Bearing health state detection system for micro turbojet engine and application method

Abstract

The application discloses a bearing health state detection system for a micro small turbojet engine and an application method, and relates to the technical field of bearing health state detection systems, and aims to provide a bearing health state detection system for a micro small turbojet engine and an application method.

Description

Technical Field

[0001] This invention relates to the field of equipment health assessment. More specifically, this invention relates to a bearing health condition detection system and application method for miniature turbojet engines. Background Technology

[0002] In the bearing rotor system of a micro turbojet engine, the bearings used in the micro turbojet engine need to withstand high speeds of over 100,000 rpm and operate stably in high-temperature environments above 300°C. These bearings are very expensive, and replacing them requires disassembling the entire engine main shaft, resulting in high labor costs.

[0003] The internal space of a micro turbojet engine is typically very compact, and the operating environment is hot, making it difficult to effectively monitor the health of bearings using internal sensors or external testing equipment. Aside from recording bearing usage time, the conventional testing method involves manually turning the rotor several times quickly while the engine is not running, observing the number of rotations due to inertia and listening to the sound of the rotor's rotation, and relying on the experience of the inspector to judge the remaining life of the bearing. In this situation, due to cautious considerations for engine safety, the safe service hours for bearings are usually conservatively defined. Because the overall operating conditions vary greatly between individual engines, in many cases, the bearing's health is still relatively good, but the specified service hours have already been reached. At this point, the bearing needs to be replaced, resulting in significant waste. In some engines, high-temperature and high-speed operating conditions account for a large proportion, causing bearings to fail before reaching the specified service hours, leading to very serious consequences. Summary of the Invention

[0004] One object of the present invention is to solve at least the above-mentioned problems and / or defects, and to provide at least the advantages described below.

[0005] To achieve these objectives and other advantages of the present invention, a bearing health condition detection system for a miniature turbojet engine is provided, comprising:

[0006] A shim-type acoustic sensor is mounted on the spindle and located between the outer ring of the bearing and the preload spring;

[0007] The engine control unit communicates with the acoustic sensor via a digital-to-analog converter module;

[0008] The display unit communicates with the engine control unit.

[0009] The voiceprint sensor is configured to include:

[0010] A piezoelectric ceramic strain gauge with lead-out electrodes mounted on it;

[0011] Upper and lower housings are installed at the top and bottom of the piezoelectric ceramic strain gauge;

[0012] A gasket placed between the upper and lower housings to spatially support and define the upper and lower end faces of the piezoelectric ceramic strain gauge;

[0013] The piezoelectric ceramic strain gauge, upper shell, lower shell, and gasket are all configured in a concentric ring structure.

[0014] Preferably, the outer diameter of each annular gasket is smaller than the outer diameter of the piezoelectric ceramic strain gauge, and it is disposed on the inner side of the upper and lower housings;

[0015] The upper and lower ends of the piezoelectric ceramic strain gauge on the side away from the annular gasket are respectively fixed between the upper and lower shells by matching damping rings.

[0016] Preferably, it also includes a thin-film substrate that cooperates with the piezoelectric ceramic strain gauge, the substrate being configured to be made of stainless steel.

[0017] Preferably, the outer edge of the lower housing is provided with a protrusion extending upward to the upper housing;

[0018] The lead-out end of the electrode passes through the lead hole on the protrusion and connects to the input interface of the digital-to-analog converter module;

[0019] The lead hole is provided with a sealing resin layer.

[0020] Preferably, the method for fabricating the voiceprint sensor is configured to include:

[0021] S1. The ceramic strain gauge is bonded to the substrate with high-temperature resin, and damping rings are bonded to the upper and lower end faces of the outer side of the ceramic strain gauge respectively.

[0022] S2. Place a gasket on the upper and lower end faces of the inner ring of the substrate, and use resistance welding to weld the two gaskets to the three parts of the substrate into a whole.

[0023] S3. Lead the electrodes of the piezoelectric ceramic strain gauge out from the edge of the gasket housing;

[0024] S4. The contact surfaces of the inner rings of the upper and lower housings with the corresponding inner rings of the gaskets, as well as the contact surfaces of the outer ring edges of the upper and lower housings, are all welded by resistance welding.

[0025] S5. The electrode lead-out positions on the lower housing are sealed with high-temperature resin.

