Gas turbine rotor vibration analysis system based on vibration analysis and machine learning technology

Abstract

The invention discloses a gas turbine rotor vibration analysis system based on vibration analysis and machine learning technologies, and relates to the technical field of gas turbine rotor vibration analysis. Comprising a high-frequency real-time acquisition sub-module, a frequency domain analysis sub-module, an axis track drawing sub-module, a characteristic parameter calculation sub-module, a fault diagnosis sub-module, a machine learning sub-module, an adaptive model updating sub-module, an automatic storage sub-module, a dual-storage mechanism sub-module and a safety protection module. The system is high in integration level, powerful in data processing and analysis function and easy to operate, a user can check and manage all collection and analysis results on a monitoring interface in real time, and the operation efficiency and user experience of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of combustion engine rotor vibration analysis, and in particular to a combustion engine rotor vibration analysis system based on vibration analysis and machine learning technology. Background Art

[0002] The rotor system of a gas-steam combined cycle unit will produce a series of vibrations during high-speed operation. If these vibrations are not effectively monitored and diagnosed, they may lead to serious equipment failures and economic losses.

[0003] At present, the technologies for monitoring and diagnosing gas turbine rotor vibration mainly focus on data acquisition, spectrum analysis, axis trajectory analysis, and fault diagnosis based on eigenvalues.

[0004] However, the existing systems still use traditional data processing methods, which have poor real-time performance and are unable to detect and warn of potential faults in a timely manner. In addition, most of the existing fault diagnosis algorithms are based on traditional methods and lack intelligent analysis and diagnosis of complex vibration signals. Summary of the invention

[0005] In view of the problems existing in the existing system, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to improve the accuracy and reliability of 9FA type gas turbine rotor vibration monitoring and diagnosis.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, an embodiment of the present invention provides a combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology, which includes a data acquisition and analysis module, including a high-frequency real-time acquisition submodule, which is used to collect vibration data during equipment operation; a frequency domain analysis submodule, which is used to perform frequency domain transformation on the collected time domain data, conduct frequency distribution characteristic analysis, and identify the occurrence and amplitude of specific frequencies or abnormal frequencies; an axis trajectory and characteristic parameter analysis module, including an axis trajectory drawing submodule, which is used to process data collected by orthogonal axis vibration eddy current sensors, and identify abnormal states of the rotor system by analyzing the shape and characteristics of the axis trajectory; a characteristic parameter calculation submodule, which is used to calculate key vibration characteristic parameters and provide rich diagnostic data support. ; Fault diagnosis and learning module, including a fault diagnosis sub-module, which is used to establish fault judgment criteria and perform intelligent diagnosis; a machine learning sub-module, which is used to use monitoring data to train the diagnosis model and perform fault classification; an adaptive model update sub-module, which is used to integrate online learning capabilities and iteratively optimize the model using continuously updated monitoring data; the storage module includes an automatic storage sub-module, which is used to determine whether the unit is in operation based on the threshold of the speed channel, and automatically trigger data storage when it is judged to be in operation; a dual storage mechanism sub-module, which is used to write the collected vibration data to the local storage in real time, and transmit it to the hyper-converged server through the network to ensure data backup and integrity; a security protection module, which is used to ensure data security and protection through physical isolation of a one-way network firewall.

[0009] As a preferred solution of the gas turbine rotor vibration analysis system based on vibration analysis and machine learning technology described in the present invention, the frequency domain analysis submodule includes: denoising the collected time domain data; segmenting the data to ensure that the length of each data segment is suitable for frequency domain transformation; using the FFT algorithm to convert the time domain data into frequency domain data and calculate the power spectrum density; identifying the main frequency components and calculating the amplitude and phase information of each frequency component; identifying the key characteristic frequencies according to the equipment characteristics, and monitoring the amplitude changes of the key characteristic frequencies.

[0010] As a preferred solution of the combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology described in the present invention, the axis trajectory drawing submodule specifically includes the following contents: ensuring the synchronous acquisition of eddy current sensor data in the X direction and the Y direction, time-aligning the acquired data, applying a bandpass filter, and removing high-frequency noise and low-frequency drift; converting the original voltage signal of the sensor into displacement information, calibrating the coordinate system, and ensuring the orthogonality of the XY coordinate system; using the XY displacement data to draw the axis trajectory, generating a static axis trajectory diagram and a dynamic trajectory animation; calculating the geometric characteristics of the trajectory, analyzing the directionality and rotation characteristics of the trajectory; classifying according to the trajectory shape, and identifying abnormal trajectory patterns.

[0011] As a preferred solution of the combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology of the present invention, the specific operation steps of the directionality of the analysis trajectory are as follows: calculate the instantaneous velocity vector of each sampling point, and the given point (a i , b i ), and the next point (a i+1 , b i+1 ), velocity vector v i It is expressed as:

[0012] v i =(a i+1 -a i ,b i+1 -b i )

[0013] The velocity vector v i Indicates the direction and speed of the trajectory; for a continuous velocity vector v i and v i+1 , calculate the cross product v i ×v i+1 , calculated by the following formula:

[0014] v i *v i+1 =(a i *b i+1 -b i *a i+1 ).

