the vibration signal is:

[0102]

[0103] Where RMS Value is the root mean square value of the vibration signal in each data acquisition window, N is the number of samples in each data acquisition window, and x i振 is the sample value of the vibration signal in each data acquisition window;

[0104] S33: Calculate the kurtosis of the vibration signal in each data acquisition window of the vibration data to identify the impact component in the signal. The calculation formula of the kurtosis of the vibration signal is:

[0105]

[0106] Where Kurtosis is the kurtosis of the vibration signal in each data acquisition window of the vibration data, N is the number of samples in the window, and x i振 is the sample value of the vibration signal in each data acquisition window, μ 振 is the mean of the signal, σ 振 is the standard deviation of the signal;

[0107] in, Where N is the number of samples in each data acquisition window, x i is the i-th vibration signal sample value;

[0108]

[0109] Where N is the number of samples in each data acquisition window, x i is the i-th vibration signal sample value, μ 振 is the average value;

[0110] S34: Calculate the root mean square value of the sound signal in each data collection window of the sound data to reflect the sound intensity characteristics. The calculation formula of the root mean square value of the sound signal is:

[0111]

[0112] Where Loudness is the root mean square value of the sound signal, x i声 is the sample value of the sound signal in each data acquisition window, and N is the number of samples in each data acquisition window;

[0113] S35: Statistical analysis and anomaly detection are performed on the vibration data and sound data, a normal state baseline is established, vibration signal and sound signal data under normal operating conditions are collected, and the mean and standard deviation of each characteristic vibration signal and sound signal are calculated. Each characteristic vibration signal and sound signal includes the peak value, root mean square value, kurtosis of the vibration signal and the root mean square value of the sound signal. A standardized score is calculated for each vibration characteristic data point and sound intensity data point collected in real time. During the calculation, the mean and standard deviation of the peak value of the vibration signal, the mean and standard deviation of the root mean square value, and the mean and standard deviation of the kurtosis are respectively substituted into the formula, and the mean and standard deviation of the root mean square value of the sound signal are substituted into the formula. The standardized score calculation formula for each characteristic is specifically as follows:

[0114]

[0115] In the formula, Z vi is the standardized score value of the vibration signal of each feature at present, Z s is the standardized score value of the current sound signal RMS value, x viis the current vibration characteristic data point, specifically the peak value, root mean square value and kurtosis of the vibration signal collected in real time in each data collection window, which are obtained by the formulas of steps S31, S32 and S33 respectively, x s is the current sound intensity data point, is the root mean square value of the sound signal collected in real time in each data collection window, obtained by the formula of the specific step S34, μ vi It is expressed as the mean of three eigenvalues, namely the peak value, root mean square value and kurtosis of the vibration signal, σ vi It is expressed as the standard deviation of three eigenvalues, namely the peak value, root mean square value and kurtosis of the vibration signal, μ s is the mean of the root mean square value of the sound signal, σ s is the standard deviation of the root mean square value of the sound signal;

[0116] S36: Perform anomaly detection on the calculated standardized score. The anomaly detection includes single feature detection and comprehensive score detection. The single feature detection is to set thresholds for the standardized score values ​​of the vibration signal and the sound signal of each feature, respectively, which are single score thresholds corresponding to each feature parameter. The vibration signal and the sound signal of each feature include the peak value, the root mean square value, the kurtosis of the vibration signal and the root mean square value of the sound signal. If the standardized score value Z of the vibration signal of any feature is vi Or the normalized score Z of the root mean square value of the sound signal s If the corresponding single score threshold is exceeded, the feature is considered abnormal;

[0117] Comprehensive scoring detection is to fuse the standardized scores of the vibration signals and sound signals of each feature, obtain the fused comprehensive abnormality scoring formula and set the comprehensive scoring threshold, such as Z combined If the value exceeds the threshold, it is determined that the machine is abnormal, and the Z obtained in step S35 is vi , that is, the peak value, RMS value and standardized score value of the vibration signal are substituted into the formula to find the average, and the RMS value Z of the sound signal is s Substituting the standardized score into the formula, the calculation formula for the integrated score value is:

[0118]

[0119] In the formula, Z combined is the comprehensive abnormality score after fusion, Z vi is the standardized score value of the vibration signal of each feature at present, that is, the standardized score value of the peak value, root mean square value and kurtosis of the vibration signal. Therefore, n in the formula is 3, and Z s It is the standardized score value of the RMS value of the current sound signal.