[0026] An application method for a bearing health status detection system for a miniature turbojet engine includes:

[0027] Step 1: When the engine rotor is running, the vibration amplitude on the bearing raceway is collected by an acoustic sensor;

[0028] Step two: The analog-to-digital conversion module amplifies the current signal collected by the acoustic sensor, converts it into a digital signal, and sends it to the engine control unit.

[0029] Step 3: The detection software on the engine control unit filters out waveforms not caused by bearing vibration, and continuously selects, crosses, mutates, and calculates fitness to determine whether the error meets the conditions. If not, it modifies each weight and threshold to further reduce the error.

[0030] Step four: The engine control unit displays the waveform and calculation results that meet the error requirements through the display unit, thus completing the bearing fault detection work;

[0031] The engine control unit determines the health status of the bearing based on whether the calculation results are within a given health value range, and displays the results through the display unit.

[0032] Preferably, the detection software uses a BP neural network for model classification and fault diagnosis. The BP neural network classifies the bearing fault type based on the signal characteristics of each frequency band in the vibration signal and trains on the bearing time-domain feature data to predict the bearing fault condition or detect the bearing health status.

[0033] The detection software uses wavelet packets to decompose the bearing vibration signal collected by the acoustic sensor into three layers of wavelet packets for the bearing signals in four states, so as to plot the energy spectrum of different fault signals for the eight frequency components obtained in different frequency bands under the same state.

[0034] Eight frequency components obtained from different frequency bands under the same condition are used as sample data for training a BP neural network to obtain the corresponding health assessment model.

[0035] Preferably, during the training of the BP neural network, the number of hidden layer neurons is set to 12, and the error of each layer is continuously calculated using gradient descent.

[0036] The training parameters include a learning rate of 0.01, a maximum number of iterations of 5000, and a target error of 0.0001.

[0037] In the genetic algorithm's running parameters, the population size is 60, the maximum number of generations is 50, the crossover probability is set to 0.70, the mutation probability is set to 0.01, and the generation gap is 0.95.

[0038] The present invention has at least the following beneficial effects: The present invention provides a bearing health condition visualization detection system, which detects the vibration of the bearing during operation based on the piezoelectric ceramic principle, performs analog-to-digital conversion through a digital-to-analog conversion module, further analyzes the vibration signal through software on the engine control unit, and intuitively displays the specific sound patterns generated when the bearing rotates through a display unit, so as to facilitate real-time detection of the bearing health condition during engine operation.

[0039] This invention provides an application method for a bearing health status visualization detection system. After collecting bearing vibration information, it uses a wavelet packet combined with an optimal hidden layer GA-BP neural network to accurately diagnose different fault or health states of the bearing, so that operators can determine whether replacement is necessary, and effectively control maintenance and equipment costs.

[0040] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the composition of the bearing health status detection system for miniature turbojet engines of the present invention.

[0042] Figure 2 This is a schematic diagram of the installation of the acoustic signature sensor of the present invention;

[0043] Figure 3 This is a schematic diagram of the structure of the pad-type acoustic sensor of the present invention;

[0044] Figure 4 This is the energy spectrum corresponding to each state of the bearing of the present invention;

[0045] Figure 5 This is a schematic diagram of the error training curve of the present invention. Detailed Implementation

[0046] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0047] It should be understood that terms such as “having,” “comprising,” and “including” as used herein do not imply the presence or addition of one or more other elements or combinations thereof.

[0048] It should be noted that in the description of this invention, the orientation or positional relationship indicated by the terms is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing this invention and simplifying the description. It does not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0049] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installed", "equipped", "sleeved / connected", "connected", etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0050] The bearing health status monitoring system for miniature turbojet engines includes: a shim-type acoustic sensor 1, a digital-to-analog converter module 2, an engine control unit 3 (with monitoring software installed), and a display screen 4. Figure 2 As shown, this system installs a sound pattern sensor in the bearing 5. Based on the piezoelectric ceramic principle, the sound pattern sensor, along with subsequent amplification, analog-to-digital conversion, and signal analysis hardware and software, intuitively displays the specific sound pattern generated when the bearing rotates, and performs real-time detection of the bearing's health status during engine operation.

[0051] like Figure 3 The pad-type acoustic sensor includes: an upper housing 10, a lower housing 11, a damping ring 12, a ceramic strain gauge 13, etc. The upper and lower housings are made of stainless steel by machining. The lower housing has thicker edges and greater axial rigidity to improve the load-bearing capacity of the pad, which is used to transmit the pressure of the preload spring to the outer ring of the bearing. The piezoelectric ceramic sheet is made into a ring shape to accommodate the main shaft 55 passing through it.