[0015] As a preferred solution of the gas turbine rotor vibration analysis system based on vibration analysis and machine learning technology described in the present invention, the sign of the cross product is used to determine the rotation direction: if the cross product is positive, it indicates that the trajectory rotates counterclockwise; if the cross product is negative, it indicates that the trajectory rotates clockwise.

[0016] As a preferred solution of the combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology of the present invention, the specific operation steps of analyzing the rotation characteristics are as follows: Calculate the instantaneous angular velocity of the trajectory: estimate the instantaneous angular velocity of the trajectory by the rate of change of the trajectory points, given the continuous points (a i , b i ), and the next point (a i+1 , b i+1 ), angular velocity ω i Calculate using:

[0017]

[0018] Among them, θ i is a point (a i , b i) is the angle relative to the center point, Δt is the time interval, and the mean and standard deviation of the angular velocity are calculated. If the standard deviation of the angular velocity is lower than the preset secondary threshold, it indicates that the rotation is stable and the operation state of the mechanical equipment is relatively stable; if the standard deviation of the angular velocity is higher than the preset primary threshold, it indicates that the rotation is unstable.

[0019] As a preferred solution of the gas turbine rotor vibration analysis system based on vibration analysis and machine learning technology described in the present invention, the machine learning submodule specifically includes the following contents: using random forest for fault diagnosis, including the following steps: constructing multiple decision trees, each tree is trained based on a randomly selected feature subset, and the final diagnosis result is determined by majority voting; based on the change trend of the characteristic parameters, the severity of the fault is evaluated:

[0020] Severity=α*ΔRMS+β*ΔCF+γ*ΔK

[0021] Among them, α, β, γ are weight coefficients, Δ represents the change of the parameter relative to the baseline value, RMS is the root mean square value, CF is the peak factor, and K is the kurtosis. Combine the results of multiple diagnostic methods to calculate the diagnostic credibility:

[0022] Confidence=(N svm +Nr f +N rule ) / (3×N total )

[0023] Among them, N svm ,Nr f ,N rule are the number of consistent samples diagnosed by SVM, random forest and rule base, respectively, N total is the total number of samples; a diagnostic report is generated based on the above content, including the detected fault type, severity, diagnostic result credibility, key characteristic parameters and their changing trends, and the diagnostic results are fed back to the characteristic parameter calculation submodule to optimize feature selection.

[0024] In a second aspect, an embodiment of the present invention provides a combustion engine rotor vibration analysis method based on vibration analysis and machine learning technology, which includes:

[0025] Start the high-frequency real-time acquisition submodule to continuously monitor the vibration data of the equipment during operation and capture the time domain signal of the equipment vibration through the vibration sensor;

[0026] The collected time-domain vibration data is input into the frequency-domain analysis submodule, the data is Fourier transformed, the frequency-domain data is generated, the frequency distribution characteristics are analyzed, and specific frequencies or abnormal frequencies and their amplitudes are identified;

[0027] The axis trajectory drawing submodule processes the data from the orthogonal axis vibration sensor to generate an axis trajectory map. By analyzing the shape and characteristics of the axis trajectory, the potential abnormal state of the rotor system can be identified.

[0028] Use the characteristic parameter calculation submodule to calculate key vibration characteristic parameters based on the axis trajectory and frequency domain analysis results;

[0029] Input the characteristic parameters into the fault diagnosis submodule, perform preliminary fault identification according to the preset fault judgment criteria, intelligently diagnose the health status of the equipment, determine whether there is an abnormality and preliminarily classify the fault type;

[0030] In the machine learning submodule, the previously collected and analyzed vibration data is used to train the diagnostic model. By combining historical data with real-time data, the accuracy of the fault diagnosis model is continuously optimized and improved.

[0031] Enable the adaptive model update submodule to use new monitoring data for online learning and model updates to ensure that the diagnostic model can adapt to changes in equipment status and maintain efficient fault classification capabilities;

[0032] Activate the automatic storage submodule to determine whether the device is in operation based on the threshold of the speed channel. If the device is in operation, the data storage function is automatically triggered to ensure the continuity of vibration data.

[0033] While storing data, the dual storage mechanism submodule is used to write real-time vibration data into local storage and transmit it to the hyper-converged server through the network for backup, ensuring the integrity and security of the data;

[0034] Physical isolation is implemented through the security protection module, and a one-way network gate is used to ensure the safe transmission and storage of vibration data, prevent external network attacks and data leakage, and ensure the overall security of the system.

[0035] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as described in the first aspect of the present invention are implemented.

[0036] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as described in the first aspect of the present invention are implemented.