[0120] In the above step S36, the specific steps of setting the comprehensive score threshold are:

[0121] S361: Collect a large amount of comprehensive scoring data Z when the system is in normal operation combined These data are from long-term operation records under normal conditions, that is, multiple Z without any abnormal failure. combined value;

[0122] S362: Calculate multiple comprehensive score values ​​Z combined The mean μ normal and standard deviation σ normal , where μ normal To calculate μ normal and σ normal The formula is:

[0123]

[0124] Where N is the comprehensive score Z combined The number of samples, Z combined，i is the comprehensive score value calculated under each normal operating condition, μ normal is the mean of multiple comprehensive ratings, σ normal is the standard deviation of multiple comprehensive score values;

[0125] S363: According to the calculated means and standard deviations of the multiple comprehensive score values, a comprehensive score threshold θ is set. The specific formula for calculating θ is:

[0126] θ=μ normal +k*σ normal

[0127] In the formula, θ is the threshold of comprehensive score detection, μ normal is the mean of multiple comprehensive rating values, σ normal is the standard deviation of multiple comprehensive score values, and k is an empirical coefficient.

[0128] The comprehensive score has only one threshold. The comprehensive abnormality score is a value obtained by comprehensively calculating the standardized score values ​​of the three characteristic values ​​of the vibration signal and the standardized score value of the root mean square value of the sound signal. The comprehensive score threshold θ is the threshold set for this value.

[0129] The value of k is usually 2 or 3. When the value is 2, the threshold contains more than 95% of normal data; when the value is 3, the threshold contains more than 99% of normal data. By adjusting the coefficient k, the sensitivity of anomaly detection can be flexibly controlled. If you want to detect more potential anomalies, you can lower the value of k; if you want to reduce false positives, you can increase the value of k. For example, in the normal operation data collected, the average of the comprehensive score is μ normal=10, the standard deviation is σ normal =2, if k=2, then the threshold is:

[0130] θ=10+2*2=14

[0131] Therefore, when the comprehensive score, Z combined When it exceeds 14, the system will determine it as abnormal.

[0132] The present invention collects vibration and sound signals in real time by installing multiple sensor nodes at key parts of the gas turbine (such as bearings and gear boxes). Each gas turbine is installed with 6-10 sensor nodes, and each node includes a vibration sensor (accelerometer) and a sound sensor, so as to comprehensively monitor the operating state of the gas turbine and improve the accuracy and timeliness of fault detection; the data processing module processes and analyzes the received vibration and sound data in real time, and extracts characteristic parameters such as vibration amplitude, frequency peak, root mean square value, kurtosis, etc. through physical value calculation, fast Fourier transform (FFT) and characteristic parameter extraction algorithm. Compared with the prior art, the present invention is more comprehensive and precise in feature extraction algorithm, and can more accurately evaluate the operating state of the gas turbine and identify potential faults; the data receiving module adopts multi-threaded data processing technology, and realizes dynamic access and efficient data flow of multiple devices and multiple types of devices through dynamic IP access and QHASH structure management. Compared with the prior art, the present invention has obvious advantages in the flexibility and efficiency of data transmission and processing, and can adapt to complex field environments and diverse equipment types; single feature detection and comprehensive score detection are performed by calculating the standardized score (Z-Score) to identify abnormal vibration conditions. Compared with the prior art, the present invention has significant advantages in the accuracy of anomaly detection.

[0133] Example 3

[0134] Based on Example 1 and Example 2, the core service system of the present invention also includes a storage module, which is connected to the alarm module and the data processing module. The storage module saves the collected dynamic signal data to the local disk, and saves the original data within a period of time before and after the alarm when an alarm occurs.

[0135] When the data processing module identifies that the operating status of the gas turbine is abnormal, it sends an abnormal signal to the alarm module. After receiving the abnormal information, the alarm module uses the QHttpMultiPart class to encapsulate the parameters. The request parameters are defined as follows:

[0136] file (required): The type is File, which means the attachment file needs to be uploaded.

[0137] unitCode (required): The type is String, which represents the device number and is used to identify a specific device.

[0138] problem (optional): String type, used to describe the problem of the device. If there is no problem, it can be left blank.

[0139] alarmTime (optional): type is String, format is date and time (such as: 2024-06-2012:13:23), recording the time when the alarm occurs.

[0140] eventType (required): The type is Integer, indicating the type of event, used to classify monitoring events.