[0052] The electrode 14 of the piezoelectric ceramic strain gauge extends from the edge of the gasket housing, and the extension position is sealed with high-temperature resin. The ceramic strain gauge is bonded to the stainless steel sheet substrate 15 with high-temperature resin. A damping ring is then bonded to the ceramic strain gauge. The damping ring with appropriate weight and rigidity can increase the frequency of the ceramic gauge itself, filter out low-frequency vibrations, and also resist additional deformation caused by large overloads during transportation, assembly, drops, and aircraft maneuvers, thereby improving the robustness of the ceramic strain gauge.

[0053] The inner ring of the stainless steel substrate containing the ceramic strain gauge is sandwiched between two layers of annular rigid stainless steel support plates 16 (or gaskets). The three parts are welded together at welding point A17 using resistance welding, which can provide support rigidity for the inner ring of the gasket when used in different installation methods.

[0054] The contact surfaces of the inner rings of the upper and lower housings and the inner ring of the annular rigid support gasket are welded together by resistance welding; finally, the welding point B18 where the outer edges of the upper and lower housings are contacted is also welded together by resistance welding. This forms a closed, flat, box-like structure. The closed structure can prevent bearing lubricating oil and dust from entering the box; in addition, the closed shell can also isolate some of the aerodynamic noise generated during engine operation.

[0055] The working principle of the pad-type acoustic sensor (hereinafter referred to as "sensor") is as follows: The sensor is installed in the bearing cavity 50, the lower housing contacts the preload spring 51, and the upper housing contacts the outer ring 52 of the bearing. When the bearing rotates, the bearing balls roll on the outer ring raceway 53 of the bearing (the outer ring raceway is between the inner ring 54 and the outer ring of the bearing). Due to the limitations of bearing manufacturing level and cost considerations, the microstructure of the bearing raceway surface has a certain roughness or pits formed by the slight peeling off of the raceway surface after the bearing is used, which will cause the outer ring raceway of the bearing to vibrate during operation. The outer ring of the bearing transmits minute vibrations to the upper housing of the sensor. The upper housing, through a ring-shaped rigid support pad, transmits the vibrations to the ceramic strain gauge. Under mechanical stress, the ceramic strain gauge causes the relative displacement of the positive and negative charge centers inside the piezoelectric material, resulting in polarization. This leads to the appearance of bound charges of opposite signs on the surfaces at both ends of the material, thus generating the piezoelectric effect. The piezoelectric effect converts the collected minute vibrations into a current signal, which is then transmitted to downstream amplification equipment for further processing via the output wire. In this solution, the piezoelectric ceramic strain gauge is made into an ultra-thin acoustic sensor by utilizing its high temperature resistance, high detection accuracy, and extremely thin profile. It can be conveniently and cost-effectively installed in existing rotor systems. In practical applications, only a 3mm increase in thickness is needed to install it in the existing bearing cavity. Through a corresponding bearing health status visualization detection system, the health status of the bearing can be detected in real time, effectively improving the economy and safety of the engine.

[0056] Furthermore, the basic principle of the bearing health status visualization system based on vibration sound patterns is as follows: the fault diagnosis process of rolling bearings is essentially a pattern recognition process. When a rolling bearing fails, the fault type can be classified according to the characteristics of each frequency band of the vibration signal. The bearing's time-domain feature data is trained using a neural network to predict the bearing's fault condition. Backpropagation (BP) neural networks are primarily used for model classification and fault diagnosis.

[0057] The specific working principle is as follows: When the engine rotor is running, the current signal collected by the sensor is amplified and converted into a digital signal inside the analog-to-digital conversion module. The signal is then sent to the engine control unit to filter out waveforms caused by non-bearing vibrations, such as rotor imbalance, aerodynamic noise, and sensor background noise. Since the filtering algorithm is a conventional technique, it will not be described here. The genetic algorithm continuously selects, crosses over, mutates, and calculates fitness to determine whether it meets the conditions. The conditions here refer to the selected weights and thresholds. These weights and thresholds can minimize the number of iterations of the BP neural network. The selection, crossover, and mutation in the genetic algorithm are all used for signal preprocessing, with the aim of optimizing the weights and thresholds to minimize their errors. The specific optimization process for weights and thresholds is as follows: 1. First, determine the topology; 2. Obtain the hidden layer structure by combining the average variance and the number of iterations; 3. Encode the initial random weights to obtain the initial population; 4. Decode to obtain the weights and thresholds; 5. Train the network using training samples; 6. Test the network using test samples; 7. Calculate the fitness; 8. Select the population with high fitness for replication; 9. Perform crossover and mutation; 10. Obtain the new population (new weights and thresholds); 11. Decode the new population; 12. Obtain the optimal weights and thresholds.