[0037] The beneficial effects of the present invention are as follows: the present invention adopts domestically produced high-precision data acquisition equipment, reduces dependence on imported equipment, saves costs, and enhances the autonomy and maintainability of the system; the present invention designs an automatic storage function, which effectively improves the accuracy and reliability of data, making fault data storage and analysis more intelligent and efficient; the system has a variety of data analysis capabilities, and generates multi-dimensional data displays such as spectrum diagrams and axis trajectory diagrams in real time, which can improve the intuitiveness and accuracy of fault diagnosis. In addition, the physical isolation of a one-way network switch is adopted to ensure the data security and protection of the production area in Zone 1, prevent external network attacks, and meet the requirements of power safety level protection. The system of the present invention has high system integration, powerful data processing and analysis functions, and is easy to operate. Users can view and manage all acquisition and analysis results in real time on the monitoring interface, which improves the system's operating efficiency and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 This is a structural diagram of the gas turbine rotor vibration analysis system based on vibration analysis and machine learning technology.

[0040] Figure 2 This is a flow chart of the gas turbine rotor vibration analysis method based on vibration analysis and machine learning technology.

[0041] Figure 3 This is the topology diagram of the 9FA gas turbine rotor vibration monitoring and diagnosis system. DETAILED DESCRIPTION

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0043] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0044] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0045] Example 1

[0046] Reference Figure 1-2 , which is the first embodiment of the present invention, and provides a combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology, including:

[0047] The data acquisition and analysis module includes a high-frequency real-time acquisition submodule, which is used to collect vibration data during equipment operation at a sampling rate of up to 20kHz to ensure that subtle changes in equipment operation are captured; a frequency domain analysis submodule, which is used to perform frequency domain transformation on the collected time domain data, conduct frequency distribution characteristic analysis, and identify the occurrence and amplitude of specific frequencies or abnormal frequencies to achieve fault diagnosis;

[0048] The axis trajectory and characteristic parameter analysis module includes an axis trajectory drawing submodule, which is used to process the data collected by the orthogonal axis vibration eddy current sensor and identify the abnormal state of the rotor system by analyzing the shape and characteristics of the axis trajectory; the characteristic parameter calculation submodule is used to calculate key vibration characteristic parameters such as RMS (root mean square value), n-fold frequency amplitude, and n-fold frequency amplitude ratio, providing rich diagnostic data support;

[0049] The fault diagnosis and learning module includes a fault diagnosis submodule, which is used to establish fault judgment criteria and perform intelligent diagnosis based on characteristic parameters and existing fault models through time domain and frequency domain characteristic value monitoring and trend analysis; a machine learning submodule, which is used to train the diagnosis model using monitoring data, gradually improve the accuracy of fault detection, and perform fault classification; an adaptive model update submodule, which is used to integrate online learning capabilities, iteratively optimize the model using continuously updated monitoring data, and improve the prediction capability of long-term operation;

[0050] The storage module includes an automatic storage submodule, which is used to determine whether the unit is in operation according to the threshold of the speed channel. When it is determined to be in operation, it automatically triggers data storage and ensures the accuracy and reliability of the data; a dual storage mechanism submodule is used to write the collected vibration data into local storage in real time and transmit it to the hyper-converged server through the network to ensure data backup and integrity and prevent data loss;

[0051] The security protection module is used to ensure data security and protection through physical isolation of a one-way network switch to meet the power safety level protection requirements.

[0052] Among them, the high-frequency real-time acquisition submodule in the data acquisition and analysis module collects vibration data and transmits it to the frequency domain analysis submodule for frequency domain analysis; the output data of the data acquisition and analysis module is transmitted to the axis trajectory and characteristic parameter analysis module for axis trajectory drawing and characteristic parameter calculation; the analysis results of the axis trajectory and characteristic parameter analysis module are transmitted to the fault diagnosis and learning module, among which the fault diagnosis submodule performs intelligent diagnosis, the machine learning submodule optimizes the model, and the adaptive model update submodule further improves the model prediction capability; the results of the fault diagnosis and learning module are transmitted to the automatic storage submodule and the dual storage mechanism submodule to ensure the safe storage of data; the physical isolation submodule of the safety protection module ensures the security of the data transmission and storage process of the entire system.

[0053] Furthermore, the high-frequency real-time acquisition submodule specifically includes the following contents:

[0054] The data acquisition instrument has an adjustable sampling frequency of up to 20kHz, which can realize the collection and storage of high-frequency real-time data, ensuring the capture of subtle changes in vibration signals and improving the ability to capture fault characteristics.

[0055] The use of domestically produced high-precision data acquisition equipment reduces dependence on imported equipment, saves costs, and enhances the autonomy and maintainability of the system.

[0056] Furthermore, the frequency domain analysis submodule specifically includes the following contents: denoising the collected time domain data, such as using a low-pass filter to eliminate high-frequency noise; segmenting the data to ensure that the length of each data segment is suitable for frequency domain transformation (usually a power of 2); using the FFT algorithm to convert the time domain data into frequency domain data, and calculating the power spectral density (PSD) to better display the energy distribution of the frequency components.

