Fan equipment fault automatic diagnosis method and device

Abstract

The invention discloses a fan equipment fault automatic diagnosis method and device. Collected fan equipment operation data is divided into low-frequency characteristic data and high-frequency vibration data, and the low-frequency characteristic data and the high-frequency vibration data are subjected to cleaning, noise reduction, characteristic extraction and dimension reduction processing respectively; a fault diagnosis model is constructed, and fault mode classification is carried out, so that automatic diagnosis and identification of fan equipment faults can be realized; and storing data for updating a fault diagnosis model data set after each diagnosis or test. And a multi-parameter diagnosis conclusion can be quickly given, the spot inspection efficiency on site is improved, and the manpower maintenance and shutdown cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of fan fault diagnosis, and in particular to a fan fault diagnosis method and device based on fault signal analysis. Background Art

[0002] Fans are widely used in industrial production, and their operating status directly affects production efficiency and safety. Fan failure may lead to production interruption, equipment damage or even safety accidents, so accurate fault diagnosis of fans is of great significance. Traditional fan fault diagnosis methods mostly rely on manual experience, and judge faults by observing the operating status of the equipment and simple vibration measurements. This method has problems such as strong subjectivity, low accuracy, and inability to detect potential faults in time. With the development of signal processing technology and machine learning algorithms, fan fault diagnosis methods based on vibration signal analysis have gradually become a research hotspot. However, the existing fault diagnosis methods based on vibration signal analysis still have some problems in practical applications, such as insufficient comprehensive signal feature extraction, inaccurate distinction between different fault types, and inability to adapt to complex working conditions.

[0003] In addition, in the process of fault diagnosis of rotating equipment, the diagnostic devices usually used by equipment maintenance personnel can only measure vibration values ​​or display frequency spectrum, and these data are only part of the information of the equipment operation status. It cannot fully reflect the operation status of the fan equipment and cannot directly draw conclusions about the health status of the equipment.

[0004] Secondly, the existing technology cannot realize automatic fault diagnosis and identification, cannot directly determine whether the equipment is running well, and cannot clearly identify the specific fault type and location of the equipment. This requires equipment maintenance personnel to rely on their own experience and judgment, and rely on manual analysis and judgment when making a diagnosis. The process is cumbersome and the accuracy depends on the operator's experience level. The influence of subjective factors is increased, which may lead to misdiagnosis or missed diagnosis.

[0005] Thirdly, most diagnostic devices cannot accurately detect the operating status during measurement, and often require manual operation and subsequent data processing and analysis, resulting in delayed fault detection and diagnosis. The inability to analyze and feedback the status of the equipment in real time and comprehensively can easily lead to untimely fault warnings when the equipment needs to run continuously for a long time, thus affecting production efficiency and safety.

[0006] Finally, the predictability is insufficient. The inability to provide timely warnings and predictions of potential failures can easily lead to equipment being discovered only after the failure has worsened, increasing maintenance and downtime costs. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide a method and device for automatically diagnosing fan equipment faults, which can quickly give multi-parameter diagnostic conclusions for fan equipment, improve on-site inspection efficiency, and reduce manpower maintenance and downtime costs. This avoids the problem that general equipment diagnostic methods and devices are not applicable, accurate, and comprehensive for fan fault troubleshooting, and require a lot of manual intervention.

[0008] In order to solve the above technical problems, the technical solution adopted by the present invention is: A method for automatically diagnosing a fault of a fan device comprises at least one of the following diagnostic steps: The collected fan equipment operation data is divided into low-frequency characteristics and high-frequency vibration data, and then cleaned, denoised, feature extracted and dimension reduced respectively; Construct a fault diagnosis model and classify fault modes to achieve automatic diagnosis and identification of fan equipment faults; After each diagnosis or test, data is saved to update the fault diagnosis model data set; The fan equipment operation data includes process data, vibration data, temperature data, environmental parameters and speed parameters.

[0009] In the above technical solution, data is saved after each diagnosis or test, samples are added to the fault diagnosis model sample library, a model update method is selected, and the accuracy of the fault diagnosis model is gradually iterated.