[0058] Meanwhile, the weights and thresholds are continuously modified. Initially, these weights and thresholds are randomly assigned, and then optimized using a genetic algorithm to minimize errors. Finally, the calculation results are displayed on a screen, completing the bearing inspection process.

[0059] Diagnostic principle of bearing fault signals: When a new bearing is in operation, due to the smoothness of the bearing raceway, the vibration amplitude and frequency caused by the balls rolling on the smooth surface with specific microstructures are small. The aforementioned system is used to collect a continuous, periodically changing signal containing raceway characteristics. After a period of operation, tiny spalling pits inevitably appear on the outer raceway surface. When a ball rotates to this pit, a small vibration is generated. At this time, due to the change in raceway surface characteristics, the periodically changing signal containing raceway surface characteristics collected by the sensor also changes accordingly. As the spalling pits and other micro-defects increase, the signal will also change accordingly. The bearing vibration signal is collected using an acoustic fingerprint sensor. Wavelet packets are used to perform three-level wavelet packet decomposition on the bearing signals of four states, with the wavelet function selected as db10. The energy spectrum of different fault signals is plotted using the eight frequency components of the obtained signal, as shown below. Figure 4 As shown.

[0060] This patent uses the energy corresponding to eight different frequency bands as network input, and the training sample data for the BP neural network is shown in Table 1. The network is trained 10 times for each number of hidden layer units to obtain the corresponding average number of iterations and average mean square value, thereby selecting the relatively most reasonable number of hidden layers. The average network performance corresponding to different numbers of hidden layers is shown in Table 2.

[0061] Table 1

[0062]

[0063]

[0064] Table 2

[0065] Based on the table above, considering both the average number of iterations and the average mean square error, the average mean square error relative to the number of 8 hidden layer units is 9.70 × 10⁻⁶. -5 The highest accuracy is achieved at 9.74×10⁻⁶. -5 The difference is only 0.04×10 -5 Therefore, the difference in precision is too small to be considered, but the difference in the number of iterations is 8. Therefore, this patent sets the number of hidden layers to 12.

[0066] After initial training, the constructed BP neural network was divided into training and testing sets based on 100 sets of bearing data. The testing set comprised 70% of the data, and the training set comprised 30%. The testing set contained 70 data points, and the training set contained 30 data points. The number of neurons in the hidden layer was 12.

[0067] The error of each layer is continuously calculated using gradient descent. The learning rate is set to 0.01, the maximum number of iterations is 5000, and the target error is 0.0001. The genetic algorithm's parameters include a population size of 60, a maximum number of generations of 50, a crossover probability of 0.70, a mutation probability of 0.01, and a generation gap of 0.95. As the number of training iterations increases, simulations are performed using MATLAB to obtain the genetic algorithm's fitness curve, as shown below. Figure 5 As shown.

[0068] Four sets of data, including normal bearings and faulty bearings, were used as test samples to test the diagnostic results. The results of the fault diagnosis are shown in Table 3.

[0069]

[0070] As can be seen from Table 3, the GA-BP neural network proposed in this patent, which uses wavelet packets combined with the optimal hidden layer after collecting bearing vibration information, can accurately diagnose different fault states of the bearing.

[0071] The above solution is merely an illustration of a preferred example and is not limited thereto. When implementing this invention, appropriate substitutions and / or modifications can be made according to the user's needs.

[0072] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention. Applications, modifications, and variations of the invention will be readily apparent to those skilled in the art.

[0073] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for the present invention. Other modifications can be readily made by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and examples shown and described herein.