[0057] Identify the main frequency components, such as fundamental frequency, harmonics and harmonics, and calculate the amplitude and phase information of each frequency component; identify key characteristic frequencies (such as rotor imbalance, bearing failure and other related frequencies) according to equipment characteristics, and monitor the amplitude changes of these characteristic frequencies.

[0058] Furthermore, the axis trajectory drawing submodule specifically includes the following contents:

[0059] First, it is necessary to ensure that the eddy current sensor data in the X and Y directions are collected synchronously, time-align the collected data, and apply a bandpass filter to remove high-frequency noise and low-frequency drift; convert the sensor's original voltage signal into displacement information, calibrate the coordinate system, and ensure the orthogonality of the XY coordinate system; use the XY displacement data to draw the axis trajectory, and generate a static axis trajectory diagram and a dynamic trajectory animation.

[0060] Furthermore, the geometric characteristics of the trajectory, such as ellipticity, roundness, maximum eccentricity, etc., are calculated to analyze the directionality and rotation characteristics of the trajectory. The specific steps are as follows:

[0061] Calculate the instantaneous velocity vector of each sampling point. i , b i ), and the next point (a i+1 , b i+1 ), velocity vector v i It can be expressed as:

[0062] v i =(a i+1 -a i ,b i+1 -b i )

[0063] The velocity vector indicates the direction and speed of the trajectory.

[0064] For a continuous velocity vector v i and v i+1 , calculate its cross product v i ×v i+1 , can be calculated by the following formula:

[0065] v i *v i+1 =(a i *b i+1 -b i *a i+1 )

[0066] The sign of the cross product can be used to determine the direction of rotation:

[0067] If the cross product is positive, it means the trajectory is rotating counterclockwise;

[0068] If the cross product is negative, it means the trajectory is rotating in a clockwise direction.

[0069] Compute the sum (or average) of the cross products of all pairs of points, if the sum is positive then most rotations are counter-clockwise; if it is negative then most rotations are clockwise.

[0070] Furthermore, the rotational characteristics analysis includes:

[0071] Calculate the instantaneous angular velocity of the trajectory:

[0072] The instantaneous angular velocity can be estimated by the rate of change of the trajectory points. Given the continuous points (a i , b i ), and the next point (a i+1 , b i+1 ), its angular velocity ω i It can be calculated as follows:

[0073]

[0074] Among them, θ i is a point (a i , b i ) is the angle relative to the center point (usually (0,0) or the centroid of the trajectory), and Δt is the time interval.

[0075] Calculate the mean and standard deviation of the angular velocity. If the standard deviation of the angular velocity is low, it indicates that the rotation is stable and the mechanical equipment is running smoothly. If it is high, it means that the rotation is unstable and there may be problems such as imbalance, looseness or mechanical wear. The standard deviation of the angular velocity can be compared with the preset threshold to make a judgment. For example, if the standard deviation of the angular velocity is lower than the preset secondary threshold, it indicates that the rotation is stable and the mechanical equipment is running smoothly. If the standard deviation of the angular velocity is higher than the preset primary threshold, it means that the rotation is unstable.

[0076] Classify based on trajectory shape (e.g., circular, elliptical, figure-of-eight, etc.) and identify abnormal trajectory patterns such as inner / outer ring, multi-petal, etc.

[0077] Optionally, generate a trend graph of the shaft position over time, analyze the long-term trend of the shaft position, and analyze the phase characteristics of the shaft trajectory in combination with the speed signal to identify forward and backward precession phenomena. For example, forward precession is usually associated with normal operation and appears as a trajectory change in the same direction as the speed. Backward precession may indicate instability or reverse precession in the equipment, which usually requires further inspection and maintenance.

[0078] It can be seen that the operation status of mechanical equipment can be effectively diagnosed by analyzing the geometric characteristics of the directionality and rotation characteristics of the trajectory. The clockwise or counterclockwise rotation direction and the stability indicators of the rotation speed (such as the average angular velocity and standard deviation) can provide valuable information to help identify potential equipment failures and maintenance needs.

[0079] Furthermore, the characteristic parameter calculation submodule specifically includes the following contents:

[0080] The calculated time domain characteristic parameters include: root mean square value (RMS), peak factor CF and kurtosis K.

[0081] The calculated frequency domain characteristic parameters include: n-fold frequency amplitude An and n-fold frequency amplitude ratio Rn.

[0082] Furthermore, in order to ensure the comparability of characteristic parameters of different dimensions, the calculated characteristic parameters are normalized.

[0083] Combine the calculated time domain and frequency domain feature parameters into a feature vector:

[0084] F = [RMS, CF, K, A1, A2, ..., An, R1, R2, ..., Rn]

[0085] Furthermore, the fault diagnosis sub-module specifically includes the following:

[0086] First, based on historical data, establish a fault diagnosis rule base, for example:

[0087] If 1.5 < CF < 4 and K > 3, there may be an early bearing fault; if A2 / A1 > 0.5, there may be an imbalance fault.