[0010] In the above technical solution, after each diagnosis or test, the validity of the test result is confirmed based on the data label, and then the diagnosis or test data is saved to the data set.

[0011] In the above technical solution, the model update method includes three modes: manual update training triggered by humans, update training based on accuracy, and regular update training based on time periods.

[0012] In the above technical solution, for low-frequency eigenvalue data, the Z-score method is used to identify outliers, and linear interpolation is used to fill the data.

[0013] In the above technical solution, for high-frequency vibration data, low-pass filtering is first performed to remove high-frequency noise, and then adaptive filtering is used to further suppress dynamic noise.

[0014] In the above technical solution, during feature extraction and dimensionality reduction, the features of the above process data, vibration data, temperature data, environmental parameters, and operating parameters are extracted, and PCA is used for feature selection.

[0015] The vibration signals include horizontal, vertical and axial vibration signals, and the temperature signals include bearing temperature and other key parts temperature signals.

[0016] In the above technical solution, a variety of characteristic parameters are extracted from the preprocessed vibration signal, including time domain characteristics (such as root mean square value, peak value, kurtosis value, etc.), frequency domain characteristics (such as amplitude and frequency of each frequency band, etc.) and time-frequency domain characteristics (such as wavelet transform coefficients, etc.).

[0017] In the above technical solution, during feature extraction and dimensionality reduction, the characteristic values ​​of each signal obtained include: the peak value and effective value of the acceleration time domain signal per second, and the frequency multiple amplitude, 2 times frequency amplitude, 3 times frequency amplitude, 4 times frequency amplitude, and 5 times frequency amplitude of the acceleration signal spectrum, the peak value and effective value of the speed time domain signal per second, the peak value and effective value of the acceleration envelope signal per second, and the amplitude of the envelope signal spectrum at the fault frequency, and the energy of the acceleration signal spectrum in the low frequency band, medium frequency band, and high frequency band; the peak value and effective value of the current time domain signal per second, and the frequency multiple amplitude, 2 times frequency amplitude, 3 times frequency amplitude, 4 times frequency amplitude, and 5 times frequency amplitude of the current signal spectrum; the average value of the temperature signal per second.

[0018] In the above technical solution, a fault diagnosis model is constructed based on the K nearest neighbor classification method.

[0019] In the above technical solution, the fault diagnosis model is based on the support vector machine algorithm and is trained using a large amount of historical fault data. It can identify common fan fault types such as imbalance, misalignment, rolling bearing fault, loose bearing seat, etc. The model learns the characteristic patterns corresponding to different fault types, classifies and judges the input characteristic parameters, and outputs preliminary fault diagnosis results.

[0020] In the above technical solution, when constructing a fault diagnosis model and classifying fault modes, the following steps are specifically included: Divide the data set: Divide the feature data set after dimensionality reduction, randomly select more than half of the feature samples as training points, and the remaining feature samples as test samples, and set the K value; Calculate the Euclidean distance from the sample to be tested to each training point; Sort each distance and select the K training points closest to the sample to be tested; Find out the equipment failure mode category to which the selected K nearest neighbor training points belong; According to the principle of minority obeys majority, the samples to be tested are classified into the equipment failure mode with the highest proportion among K training points to complete the failure mode classification.

[0021] In the above technical solution, for the situation where a fault is judged, the severity and development trend of the fault are further analyzed. On the one hand, the extracted characteristic parameters are compared with the preset threshold value. When the characteristic parameters exceed a certain threshold value, it is considered that the degree of the fault has increased. On the other hand, the characteristic parameters are trend analyzed to observe their changing trend over time and predict the development direction and speed of the fault. At the same time, combined with the operation time, maintenance history and other information of the fan, the severity of the fault is comprehensively evaluated to provide a basis for subsequent maintenance work.

[0022] In the above technical solution, specific maintenance suggestions and repair strategies are given based on the fault diagnosis results and severity assessment. For example, for minor imbalance faults, it is recommended to perform simple dynamic balancing adjustments when the equipment is shut down; for serious bearing faults, it is necessary to arrange shutdown and bearing replacement as soon as possible. Finally, the diagnosis results, maintenance suggestions and other information are organized into report outputs and stored in the database for subsequent query and analysis, providing data support for the full life cycle management of the equipment.

