Liquid rocket engine fault simulation system, method, medium and equipment

Abstract

The invention discloses a liquid rocket engine fault simulation system, method, medium and equipment, the system comprises a fault simulation module, an environment interference simulation module and an image acquisition and fault identification module, the fault simulation module is used for simulating various fault states of a liquid rocket engine in a test run process; the environment infection simulation module is used for simulating various environment interferences occurring in the test run process of the liquid rocket engine; and the image acquisition and fault identification module is used for acquiring video image data of the liquid rocket engine in the fault simulation and environment interference simulation process so as to carry out real-time identification and diagnosis on the fault of the liquid rocket engine. According to the invention, the fault of the liquid rocket engine in the test run process can be accurately identified.

Description

Technical Field

[0001] The present invention belongs to the field of liquid rocket detection technology, and in particular relates to a liquid rocket engine fault simulation system, method, medium and equipment. Background Art

[0002] Liquid rocket engines are the core components of launch vehicles. With the introduction of engine health management systems and reusable launch vehicles, higher requirements are placed on the safety, reliability and economy of liquid rocket engines. Therefore, strict condition monitoring and fault diagnosis of key engine components and the operation of the entire machine are required.

[0003] The existing fault diagnosis technology of rocket engine test benches mostly adopts the redline algorithm, which collects and analyzes the slow-speed data and video image data after the engine test by high-speed cameras arranged in fixed positions to monitor and diagnose engine faults, and then makes corresponding optimizations and improvements. Although the existing video image acquisition system and redline mechanism monitoring system have played an important role in engine fault detection, the fixed-angle camera cannot fully and comprehensively monitor and collect the video image data of the engine during the test. Some parts of the engine cannot be captured by the camera due to overlapping devices, and there may be problems of missed faults. At the same time, the shutdown decision of the redline system is highly dependent on the setting of upper and lower thresholds, and the process of manually processing relevant video image data after the test is very time-consuming and laborious. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present disclosure is to provide a liquid rocket engine fault simulation system, method, medium and equipment, which can achieve comprehensive simulation and monitoring of faults and environmental interference that may occur in the test process of a liquid rocket engine.

[0005] To achieve the above objectives, the present disclosure provides the following technical solutions:

[0006] A liquid rocket engine fault simulation system, the system comprising: a fault simulation module, an environmental interference simulation module and an image acquisition and fault identification module, wherein the fault simulation module is used to simulate various fault states that may occur during a test run of a liquid rocket engine; the environmental interference simulation module is used to simulate various environmental interferences that may occur during a test run of a liquid rocket engine; the image acquisition and fault identification module is used to collect video image data of the liquid rocket engine during the fault simulation and environmental interference simulation processes, so as to identify and diagnose the faults of the liquid rocket engine in real time.

[0007] Optionally, the fault simulation module includes: an abnormal flame simulation unit, a leakage simulation unit and a sensor detachment simulation unit, wherein the abnormal flame simulation unit is used to simulate an abnormal flame fault occurring in a liquid rocket engine during a test run; the leakage simulation unit is used to simulate a high-pressure liquid leakage fault occurring in a liquid rocket engine during a test run; and the sensor detachment simulation unit is used to simulate an abnormal sensor detachment fault occurring in a liquid rocket engine during a test run.

[0008] Optionally, the sensor detachment simulation unit includes: an iron block, a de-energized electromagnet, an electromagnet fixing bracket and an engine model wall, wherein the engine model wall is fixed on a liquid rocket engine, the iron block is attached to the engine model wall, the electromagnet fixing bracket is fixed in the liquid rocket engine, and the de-energized electromagnet is fixed by the electromagnet fixing bracket.

[0009] Optionally, the image acquisition and fault identification module includes: an inspection robot, the inspection robot including a robot body, an AI vision robotic arm being arranged on the top of the robot body, wherein the inspection robot body is used to perform automatic inspections on liquid rocket engines during fault simulation and environmental interference simulation; the AI ​​vision robotic arm is used to dynamically collect image data of the liquid rocket engine during the automatic inspection process along with the inspection robot body, so as to identify the faults of the liquid rocket engine in real time.