Claims

1. A bearing health status detection system for a miniature turbojet engine, characterized in that, include: A shim-type acoustic sensor is mounted on the spindle and located between the outer ring of the bearing and the preload spring; The engine control unit communicates with the acoustic sensor via a digital-to-analog converter module; The display unit communicates with the engine control unit. The voiceprint sensor is configured to include: A piezoelectric ceramic strain gauge with lead-out electrodes mounted on it; Upper and lower housings are installed at the top and bottom of the piezoelectric ceramic strain gauge; A gasket placed between the upper and lower housings to spatially support and define the upper and lower end faces of the piezoelectric ceramic strain gauge; The piezoelectric ceramic strain gauge, upper shell, lower shell, and gasket are all configured in a concentric ring structure. The method for manufacturing the voiceprint sensor is configured to include: S1. The ceramic strain gauge is bonded to the substrate with high-temperature resin, and damping rings are bonded to the upper and lower end faces of the outer side of the ceramic strain gauge respectively. S2. Place a gasket on the upper and lower end faces of the inner ring of the substrate, and use resistance welding to weld the two gaskets to the three parts of the substrate into a whole. S3. Lead the electrodes of the piezoelectric ceramic strain gauge out from the edge of the gasket housing; S4. The contact surfaces of the inner rings of the upper and lower housings with the corresponding inner rings of the gaskets, as well as the contact surfaces of the outer ring edges of the upper and lower housings, are all welded by resistance welding. S5. The electrode lead-out positions on the lower housing are sealed with high-temperature resin.

2. The bearing health status detection system for miniature turbojet engines as described in claim 1, characterized in that, The outer diameter of each annular gasket is smaller than that of the piezoelectric ceramic strain gauge, and they are arranged on the inner side of the upper and lower housings; The upper and lower ends of the piezoelectric ceramic strain gauge on the side away from the annular gasket are respectively fixed between the upper and lower shells by matching damping rings.

3. The bearing health status detection system for miniature turbojet engines as described in claim 2, characterized in that, It also includes a thin-film substrate that mates with a piezoelectric ceramic strain gauge, the substrate being configured to be made of stainless steel.

4. The bearing health status detection system for miniature turbojet engines as described in claim 3, characterized in that, The outer edge of the lower housing is provided with a protrusion extending upward to the upper housing; The lead-out end of the electrode passes through the lead hole on the protrusion and connects to the input interface of the digital-to-analog converter module; The lead hole is provided with a sealing resin layer.

5. An application method of the bearing health status detection system for miniature turbojet engines as described in any one of claims 1-4, characterized in that, include: Step 1: When the engine rotor is running, the vibration amplitude on the bearing raceway is collected by an acoustic sensor; Step two: The analog-to-digital conversion module amplifies the current signal collected by the acoustic sensor, converts it into a digital signal, and sends it to the engine control unit. Step 3: The detection software on the engine control unit filters out waveforms not caused by bearing vibration, and continuously selects, crosses, mutates, and calculates fitness to determine whether the error meets the conditions. If not, it modifies each weight and threshold to further reduce the error. Step four: The engine control unit displays the waveform and calculation results that meet the error requirements through the display unit, thus completing the bearing fault detection work; The engine control unit determines the health status of the bearing based on whether the calculation results are within a given health value range, and displays the results through the display unit.

6. The application method of the bearing health status detection system for miniature turbojet engines as described in claim 5, characterized in that, The detection software uses a BP neural network for model classification and fault diagnosis. The BP neural network classifies the bearing fault types based on the signal characteristics of each frequency band in the vibration signal and trains on the bearing time-domain feature data to predict the bearing fault condition or detect the bearing health status. The detection software uses wavelet packets to decompose the bearing vibration signal collected by the acoustic sensor into three layers of wavelet packets for the bearing signals in four states, so as to plot the energy spectrum of different fault signals for the eight frequency components obtained in different frequency bands under the same state. Eight frequency components obtained from different frequency bands under the same condition are used as sample data for training a BP neural network to obtain the corresponding health assessment model.

7. The application method of the bearing health status detection system for miniature turbojet engines as described in claim 6, characterized in that, During the training of the BP neural network, the number of neurons in the hidden layer is set to 12, and the error of each layer is continuously calculated using gradient descent. The training parameters include a learning rate of 0.01, a maximum number of iterations of 5000, and a target error of 0.0001. In the genetic algorithm's running parameters, the population size is 60, the maximum number of generations is 50, the crossover probability is set to 0.70, the mutation probability is set to 0.01, and the generation gap is 0.95.

Images