[0023] A fan equipment fault automatic diagnosis device, comprising: The data acquisition module is used to collect various operating data of the fan equipment, including process data, vibration data, temperature data, environmental parameters and rotation speed; the process data collected include: damper opening, fan inlet and outlet pressure, fan inlet and outlet temperature, lubricating oil pressure and temperature; The computing module is used to clean and reduce noise, extract features and reduce the dimension of the collected data, and classify fault modes using the feature data after dimension reduction to achieve automatic diagnosis and identification; The model training and updating module is used to build a fault diagnosis model based on the K-nearest neighbor classification method, and to confirm and save the data after each diagnosis or test and update it to the sample library. The accuracy of the fault diagnosis model is gradually iterated by manually triggering training, updating based on accuracy, or conducting regular training based on time periods.

[0024] The modules work closely together to complete the diagnosis of fan failures.

[0025] In the above technical solution, the data acquisition module includes a built-in data access module, a data acquisition module and an environmental parameter acquisition module, and a configuration module is set for users to select appropriate initial parameters of the equipment; the data access module is used to acquire process data; the data acquisition module is used to collect vibration data, temperature data and rotation speed; the environmental parameter acquisition module is used to collect real-time environmental humidity parameters.

[0026] In the above technical solution, the configuration module is used to configure the initial parameters of the equipment, including the equipment type, bearing model, number of gear teeth, and number of impeller blades.

[0027] In the above technical solution, the configuration module is preferably configured as an input device or interface.

[0028] In the above technical solution, the data access module is built into the diagnostic device and is configured with relevant information of the on-site process data server for connecting to external devices and signals according to standard protocols to obtain the process data of the equipment in real time.

[0029] In the above technical solution, the data acquisition module is configured as follows: an acceleration sensor is installed on each bearing seat of the fan, connected to the inspection instrument through an aerial plug, and the vibration data of the equipment is collected.

[0030] In the above technical solution, the data acquisition module is configured as follows: a temperature sensor is installed on each bearing seat of the fan to collect temperature data in the same way as vibration.

[0031] In the above technical solution, the data acquisition module is configured to use a speed sensor to collect the working speed of the equipment as a speed parameter.

[0032] That is, the data acquisition module is configured to install a rotation speed sensor, an acceleration sensor, and a temperature sensor on each bearing seat of the fan when collecting vibration data.

[0033] Preferably, the rotation speed sensor and temperature sensor are configured to be connected to the fan equipment fault diagnosis device through an aviation plug to collect equipment vibration data.

[0034] In addition, the present invention also provides a storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of the above-mentioned fan fault diagnosis method can be implemented. By installing the program in the storage medium on a device, the device can have the function of fan fault diagnosis, and realize real-time monitoring of the fan operating status and automatic diagnosis of faults.

[0035] Compared with the prior art, the beneficial effects of the present invention are: Although vibration analysis is a mature fault detection technology, existing vibration detection equipment still has many limitations, especially in the automation and intelligence of fault diagnosis. Most existing detection and diagnosis devices can only provide basic vibration data, such as acceleration, velocity and displacement, or simple spectrum diagrams. Although these data provide experts with clues to the operating status of the equipment, it is difficult to automatically and accurately determine the type of fault and its location without the participation of professionals. Therefore, existing equipment status monitoring technology often needs to rely on experienced professionals to perform complex spectrum analysis and judgment. Such manual diagnosis is not only time-consuming and labor-intensive, but also easily affected by subjective factors. The device of the present invention has built-in multiple eigenvalue calculation methods and mechanism diagnosis rules. From feature extraction to fault self-diagnosis, it can quickly give a diagnostic conclusion, improve on-site inspection efficiency, and reduce manpower.

[0036] At present, the better diagnostic devices on the market, such as Emerson CSI2140 series products, only have the functions of vibration data collection, vibration characteristic value calculation, and waveform presentation. Specific analysis still requires manual analysis of the spectrum to locate the fault. In addition, VM-63A of RION is also a popular product, but it only has three characteristic value display functions, which can only display the size of vibration and cannot achieve accurate analysis of faults.