[0010] Optionally, the AI ​​vision robotic arm includes: a bearing and a bearing frame, the bearing being fixed to the robotic arm assembly hole at the top of the robot body through the bearing frame, the robotic arm assembly hole being connected to the robot body through a support column to form a servo slot; a third high-voltage bus servo is fixed to the bottom end of the bearing, the third high-voltage bus servo is located in the servo slot, a second high-voltage bus servo is arranged on the bearing frame, the second high-voltage bus servo is connected to one end of a second aluminum alloy bracket, and the other end of the second aluminum alloy bracket is connected to the first high-voltage bus servo; the first high-voltage bus servo is also movably connected to the first aluminum alloy bracket, a camera mounting bracket is arranged on the first aluminum alloy bracket, and a high-definition camera is fixedly connected to the camera mounting bracket.

[0011] Optionally, the inspection robot is also provided with an emergency stop switch.

[0012] Optionally, the environmental interference simulation module includes: a high-brightness lighting lamp, a water mist machine and a vibration motor, wherein the high-brightness lighting lamp is arranged in the nozzle of the liquid rocket engine to simulate the strong light interference when the nozzle is working; the water mist machine is arranged on the outside of the liquid rocket engine to simulate the water mist environment when the liquid rocket engine is working; the vibration motor is arranged on the outside of the liquid rocket engine to simulate the vibration of the liquid rocket engine.

[0013] The present disclosure also provides a liquid rocket engine fault simulation method, the method comprising: performing fault simulation and environmental interference simulation on the liquid rocket engine, and collecting image data of the liquid rocket engine during the simulation; preprocessing the image data; analyzing the preprocessed image data to identify the fault characteristics of the liquid rocket engine and record the fault information, and generating a fault detection report.

[0014] The present disclosure also provides a storage medium, which includes instructions. When the storage medium is executed on a computer, the computer is enabled to execute the above method.

[0015] The present disclosure also provides an electronic device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the aforementioned method when executing the program.

[0016] Compared with the prior art, the beneficial effects brought by the present disclosure are as follows: the present disclosure realizes comprehensive simulation and real-time monitoring of various fault conditions and environmental interferences that may occur in the test process of liquid rocket engines by integrating fault simulation modules, environmental interference simulation modules and image acquisition and fault identification modules. The fault simulation module can accurately simulate faults such as abnormal flames, high-pressure liquid leakage and sensor detachment. The environmental interference simulation module can simulate complex conditions such as strong light interference, water mist environment and engine vibration. The image acquisition and fault identification module realizes all-round image acquisition and real-time fault diagnosis of various parts of the engine through the collaborative work of the inspection robot and the AI ​​visual robotic arm. This comprehensive simulation and monitoring system not only improves the efficiency and accuracy of fault detection, but also enhances the adaptability and reliability of the system, providing a strong guarantee for the safety and reliability of liquid rocket engines. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the overall structure of a liquid rocket engine fault simulation system provided by an embodiment of the present disclosure;

[0018] Figure 2 is a structural schematic diagram of an inspection robot provided by another embodiment of the present disclosure;

[0019] Figure 3 is a schematic structural diagram of a robot body provided by another embodiment of the present disclosure;

[0020] Figure 4 is a structural schematic diagram of an AI visual robotic arm provided by another embodiment of the present disclosure;

[0021] Figure 5 is a schematic diagram of a control system of an image acquisition and fault identification module provided by another embodiment of the present disclosure;

[0022] Figure 6 is a front view of a sensor shedding simulation module provided by another embodiment of the present disclosure;

[0023] Figure 7 is a rear view of a sensor shedding simulation module provided by another embodiment of the present disclosure;

[0024] Figure 8 is a top view of a sensor shedding simulation module provided by another embodiment of the present disclosure;

[0025] Fig. 9 is a structural schematic diagram of a leakage simulation unit provided by another embodiment of the present disclosure;