[0088] Preferably, use Support Vector Machine (SVM) for fault classification:

[0089] For the feature vector F, solve the optimization problem:

[0090] min(1 / 2 × ||w|| 2 + C × ∑ξi)

[0091]

[0092] where w is the weight vector, b is the bias, C is the penalty parameter, is the kernel function, ξ is the slack variable used to handle noise or inseparable samples in the dataset, yi is the class label indicating the true class of each sample, and t is the time step.

[0093] Furthermore, the machine learning sub-module specifically includes the following:

[0094] Using the decision tree ensemble method, the present invention uses random forest for fault diagnosis, including the following steps:

[0095] Construct multiple decision trees, each tree is trained based on a randomly selected subset of features, and the final diagnosis result is determined by majority voting.

[0096] Based on the changing trend of the characteristic parameters, evaluate the severity of the fault:

[0097] Severity = α * ΔRMS + β * ΔCF + γ * ΔK

[0098] where α, β, γ are the weight coefficients respectively, and Δ represents the change amount of the parameter relative to the baseline value.

[0099] Combining the results of multiple diagnostic methods, calculate the diagnostic confidence:

[0100] Confidence = (N svm + Nr f + N rule ) / (3 × N total)

[0101] Among them, N svm ,Nr f ,N rule are the number of consistent samples diagnosed by SVM, random forest and rule base, respectively, N total is the total number of samples.

[0102] A diagnostic report is generated based on the above content, including the detected fault type, severity, diagnostic result credibility, key characteristic parameters and their changing trends. The diagnostic results are fed back to the characteristic parameter calculation submodule for optimizing feature selection, including:

[0103] Calculate the contribution of each characteristic parameter to the diagnosis result, adjust the weight of the characteristic parameter, and enhance the weight of the parameter that is sensitive to the fault.

[0104] Furthermore, the fault diagnosis threshold is dynamically adjusted based on the equipment operating status and environmental factors:

[0105] Threshold new =Threshold old ×(1+λ×ΔOperating Condition )

[0106] Among them, λ is the learning rate, ΔOperating Condition is the change in operating conditions.

[0107] Furthermore, the adaptive model update submodule specifically includes the following contents:

[0108] Incremental learning methods such as Online Stochastic Gradient Descent (Online SGD) are used to update model parameters.

[0109] The parameter update formula is:

[0110]

[0111] Among them, θt is the model parameter at time t, η is the learning rate, L is the loss function, xt and yt are the new input data and labels respectively.

[0112] Combining the newly learned knowledge with the existing model can use techniques such as ensemble methods or parameter averaging.

[0113] The present invention uses weighted average:

[0114] fnew(x)=α 0 ×fold(x)+(1-α 0 )×fupdate(x)

[0115] Among them, fnew is the updated model, fold is the original model, fupdate is the model trained based on new data, α 0 is the weight coefficient.

[0116] Furthermore, monitor model performance and detect changes in data distribution. Specifically, if obvious concept drift is detected, it may be necessary to retrain the entire model or adjust the model structure.

[0117] Furthermore, the automatic storage submodule specifically includes the following contents:

[0118] Continuously monitor the threshold of the speed channel to determine the operating status of the unit (such as startup, operation, shutdown, etc.). When it is determined that the equipment is in operation, automatically start the data storage process, select key parameters and diagnostic results for recording, and ensure that all relevant data are stored completely.

[0119] The collected data is organized according to timestamps to ensure the continuity and integrity of the data. According to the set storage strategy, the data is regularly transferred from the temporary storage area to the long-term storage area. During the data storage process, data verification is performed to ensure the accuracy and integrity of the data and prevent data loss or tampering.

[0120] It can be seen that the system is designed with an automatic storage function, which determines whether the unit is in operation based on the threshold of the speed channel. When it is judged to be in operation, it automatically triggers data storage. This data preprocessing method effectively improves the accuracy and reliability of the data, making fault data storage and analysis more intelligent and efficient.

[0121] Furthermore, the dual storage mechanism submodule specifically includes the following contents:

[0122] The vibration data and analysis results generated by the high-frequency real-time acquisition submodule and other analysis submodules are stored in real time in the local storage device to ensure fast access and data security.

[0123] Utilize network connections to transmit real-time stored data to remote hyper-converged servers, using data compression and transmission optimization algorithms to ensure efficient and secure data transmission.

[0124] Back up data on the hyper-converged server to keep local storage and remote storage data synchronized. Check data consistency regularly to prevent data mismatch or loss, and simultaneously establish a fault recovery mechanism to ensure that when a local storage device or network transmission fails, data can be quickly recovered from the hyper-converged server to avoid data loss.

[0125] It can be seen that the system adopts a dual storage mechanism. Whenever the industrial computer collects new vibration data, the data will be written to the local storage in real time and transmitted to the hyper-converged server through the network. This strategy ensures that even if the hyper-converged device fails, the industrial computer can still retain the data. It effectively ensures data backup and integrity and prevents data loss caused by hyper-converged server failure.

[0126] Furthermore, the security protection module specifically includes the following contents:

[0127] Deploy a one-way network firewall to physically isolate the internal network from the external network to prevent external network attacks from threatening the data and equipment of the vibration analysis system.