[0037] The automatic fault diagnosis device for fan equipment of the present invention has multiple built-in data collection and processing methods, and is specially designed and developed for fan equipment, and is used to collect multiple operating data of fan equipment, including process data, vibration data, temperature data, environmental parameters and speed parameters; it can process the data including the damper opening of the equipment, the inlet and outlet pressure of the fan, the inlet and outlet temperature of the fan, the lubricating oil pressure and temperature, and can judge and process more faults more accurately and comprehensively. Through multi-dimensional signal acquisition and comprehensive feature extraction, the present invention can more comprehensively reflect the operating status of the fan and improve the accuracy of fault diagnosis. It can accurately identify the fault type of the fan, evaluate the severity and development trend of the fault, and give targeted maintenance suggestions, which effectively improves the operating reliability and maintenance efficiency of the fan equipment, reduces the losses and risks caused by equipment failures, has significant economic and social benefits, and can be widely used in the field of fan equipment management in various industrial productions.

[0038] Compared with general intelligent diagnostic tools, the automatic fault diagnosis device for fan equipment of the present invention sets a diagnostic model for the fan, and sets a self-update module to match the diagnostic model, and sets a manual update mode, an accuracy update mode, and a time update mode to select the update mode. The model sample library can be updated based on the test results, thereby enhancing the diagnostic capability in the model library, so that the equipment has a self-learning function, and the diagnostic accuracy can be gradually iterated.

[0039] It can not only determine the type of fault, but also evaluate the severity and development trend of the fault, provide more targeted suggestions for equipment maintenance, and effectively reduce equipment downtime and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 It is a flow chart of an embodiment of a method for automatically diagnosing faults of a fan device according to the present invention.

[0041] Figure 2 It is a working principle diagram of a data acquisition module in the automatic diagnosis device for fan equipment faults of the present invention.

[0042] Figure 3It is a working principle diagram of the calculation module in the automatic diagnosis device for fan equipment faults of the present invention.

[0043] Figure 4 It is a working principle diagram of the model training and updating module in the automatic diagnosis device for fan equipment faults of the present invention.

[0044] Figure 5 The present invention is a working principle diagram of a configuration module included in a data acquisition module in the automatic diagnosis device for fan equipment faults.

[0045] Figure 6 This is a diagram of data collected during actual diagnosis according to an embodiment of the present invention (before noise reduction).

[0046] Figure 7 yes Figure 6 Schematic diagram of data after denoising.

[0047] Figure 8 yes Figure 7 Data graph after 2D denoising. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] Example 1 This embodiment provides a method for automatically diagnosing a fault of a fan device, the process of which is shown in FIG1 , and the specific steps are as follows: Multiple sensors installed on the fan are used to collect operating data such as vibration signals, speed, temperature, humidity signals, etc. Vibration signals are collected from the horizontal, vertical and axial directions to fully reflect the vibration status of the fan; temperature signals focus on collecting the temperature of key parts such as bearings, so as to promptly detect abnormal conditions such as overheating.

[0050] The collected signals are preprocessed. First, filtering and denoising are performed to remove high-frequency noise and interference components in the signal and improve the signal-to-noise ratio; then normalization is performed to unify data of different dimensions and ranges into the same range to facilitate subsequent feature extraction and analysis.

[0051] Extract multiple characteristic parameters from the preprocessed vibration signal. In the time domain, calculate the statistical features such as the root mean square value, peak value, and kurtosis value, which can reflect the overall energy and impact characteristics of the signal; in the frequency domain, convert the signal to the frequency domain through fast Fourier transform (FFT), extract the amplitude and frequency information of each frequency band, and identify different vibration frequency components; in the time-frequency domain, use methods such as wavelet transform to obtain the time-frequency domain characteristics of the signal, which can capture the transient changes and local characteristics of the signal.

[0052] The extracted feature parameters are input into a pre-trained fault diagnosis model, which is based on the support vector machine algorithm and trained using a large amount of historical fault data. It can identify common fan fault types such as imbalance, misalignment, rolling bearing fault, and loose bearing seat. The model learns the feature patterns corresponding to different fault types, classifies and judges the input feature parameters, and outputs preliminary fault diagnosis results.