[0141] The implementation steps are as follows:

[0142] 1. Create a QNetworkAccessManager instance to handle network requests.

[0143] 2. Build a QHttpMultiPart object and encapsulate the request parameters as a multi-part form.

[0144] 3. Set the file, device number, problem description, alarm time and event type parameters to

[0145] QHttpMultiPart.

[0146] 4. Call QNetworkAccessManager::post to send a POST request to the specified URL.

[0147] 5. Process the server response and parse the returned results.

[0148] This application uses QHttpMultiPart to construct and send HTTP requests containing multiple parts (such as text and files). The advantages are: 1. Qt provides a cross-platform API, and the code can be easily ported on Windows, Linux, and macOS. 2. The signal and slot mechanism can easily handle asynchronous network operations, and the code structure is clear. Using other libraries, such as libcurl or Boost.Asio, their cross-platform performance and code writing complexity are worse than Qt.

[0149] When the data processing module finds data abnormalities, it sends an abnormal signal to the alarm module. The storage module saves the collected vibration and sound data to the local disk, and saves the original data for a period of time before and after the alarm occurs to ensure the integrity and traceability of the data. The storage module supports two storage methods: CSV format and MySQL database, which facilitates data management and query.

[0150] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications are also within the protection scope of the present invention.

Claims

1. A multi-sensor based gas turbine fault diagnosis system, characterized in that: It includes data acquisition equipment, a data core service system and a monitoring center connected in sequence, wherein the data acquisition equipment is composed of a sensor node and a data acquisition module; the data core service system is composed of a data receiving module, a data processing module and an alarm module, the data receiving module, the data processing module and the alarm module are connected in sequence, and the alarm module is connected to the data processing module; The sensor nodes of the data acquisition equipment include multiple types of sensors, which are installed in different parts of the gas turbine to collect dynamic signals from multiple parts of the gas turbine in real time and monitor the status of the gas turbine; The data acquisition module of the data acquisition equipment is used to connect to the sensor nodes, receive dynamic signals and transmit them to the data receiving module; The data receiving module of the data core service system receives the dynamic signal and transmits it to the data processing module. The data processing module includes a feature parameter extraction algorithm, which extracts each feature parameter through the feature parameter extraction algorithm, calculates the mean and standard deviation of each feature parameter, and performs standardized score calculation for each dynamic signal feature data point collected in real time, and sets the standardized score threshold of each feature parameter, which is called a single scoring threshold; the data processing module also fuses the standardized scores of each feature parameter, obtains the fused comprehensive abnormal scoring formula and sets the comprehensive scoring threshold; if the feature data point of the current dynamic signal exceeds the comprehensive scoring threshold or the single scoring threshold corresponding to each feature parameter, the data processing module sends an abnormal signal to the alarm module; The monitoring center is connected to the alarm module, which receives abnormal signals and sends alarm information to the monitoring center for display to remind operation and maintenance personnel.

2. A multi-sensor based gas turbine fault diagnosis system according to claim 1, characterized in that: The sensor nodes of the data acquisition equipment include multiple vibration sensors and sound sensors, which collect vibration signals and sound signals from multiple parts of the gas turbine in real time. They are installed on the bearings and gearboxes of the gas turbine. Each gas turbine is installed with 6-10 sensor nodes, and each node contains a vibration sensor and a sound sensor to ensure the stability and accuracy of signal transmission.

3. A multi-sensor based gas turbine fault diagnosis system according to claim 1, characterized in that: The data acquisition module of the data core service system supports a variety of device types, including 8 channels and 16 channels. The sampling rate and resolution of each channel are consistent. The data receiving module can support multiple quantities and models of acquisition devices and dynamic IP access.

4. A multi-sensor based gas turbine fault diagnosis system according to claim 1, characterized in that: The data core service system also includes a storage module, which is connected to the alarm module and the data processing module. The storage module saves the collected dynamic signal data to the local disk and saves the original data for a period of time before and after the alarm when an alarm occurs.