[0128] Continuously monitor data flow, identify abnormal data transmission behavior, and trigger security alarms when anomalies are detected.

[0129] According to the power safety level protection requirements, set access rights for different users and modules to ensure that only authorized personnel can access sensitive data and perform key operations.

[0130] Furthermore, this embodiment also provides a combustion turbine rotor vibration analysis method based on vibration analysis and machine learning technology, comprising the following steps:

[0131] Start the high-frequency real-time acquisition submodule to continuously monitor the vibration data of the equipment during operation, and capture the time domain signal of the equipment vibration through the vibration sensor; input the collected time domain vibration data into the frequency domain analysis submodule, perform Fourier transform on the data, generate frequency domain data, analyze the frequency distribution characteristics, and identify specific frequencies or abnormal frequencies and their amplitudes; use the axis trajectory drawing submodule to process the data from the orthogonal axis vibration sensor, generate an axis trajectory diagram, and identify the potential abnormal state of the rotor system by analyzing the shape and characteristics of the axis trajectory; use the characteristic parameter calculation submodule to calculate the key vibration characteristic parameters based on the axis trajectory and frequency domain analysis results; input the characteristic parameters into the fault diagnosis submodule, perform preliminary fault identification based on the preset fault judgment criteria, intelligently diagnose the health status of the equipment, determine whether there is an abnormality and preliminarily classify the fault type; in the machine learning submodule, use the previously collected and analyzed The vibration data training diagnosis model is used to continuously optimize and improve the accuracy of the fault diagnosis model through the combination of historical data and real-time data; the adaptive model update submodule is enabled to use new monitoring data for online learning and model update to ensure that the diagnosis model can adapt to changes in the equipment status and maintain efficient fault classification capabilities; the automatic storage submodule is activated to determine whether the equipment is in operation based on the threshold of the speed channel. If the equipment is running, the data storage function is automatically triggered to ensure the continuity of the vibration data; while storing data, the dual storage mechanism submodule is used to write real-time vibration data to local storage, and transmit it to the hyper-converged server through the network for backup to ensure data integrity and security; physical isolation is implemented through the security protection module, and a one-way network firewall is used to ensure the safe transmission and storage of vibration data, prevent external network attacks and data leakage, and ensure the overall security of the system.

[0132] This embodiment also provides a computer device, which is suitable for the case of a gas turbine rotor vibration analysis system based on vibration analysis and machine learning technology, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the gas turbine rotor vibration analysis system based on vibration analysis and machine learning technology as proposed in the above embodiment.

[0133] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0134] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, a combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as proposed in the above embodiment is implemented.

[0135] In summary, the present invention adopts domestically produced high-precision data acquisition equipment, reduces dependence on imported equipment, saves costs, and enhances the autonomy and maintainability of the system; the present invention is designed with an automatic storage function, which effectively improves the accuracy and reliability of data, making fault data storage and analysis more intelligent and efficient; the system has a variety of data analysis capabilities, and generates multi-dimensional data displays such as spectrum diagrams and axis trajectory diagrams in real time, which can improve the intuitiveness and accuracy of fault diagnosis. In addition, the physical isolation of a one-way network switch is adopted to ensure the data security and protection of the production area in Zone 1, prevent external network attacks, and meet the requirements of power safety level protection. The system of the present invention has high system integration, powerful data processing and analysis functions, and is easy to operate. Users can view and manage all acquisition and analysis results in real time on the monitoring interface, which improves the system's operating efficiency and user experience.

[0136] Example 2

[0137] Reference Figure 3 , which is the second embodiment of the present invention, provides a combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0138] According to the on-site conditions of Nanjing gas turbine power plant, the layout scheme of 9FA gas turbine rotor vibration monitoring and diagnosis system was designed.

[0139] The 9FA engine has two bearings, located at the front end of the compressor and the rear end of the turbine. Each bearing is equipped with four onboard vibration measuring points, two of which are eddy current displacement sensors (shaft vibration), which are installed vertically to each other at an angle of 45° and 135° from the horizontal plane. Two are acceleration sensors (bearing vibration), which are installed adjacently just above the bearings. Two axial displacement sensors and one key phase sensor are also installed at the compressor bearings.

[0140] At present, the displacement sensor signal and key phase sensor signal are connected to the Bentley 3500 vibration data collector, which has an OEM-approved data interface reserved for third-party equipment. This system collects and post-processes vibration signals through this interface.

[0141] The data acquisition instrument includes: the data acquisition instrument has 24 independent 24-bit AD channels, which can collect 20 vibration signals (including displacement and acceleration channels), 1 key phase and 1 speed signal in real time and synchronously. At the same time, the chassis has an expandable slot (16 channels) for subsequent access to vibration data; the sampling frequency of the acquisition instrument is adjustable, up to 20kHz; the output signal of Bentley 3500 is a voltage signal of ±20V, and the input range of the acquisition instrument is ±20V. The data acquisition accuracy within the range is better than ±0.05%; the data acquisition instrument has the ability to work continuously and stably for a long time, and the continuous working time is not less than 12 months; the acquisition instrument must have the function of resisting strong electromagnetic interference (national standard level 4).