[0053] For situations where a fault is determined to exist, the severity and development trend of the fault are further analyzed. On the one hand, the extracted characteristic parameters are compared with the preset thresholds. When the characteristic parameters exceed a certain threshold, the fault is considered to be aggravated. On the other hand, trend analysis is performed on the characteristic parameters to observe their changing trends over time and predict the direction and speed of the fault. At the same time, combined with the operation time, maintenance history and other information of the fan, the severity of the fault is comprehensively evaluated to provide a basis for subsequent maintenance work.

[0054] For example, when diagnosing imbalance, based on the ratio of 1st frequency to 2nd frequency, it can be determined whether the fan rotor is dynamically unbalanced.

[0055] Diagnose misalignment based on the ratio of 2 times the frequency to 1 times the frequency; it can be determined whether the fan rotor is poorly aligned.

[0056] As a preference, this embodiment can also provide specific maintenance suggestions and repair strategies based on the fault diagnosis results and severity assessment. For example, for minor imbalance faults, it can be recommended to perform simple dynamic balancing adjustments when the equipment is shut down; for serious bearing faults, it is necessary to arrange to shut down and replace the bearings as soon as possible. Finally, the diagnostic results, maintenance suggestions and other information are organized into report outputs and stored in the database for subsequent query and analysis, providing data support for the full life cycle management of the equipment.

[0057] For example, set the suggested matching rules in the model: If the shaft vibration difference on the intake side is large, check the condition of the vibration probe on the intake side; The shaft vibration difference on the outlet side is large, check the vibration probe on the outlet side; If the bearing temperature difference is large, it is recommended to check the bearing temperature probe.

[0058] The equipment base is loose. It is recommended to 1. Check the tightness of the bearing seat foundation and tighten the bolts as required; 2. Check whether the butterfly spring is deformed and the foundation is loose; 3. Check whether the frame is corroded, causing the bearing seat to vibrate.

[0059] Oil film vortex, check whether the bearing is worn, and adjust the bearing clearance and bearing oil supply pressure if necessary; Friction between moving and static parts, suggestions: 1. Adjust parameters to ensure uniform thermal expansion of the unit; 2. Inspect and adjust the concentricity of the rotor and stator; 3. Measure the horizontality of the foundation to eliminate the influence of settlement; 4. Check the performance of the unit to determine whether an overhaul is required.

[0060] The rotor is unbalanced. It is recommended to 1. steam purge and soak the impeller in ammonia water; 2. perform dynamic and static balancing; 3. check whether the blades are damaged or cracked.

[0061] The fan rotor is not well centered. Suggestions: 1. Check the coupling alignment deviation; 2. Check the concentricity of the shaft installation at both ends of the equipment; 3. Check whether the foundation has any settlement; 4. Identify whether the shaft is (thermally) bent by signal characteristics.

[0062] Example 2 The automatic fault diagnosis device for fan equipment of the present invention has built-in multiple characteristic value calculation methods and mechanism diagnosis rules, and can quickly give a diagnosis conclusion from feature extraction to fault self-diagnosis, thereby improving on-site inspection efficiency and reducing manpower.

[0063] The structure of the automatic fault diagnosis device for fan equipment of the present invention is as follows: Figure 1-Figure 5 As shown, the details are as follows: It includes data acquisition module, calculation module, model training and update module.

[0064] Among them, the data acquisition module is as follows Figure 2 As shown, it includes a data access module, a data acquisition module and an environmental parameter acquisition module, and a configuration module is set for users to select appropriate initial parameters of the equipment. The data acquisition module is used to collect various operating data of the fan equipment, including process data, vibration data, temperature data, environmental parameters and speed parameters. Among them, the data access module is built into the diagnostic device and is configured with relevant information of the on-site process data server. It is used to connect to external equipment and signals according to standard protocols and obtain the process data of the equipment in real time. The process data collected include: damper opening, fan inlet and outlet pressure, fan inlet and outlet temperature, lubricating oil pressure and temperature.

[0065] The data acquisition module is used to collect vibration data, temperature data and speed parameters.