5. A control method for a gas turbine fault diagnosis system based on any one of claims 1 to 4, characterized in that: The steps include: S1: The data acquisition module of the data acquisition device receives the dynamic signal data collected by the sensor node and transmits it to the data receiving module, and the data receiving module transmits the dynamic signal data to the data processing module; S2: The data processing module processes and analyzes the dynamic signal data in real time. By calculating the physical value of the dynamic signal data, the original collected data is converted into a value with actual physical meaning for subsequent signal analysis and processing. By performing fast Fourier transform on the dynamic signal data, the time domain signal is converted into a frequency domain signal for analyzing the frequency components in the dynamic signal. S3: The data processing module extracts the characteristic parameters of the dynamic signal through the characteristic parameter extraction algorithm, and evaluates the operating status of the gas turbine by analyzing the extracted characteristic parameters to identify whether an abnormal situation occurs; S4: When the data processing module identifies an abnormal situation, it sends an abnormal signal to the alarm module.

6. The control method of a multi-sensor based gas turbine fault diagnosis system according to claim 5, characterized in that: In step S1, the specific communication steps of the data acquisition module, data receiving module and data processing module of the data core service system are as follows: S11: The data acquisition module receives the dynamic signal of the gas turbine collected in real time by the sensor node, and sends a detection packet 80 to the data receiving module after performing signal processing on the dynamic signal. The detection packet 80 includes the device number and type of the data acquisition module; S12: The data receiving module obtains the current device number and type from the detection packet 80 in real time, and obtains the data processing thread corresponding to the current device number from the cache. If the device number exists, the device IP address is updated. If the device number does not exist, a corresponding data processing thread is created according to the current device number and type, and the device number and type as well as the created data processing thread information are stored in the QHASH structure for management, so as to facilitate subsequent data flow; S13: The data receiving module sends the detection packet 80 to the data processing thread, and the data processing module sends the configuration parameters 81 of the data acquisition module to the data receiving module. The data processing module sends the configuration parameters 81 of the data acquisition device to the data receiving module according to the parameter configuration. The configuration parameters 81 include the acquisition frequency, sampling points and number of channels, and are forwarded by the data receiving module to the data acquisition module, waiting for the confirmation of the data acquisition module. S14: After the data acquisition module receives the device configuration parameters 81, the data acquisition module sends the same confirmation message to the data receiving module. After receiving the message, the data receiving module forwards the data request message to the data acquisition module to inform the data receiving module that it is ready. The data acquisition module starts to collect sensor data, and the sensor data is integrated into raw data 82. The data acquisition module transmits the raw data 82 to the data receiving module, and the data receiving module then forwards the raw data 82 to the data processing module.