[0142] Industrial computer parameters include:

[0143] CPU: Intel i5-14600K;

[0144] Memory: 32G DDR5;

[0145] Hard disk: 1T solid state hard disk + 8T mechanical hard disk;

[0146] System software parameters include:

[0147] Acquisition software configuration: You can set parameters such as sensor type, sensor number, and sensor sensitivity for each channel.

[0148] The automatic start and end of storage can be achieved according to the threshold of the set speed channel.

[0149] Data processing and analysis: The spectrum of vibration data can be calculated and analyzed in real time, and 30 characteristic parameters such as RMS (mean square error), n-fold frequency amplitude, and n-fold frequency amplitude ratio of all vibration channels can be calculated.

[0150] like Figure 3As shown in the figure, this system includes domestic data collectors, industrial computers, photoelectric converters, connecting optical fibers (network cables), switches, one-way gateways, hyper-converged servers and other equipment. The system connects the signal to the domestic data collector from the third-party data interface, and the original data is stored and sent in the industrial computer. The data enters the second area of ​​the power plant through the switch, photoelectric converter and one-way gateway.

[0151] The data is then transmitted to the hyper-converged server located in the third area of ​​the power plant through the existing optical fiber channel of the power plant. The original data is stored again and analyzed and processed in the hyper-converged server, including spectrum analysis, axis trajectory analysis, characteristic value trend analysis and characteristic indicator fault diagnosis, fault diagnosis analysis based on machine learning, etc. Finally, the analyzed and processed data is sent to the rotor vibration monitoring and diagnosis interface for display through the third-party TCP protocol. Among them, the original data stored on the industrial computer is used for backup to prevent data loss caused by equipment failures such as switches and one-way network switches.

[0152] Among them, the signals that the vibration data collector can output also include the displacement sensor signals at the bearings of the turbine and generator of the combined cycle unit. Therefore, this system can also monitor the shaft system vibration of the 9FA gas turbine and the combined cycle unit, and a fault diagnosis module for the shaft system vibration of the unit can be developed in the later stage.

[0153] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A gas turbine rotor vibration analysis system based on vibration analysis and machine learning technology, characterized by: include, Data acquisition and analysis module, including high-frequency real-time acquisition submodule, used to collect vibration data during equipment operation; The frequency domain analysis submodule is used to perform frequency domain transformation on the collected time domain data, conduct frequency distribution characteristic analysis, and identify the occurrence and amplitude of specific frequencies or abnormal frequencies; The axis trajectory and characteristic parameter analysis module includes an axis trajectory drawing submodule, which is used to process the data collected by the orthogonal axis vibration eddy current sensor and identify the abnormal state of the rotor system by analyzing the shape and characteristics of the axis trajectory; The characteristic parameter calculation submodule is used to calculate key vibration characteristic parameters and provide rich diagnostic data support; Fault diagnosis and learning module, including fault diagnosis submodule, used to establish fault judgment criteria and perform intelligent diagnosis; The machine learning submodule is used to train the diagnostic model using monitoring data and perform fault classification; The adaptive model update submodule is used to integrate online learning capabilities and iteratively optimize the model using continuously updated monitoring data; The storage module includes an automatic storage submodule, which is used to determine whether the unit is in operation according to the threshold value of the speed channel, and automatically trigger data storage when it is determined to be in operation; Dual storage mechanism submodule, which is used to write the collected vibration data to local storage in real time and transmit it to the hyper-converged server through the network to ensure data backup and integrity; The security protection module is used to ensure data security and protection through physical isolation of a one-way network firewall.

2. The combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as claimed in claim 1, characterized in that: The frequency domain analysis submodule includes: De-noising the collected time domain data; Segment the data to ensure that the length of each data segment is suitable for frequency domain transformation; Use FFT algorithm to convert time domain data into frequency domain data and calculate power spectrum density; Identify the main frequency components and calculate the amplitude and phase information of each frequency component; According to the equipment characteristics, the key characteristic frequency is identified, and the amplitude change of the key characteristic frequency is monitored.

3. The combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as claimed in claim 2, characterized in that: The axis trajectory drawing submodule specifically includes the following contents: Ensure synchronous data collection of eddy current sensors in the X and Y directions, time align the collected data, and apply a bandpass filter to remove high-frequency noise and low-frequency drift; Convert the original voltage signal of the sensor into displacement information, perform coordinate system calibration, and ensure the orthogonality of the XY coordinate system; Use XY displacement data to draw the axis trajectory, generate static axis trajectory diagram and dynamic trajectory animation; Calculate the geometric characteristics of the trajectory and analyze the directionality and rotation characteristics of the trajectory; Classify based on trajectory shapes and identify abnormal trajectory patterns.