[0066] Vibration data collection: Install acceleration sensors on each bearing seat of the fan, connect them to the inspection instrument through an aerial plug, and collect equipment vibration data.

[0067] Temperature data: Install temperature sensors on each bearing seat of the fan to collect temperature data in the same way as vibration.

[0068] The speed sensor is used to collect the working speed of the equipment as a speed parameter.

[0069] The environmental parameter acquisition module is composed of the temperature and humidity module built into the diagnostic device to collect real-time environmental humidity.

[0070] Computational modules such as Figure 3 As shown, it is used to clean and reduce noise, extract features and reduce dimension of the collected data, and classify fault modes using the feature data after dimension reduction to achieve automatic diagnosis and identification; Model training and updating modules such as Figure 4 As shown, it is used to build a fault diagnosis model based on the K nearest neighbor classification method, and to confirm and save the diagnosis or test data after each diagnosis or test and update it to the sample library, and gradually iterate the accuracy of the fault diagnosis model by manually triggering training, updating based on accuracy or performing regular training based on time periods.

[0071] Configuration modules such as Figure 5 As shown, the automatic fault diagnosis device for fan equipment of the present invention constitutes a part of the data acquisition module through a configuration module such as an input device or an input interface, which is used for allowing users to select appropriate initial parameters of the equipment, including equipment type, bearing model, number of gear teeth, and number of impeller blades.

[0072] Then install the speed sensor, temperature sensor and acceleration sensor on each bearing seat of the fan.

[0073] Preferably, the above-mentioned rotation speed sensor, temperature sensor and acceleration sensor are configured to be connected to the fan equipment fault diagnosis device through an aviation plug to collect equipment vibration data.

[0074] Example 3 like Figure 1-Figure 5 As shown, the automatic diagnosis method for fan equipment faults of the present invention can identify common fan faults such as imbalance, misalignment, rolling bearing faults, loose bearing seats, etc., and includes the following steps: The collected fan equipment operation data is divided into low-frequency characteristics and high-frequency vibration data, and then cleaned, denoised, feature extracted and dimension reduced respectively; Construct a fault diagnosis model and classify fault modes to achieve automatic diagnosis and identification of fan equipment faults; After each diagnosis or test, the diagnosis or test data is saved to update the fault diagnosis model data set.

[0075] The data collection process is as follows: The collected data include the equipment’s process data, vibration data, temperature data, environmental parameters, and speed parameters; The process data collected include: damper opening, fan inlet and outlet pressure, fan inlet and outlet temperature, lubricating oil pressure and temperature. The process data is realized by the built-in data access module of the diagnostic device. The data access module is configured with the relevant information of the on-site process data server to obtain the process data of the equipment in real time. Vibration data collection: Install acceleration sensors on each bearing seat of the fan, connect them to the inspection instrument through an aerial plug, and collect equipment vibration data.

[0076] Temperature data: Install temperature sensors on each bearing seat of the fan to collect temperature data in the same way as vibration. Environmental parameters: The diagnostic device has a built-in temperature and humidity module to collect real-time environmental parameters; The speed sensor is used to collect the working speed of the equipment as the working condition parameter.

[0077] The data cleaning and noise reduction process is as follows: Low-frequency eigenvalue data: Since low-frequency eigenvalue data is a slow variable data in the field, the data will not have sudden changes such as sudden jumps, so the interference (sudden jump data) in the identification process is reduced to reduce the interference to the diagnosis conclusion. The following data cleaning methods are provided: Use the Z-score method to identify outliers in feature data: The Z-score is the difference between a data point and the mean of the data set, expressed in units of standard deviation: ; in: It is a specific data point in the data set; is the mean of the data set; It is the standard deviation of the data set. From a statistical point of view, usually Z ≥ 3 or Z ≤ -3, the data point is considered an outlier.

[0078] After the outliers are identified, the data is filled using linear interpolation.

[0079] High-frequency vibration data: Various strong electric and magnetic field interferences in industrial sites will interfere with high-frequency vibration data. Therefore, noise reduction methods are used to filter out noise data in high-frequency vibration signals.