7. The control method of a multi-sensor based gas turbine fault diagnosis system according to claim 5, characterized in that: In step S3, the dynamic signal includes vibration data and sound data, which are collected by vibration sensors and sound sensors installed at different parts of the gas turbine, and the characteristic parameter extraction algorithm is applied, which specifically includes the following steps: S31: Calculate the peak value of the vibration signal in each data acquisition window of the vibration data. The peak value is the maximum intensity of the signal. The peak value calculation formula of the vibration data is: Peak Value=max(|x i振 |) Where Peak Value is the peak value of the vibration signal in each data acquisition window, x i振 is the sample value of the vibration signal in each data acquisition window; S32: Calculate the root mean square value of the vibration signal in each data acquisition window of the vibration data. The root mean square value reflects the average strength of the signal. The calculation formula of the root mean square value of the vibration signal is: Where RMS Value is the root mean square value of the vibration signal in each data acquisition window, N is the number of samples in each data acquisition window, and x i振 is the sample value of the vibration signal in each data acquisition window; S33: Calculate the kurtosis of the vibration signal in each data acquisition window of the vibration data to identify the impact component in the signal. The calculation formula of the kurtosis of the vibration signal is: Where Kurtosis is the kurtosis of the vibration signal in each data acquisition window of the vibration data, N is the number of samples in the window, and x i振 is the sample value of the vibration signal in each data acquisition window, μ 振 is the mean of the signal, σ 振 is the standard deviation of the signal; in, Where N is the number of samples in each data acquisition window, x i is the i-th vibration signal sample value; Where N is the number of samples in each data acquisition window, x i is the i-th vibration signal sample value, μ 振 is the average value; S34: Calculate the root mean square value of the sound signal in each data collection window of the sound data to reflect the sound intensity characteristics. The calculation formula of the root mean square value of the sound signal is: Where Loudness is the root mean square value of the sound signal, x i声 is the sample value of the sound signal in each data acquisition window, and N is the number of samples in each data acquisition window; S35: Statistical analysis and anomaly detection are performed on the vibration data and sound data, a normal state baseline is established, vibration signal and sound signal data under normal operating conditions are collected, and the mean and standard deviation of each characteristic vibration signal and sound signal are calculated. Each characteristic vibration signal and sound signal includes the peak value, root mean square value, kurtosis of the vibration signal and the root mean square value of the sound signal. A standardized score is calculated for each vibration characteristic data point and sound intensity data point collected in real time. During the calculation, the mean and standard deviation of the peak value of the vibration signal, the mean and standard deviation of the root mean square value, and the mean and standard deviation of the kurtosis are respectively substituted into the formula, and the mean and standard deviation of the root mean square value of the sound signal are substituted into the formula. The standardized score calculation formula for each characteristic is specifically as follows: In the formula, Z vi is the standardized score value of the vibration signal of each feature at present, Z s is the standardized score value of the current sound signal RMS value, x vi is the current vibration characteristic data point, specifically the peak value, root mean square value and kurtosis of the vibration signal collected in real time in each data collection window, which are obtained by the formulas of steps S31, S32 and S33 respectively, x s is the current sound intensity data point, is the root mean square value of the sound signal collected in real time in each data collection window, obtained by the formula of the specific step S34, μ vi It is expressed as the mean of three eigenvalues, namely the peak value, root mean square value and kurtosis of the vibration signal, σ vi It is expressed as the standard deviation of three eigenvalues, namely the peak value, root mean square value and kurtosis of the vibration signal, μ s is the mean of the root mean square value of the sound signal, σ s is the standard deviation of the root mean square value of the sound signal; S36: Perform anomaly detection on the calculated standardized score. The anomaly detection includes single feature detection and comprehensive score detection. The single feature detection is to set thresholds for the standardized score values ​​of the vibration signal and sound signal of each feature respectively, which are single score thresholds corresponding to each feature parameter. The vibration signal and sound signal of each feature include the peak value, root mean square value, kurtosis of the vibration signal and the root mean square value of the sound signal. If the standardized score value Z of the vibration signal of any feature obtained in step S35 is vi Or the normalized score Z of the root mean square value of the sound signal s If the corresponding single score threshold is exceeded, the feature is considered abnormal; Comprehensive scoring detection is to fuse the standardized scores of the vibration signals and sound signals of each feature, obtain the fused comprehensive abnormality scoring formula and set the comprehensive scoring threshold, such as Z combined If the value exceeds the threshold, it is determined that the machine is abnormal, and the Z obtained in step S35 is vi , that is, the peak value, RMS value and standardized score value of the vibration signal are substituted into the formula to find the average, and the RMS value Z of the sound signal is s Substituting the standardized score into the formula, the calculation formula for the integrated score value is: In the formula, Z combined is the comprehensive abnormality score after fusion, Z vi is the standardized score value of the vibration signal of each feature at present, that is, the standardized score value of the peak value, root mean square value and kurtosis of the vibration signal, Z s It is the standardized score value of the RMS value of the current sound signal.

8. A multi-sensor based gas turbine fault diagnosis system according to claim 7, characterized in that: In step S36, the abnormal score value Z of the comprehensive score detection is specifically set combined The steps of threshold θ are: S361: Collect multiple comprehensive score values ​​Z of the system in normal operation combined ,These data are from long-term operation records under normal conditions; S362: Calculate multiple comprehensive score values ​​Z combined The mean μ normal and standard deviation σ normal , calculate μ normal and σ normal The formula is: Where N is the comprehensive score Z combined The number of samples, Z combined,i is the comprehensive score value calculated under each normal operating condition, μ normal is the mean of multiple comprehensive ratings, σ normal is the standard deviation of multiple comprehensive score values; S363: According to the calculated means and standard deviations of the multiple comprehensive score values, a comprehensive score threshold θ is set. The specific formula for calculating θ is: θ=μ normal +k*s normal In the formula, θ is the threshold of comprehensive score detection, μ normal is the mean of multiple comprehensive score values, σ normal is the standard deviation of multiple comprehensive score values, and k is an empirical coefficient.

9. The control method of a multi-sensor based gas turbine fault diagnosis system according to claim 8, characterized in that: The value of k is 2 or 3. When the value is 2, the threshold includes more than 95% of the normal data; when the value is 3, the threshold includes more than 99% of the normal data. The sensitivity of anomaly detection is controlled by adjusting the coefficient k.

10. The control method of a multi-sensor based gas turbine fault diagnosis system according to claim 5, characterized in that: After receiving the exception information, the alarm module uses the QHttpMultiPart class to encapsulate the parameters.

Images