4. The combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as claimed in claim 3, characterized in that: The specific operation steps of analyzing the directionality of the trajectory are as follows: Calculate the instantaneous velocity vector of each sampling point. i , b i ), and the next point (a i+1 , b i+1 ), velocity vector v i It is expressed as: v i =(a i+1 -a i ,b i+1 -b i ) The velocity vector v i Indicates the direction and speed of the trajectory; For a continuous velocity vector v i and v i+1 , calculate the cross product v i ×v i+1 , calculated by the following formula: v i *v i+1 =(a i *b i+1 -b i *a i+1 )。 5. The combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as claimed in claim 4, characterized in that: The sign of the cross product is used to determine the direction of rotation: If the cross product is positive, it means the trajectory is rotating counterclockwise; If the cross product is negative, it means the trajectory is rotating in a clockwise direction.

6. The combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as claimed in claim 5, characterized in that: The specific steps for analyzing the rotation characteristics are as follows: Calculate the instantaneous angular velocity of the trajectory: The instantaneous angular velocity of the trajectory is estimated by the rate of change of the trajectory points. Given a continuous point (a i , b i ), and the next point (a i+1 , b i+1 ), angular velocity ω i Calculate using: Among them, θ i is a point (a i , b i ) is the angle relative to the center point, Δt is the time interval; Calculate the mean and standard deviation of the angular velocity. If the standard deviation of the angular velocity is lower than the preset secondary threshold, it indicates that the rotation is stable and the operation state of the mechanical equipment is relatively stable. If the standard deviation of the angular velocity is higher than a preset first-level threshold, it indicates that the rotation is unstable.

7. The combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as claimed in claim 6, characterized in that: The machine learning submodule specifically includes the following contents: Fault diagnosis using random forests includes the following steps: Construct multiple decision trees, each of which is trained based on a randomly selected subset of features, and the final diagnosis result is determined by majority voting; Evaluate the severity of the fault based on the changing trend of characteristic parameters: Severity=α*ΔRMS+β*ΔCF+γ*ΔK Among them, α, β, γ are weight coefficients, Δ represents the change of the parameter relative to the baseline value, RMS is the root mean square value, CF is the peak factor, and K is the kurtosis; Combining the results of multiple diagnostic methods, the diagnostic confidence is calculated: Confidence=(N svm +Nr f +N rule ) / (3×N total ) Among them, N svm ,Nr f ,N rule are the number of consistent samples diagnosed by SVM, random forest and rule base, respectively, N total is the total sample number; A diagnostic report is generated based on the above content, including the detected fault type, severity, diagnostic result credibility, key characteristic parameters and their changing trends. The diagnostic results are fed back to the characteristic parameter calculation submodule to optimize feature selection.

8. A method for analyzing the vibration of a gas turbine rotor based on vibration analysis and machine learning technology, based on a system for analyzing the vibration of a gas turbine rotor based on vibration analysis and machine learning technology as claimed in any one of claims 1 to 7, characterized in that: Also includes, Start the high-frequency real-time acquisition submodule to continuously monitor the vibration data of the equipment during operation and capture the time domain signal of the equipment vibration through the vibration sensor; The collected time-domain vibration data is input into the frequency-domain analysis submodule, the data is Fourier transformed, the frequency-domain data is generated, the frequency distribution characteristics are analyzed, and specific frequencies or abnormal frequencies and their amplitudes are identified; The axis trajectory drawing submodule processes the data from the orthogonal axis vibration sensor to generate an axis trajectory map. By analyzing the shape and characteristics of the axis trajectory, the potential abnormal state of the rotor system can be identified. Use the characteristic parameter calculation submodule to calculate key vibration characteristic parameters based on the axis trajectory and frequency domain analysis results; Input the characteristic parameters into the fault diagnosis submodule, perform preliminary fault identification according to the preset fault judgment criteria, intelligently diagnose the health status of the equipment, determine whether there is an abnormality and preliminarily classify the fault type; In the machine learning submodule, the previously collected and analyzed vibration data is used to train the diagnostic model. By combining historical data with real-time data, the accuracy of the fault diagnosis model is continuously optimized and improved. Enable the adaptive model update submodule to use new monitoring data for online learning and model updates to ensure that the diagnostic model can adapt to changes in equipment status and maintain efficient fault classification capabilities; Activate the automatic storage submodule to determine whether the device is in operation based on the threshold of the speed channel. If the device is in operation, the data storage function is automatically triggered to ensure the continuity of vibration data. While storing data, the dual storage mechanism submodule is used to write real-time vibration data into local storage and transmit it to the hyper-converged server through the network for backup, ensuring the integrity and security of the data; Physical isolation is implemented through the security protection module, and a one-way network gate is used to ensure the safe transmission and storage of vibration data, prevent external network attacks and data leakage, and ensure the overall security of the system.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the combustion turbine rotor vibration analysis system based on vibration analysis and machine learning technology as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the combustion engine rotor vibration analysis system based on vibration analysis and machine learning technology as described in any one of claims 1 to 7 are implemented.

Images