[0080] In the signal preprocessing stage, the collected vibration acceleration signal is first low-pass filtered to filter out high-frequency noise components while retaining low-frequency information related to mechanical failures. Then, dynamic noise is further suppressed through adaptive filtering, which adjusts filter parameters based on real-time feedback to ensure that the filtering effect can adapt to signal changes under different working conditions.

[0081] The feature selection and dimensionality reduction process is as follows: By extracting the features of the above-mentioned process data, vibration data, temperature data, environmental parameters, and operating parameters, and using PCA for feature selection; the characteristic values ​​of each signal obtained include: the peak value, effective value of the acceleration time domain signal per second, and the frequency multiple amplitude, 2 times amplitude, 3 times amplitude, 4 times amplitude, and 5 times amplitude of the acceleration signal spectrum, the peak value and effective value of the speed time domain signal per second, the peak value, effective value of the acceleration envelope signal per second, and the amplitude of the envelope signal spectrum at the fault frequency, the energy of the acceleration signal spectrum per second in the low frequency band, medium frequency band, and high frequency band; the peak value, effective value of the current time domain signal per second, and the frequency multiple amplitude, 2 times amplitude, 3 times amplitude, 4 times amplitude, and 5 times amplitude of the current signal spectrum; the average value of the temperature signal per second.

[0082] The model operation and diagnosis process is as follows: The output value of the previous step is input into the model, and the final result is generated to assist the point inspection personnel in operation and maintenance judgment. The model construction is based on the K nearest neighbor classification method.

[0083] The K nearest neighbor algorithm is used to classify the features after dimensionality reduction into classification modes, including: 1) Divide the data set; divide the feature data set after dimensionality reduction, randomly select 70% of the feature samples as training points, which are the sources of prior information for fault classification, and the remaining 30% of the feature samples as test samples, and set the K value; the above-mentioned K nearest neighbor algorithm is a classification method. The so-called K nearest neighbors refer to the K nearest neighbors. Each feature sample to be tested can be represented by the K training points closest to it; K: represents the number of neighboring training points selected in the classification process of the feature sample to be tested; 2) Calculate the Euclidean distance from the sample to be tested to each training point; 3) Sort each distance and select the K training points closest to the sample to be tested; 4) Find out the equipment failure mode category to which the selected K nearest neighbor training points belong; 5) According to the principle of minority obeys majority, the samples to be tested are classified into the type of equipment failure mode that accounts for the highest proportion among the K training points to complete the failure mode classification.

[0084] The model self-update process is as follows: After each diagnosis or test, the diagnosis or test data will be saved. The inspection personnel will compare the judgment result of the diagnostic device with the actual manual judgment result on site, and input the actual judgment result into the diagnostic device to form a diagnostic closed loop. The sample will be added to the sample library.

[0085] The diagnostic device is equipped with a variety of training methods, which can trigger training manually, train based on accuracy, or conduct regular training mechanisms based on time periods to gradually iterate the accuracy of the model.

[0086] The accuracy rate can be set to different range thresholds. For example, for different fault scenarios, if a higher accuracy rate is required for more complex situations, iterations can be performed, while for more regular faults, a lower fault accuracy rate can be applied.

[0087] Example 4 The present invention provides a practical application example of the method and device for automatically diagnosing fan equipment faults.

[0088] The project sensors have been installed and all on-site data has been uploaded to the platform, which is used to verify that some of the early warning and diagnosis data are real data from on-site equipment.

[0089] The data is organized into data sets to form a standard data format, and the results are generated by batch running of the program and compared with the actual data. During the test, the diagnostic model is tuned and compared to optimize the diagnostic model parameters.

[0090] The collected data were used for specific diagnosis, and the Z-score method was used to reduce noise on the collected data. The data before and after noise reduction were as follows: Figure 6 and Figure 7 shown.

[0091] The data after dimensionality reduction is as follows Figure 8 shown.

[0092] The classification model is generated by training the reduced-dimensional data, and the data collected on site is input into the classification model.

[0093] The diagnostic device generates the diagnostic results. After the results are manually confirmed, the user chooses whether to add them to the sample library and labels the data, which can continuously improve the accuracy of the model. After comparison, the following can be achieved: 99% continuous operation rate of equipment; 20% increase in personnel inspection efficiency; 60% fault diagnosis accuracy.

[0094] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A method for automatically diagnosing fan equipment faults, characterized in that: Include at least one of the following diagnostic steps: The collected fan equipment operation data is divided into low-frequency characteristics and high-frequency vibration data, and then cleaned, denoised, feature extracted and dimension reduced respectively; Construct a fault diagnosis model and classify fault modes to achieve automatic diagnosis and identification of fan equipment faults; After each diagnosis or test, data is saved to update the fault diagnosis model data set; The fan equipment operation data includes process data, vibration data, temperature data, environmental parameters and speed parameters; the process data includes the equipment's damper opening, fan inlet and outlet pressures, fan inlet and outlet temperatures, lubricating oil pressure and temperature data.

2. The automatic diagnosis method for fan equipment fault according to claim 1, characterized in that: Save the data after each diagnosis or test, add the samples to the fault diagnosis model sample library, select the model update method, and gradually iterate the accuracy of the fault diagnosis model.

3. The automatic diagnosis method for fan equipment fault according to claim 1, characterized in that: For low-frequency eigenvalue data, the Z-score method is used to identify outliers, and linear interpolation is used to fill the data.

4. The automatic diagnosis method for fan equipment fault according to claim 1, characterized in that: For high-frequency vibration data, low-pass filtering is first performed to remove high-frequency noise, and then adaptive filtering is used to further suppress dynamic noise.

5. The automatic diagnosis method for fan equipment fault according to claim 1, characterized in that: During feature extraction and dimensionality reduction, the features of the above-mentioned process data, vibration data, temperature data, environmental parameters, and operating parameters are extracted, and PCA is used for feature selection.

6. The automatic diagnosis method for fan equipment fault according to claim 1, characterized in that: During feature extraction and dimensionality reduction, the characteristic values ​​of each signal obtained include: the peak value and effective value of the acceleration time domain signal per second, and the frequency multiple amplitude, 2 times frequency amplitude, 3 times frequency amplitude, 4 times frequency amplitude, and 5 times frequency amplitude of the acceleration signal spectrum, the peak value and effective value of the speed time domain signal per second, the peak value and effective value of the acceleration envelope signal per second, and the amplitude of the envelope signal spectrum at the fault frequency, the energy of the acceleration signal spectrum per second in the low frequency band, medium frequency band, and high frequency band; the peak value and effective value of the current time domain signal per second, and the frequency multiple amplitude, 2 times frequency amplitude, 3 times frequency amplitude, 4 times frequency amplitude, and 5 times frequency amplitude of the current signal spectrum; the average value of the temperature signal per second.

7. The automatic diagnosis method for fan equipment fault according to claim 1, characterized in that: A fault diagnosis model is constructed based on the K nearest neighbor classification method.

8. A fan equipment fault automatic diagnosis device, characterized in that include: Data acquisition module, used to collect various operating data of fan equipment, including process data, vibration data, temperature data, environmental parameters and speed; The process data collected include: damper opening, fan inlet and outlet pressure, fan inlet and outlet temperature, lubricating oil pressure and temperature; The computing module is used to clean and reduce noise, extract features and reduce the dimension of the collected data, and classify fault modes using the feature data after dimension reduction to achieve automatic diagnosis and identification; The model training and updating module is used to build a fault diagnosis model based on the K-nearest neighbor classification method, and to confirm and save the data after each diagnosis or test and update it to the sample library. The accuracy of the fault diagnosis model is gradually iterated by manually triggering training, updating based on accuracy, or conducting regular training based on time periods.

9. The automatic fault diagnosis device for fan equipment according to claim 8, characterized in that: The data acquisition module includes a built-in data access module, a data acquisition module and an environmental parameter acquisition module, and sets a configuration module for selecting the initial parameters of the equipment; the data access module is used to obtain process data; the data acquisition module is used to collect vibration data, temperature data and rotation speed; the environmental parameter acquisition module is used to collect ambient humidity.

10. The automatic fault diagnosis device for fan equipment according to claim 8, characterized in that: The data acquisition module is configured to install a rotation speed sensor, a temperature sensor, and an acceleration sensor on each bearing seat of the fan when collecting vibration data.

Images