[0026] Fig.10 is a schematic diagram of the structure of a convolutional neural network provided by another embodiment of the present disclosure;

[0027] The following are the descriptions of the reference numerals:

[0028] 100. Liquid rocket engine; 200. Abnormal flame simulation unit; 300. Leakage simulation unit; 400. Sensor detachment simulation unit; 500. High-brightness lighting; 600. Water mist machine; 700. Vibration motor; 800. Image acquisition and fault identification module; 1. Industrial camera; 2. Pan / tilt; 3. AI visual robotic arm; 4. Main control board; 5. Inspection robot; 6. Lithium battery; 7. Luminous ultrasonic sensor; 8. Emergency stop switch; 9. Mechanical foot; 10. Box; 11. High-definition camera; 12. Camera mounting bracket; 13. First high-voltage bus servo; 14. First aluminum alloy bracket; 15. Second aluminum alloy bracket; 16. Second high-voltage bus servo; 17. Bearing frame; 18. Bearing; 19. Third high-voltage bus servo; 20. Iron block; 21. Power-off type electromagnet; 22. Electromagnet fixing bracket; 23. Engine model wall. DETAILED DESCRIPTION

[0029] The following will refer to the attached Figures 1 to 10Specific embodiments of the present disclosure are described in detail. Although specific embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0030] It should be noted that certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that technicians may use different nouns to refer to the same component. This specification and claims do not use the difference in nouns as a way to distinguish components, but use the difference in the functions of the components as the criterion for distinction. As mentioned throughout the specification and claims, "including" or "comprising" is an open term, so it should be interpreted as "including but not limited to". The subsequent description of the specification is a preferred embodiment of the present disclosure, but the description is based on the general principles of the specification and is not used to limit the scope of the present disclosure. The scope of protection of the present disclosure shall be determined by the attached claims.

[0031] To facilitate understanding of the embodiments of the present disclosure, further explanation will be given below using specific embodiments as examples in conjunction with the accompanying drawings, and each of the accompanying drawings does not constitute a limitation on the embodiments of the present disclosure.

[0032] Figure 1 FIG. 1 is a schematic diagram of a liquid rocket engine fault simulation system provided by an embodiment of the present disclosure. Figure 1 As shown, the system includes: a fault simulation module, an environmental interference simulation module and an image acquisition and fault identification module, wherein the fault simulation module is used to simulate various fault states that occur during the test of the liquid rocket engine; the environmental interference simulation module is used to simulate various environmental interferences that occur during the test of the liquid rocket engine; the image acquisition and fault identification module is used to collect video image data of the liquid rocket engine during the fault simulation and environmental interference simulation processes, so as to identify and diagnose the faults of the liquid rocket engine in real time.

[0033] The present invention realizes comprehensive simulation of liquid rocket engines by integrating fault simulation modules, environmental interference simulation modules and image acquisition and fault identification modules. This comprehensive simulation and monitoring system not only improves the efficiency and accuracy of fault detection, but also enhances the adaptability and reliability of the system, providing a strong guarantee for the safety and reliability of liquid rocket engines.

[0034] In another exemplary embodiment, the fault simulation module includes an abnormal flame simulation unit 200, a leakage simulation unit 300 and a sensor detachment simulation unit 400, wherein the abnormal flame simulation unit 200 is used to simulate an abnormal flame fault occurring during the test run of the engine; the leakage simulation unit 300 is used to simulate a high-pressure liquid leakage fault occurring during the test run of the engine; and the sensor detachment simulation unit 400 is used to simulate an abnormal sensor detachment fault occurring during the test run of the engine.

[0035] In this embodiment, the abnormal flame simulation unit 200 adopts a small flame spray device, which is placed in the combustion elbow and combustion chamber of the engine model. The small flame spray device is connected to the STM32 controller, and a control signal can be sent to the STM32 controller through a wireless device (such as a wireless handle or a mobile phone). After receiving the control signal, the STM32 controller parses the signal through its built-in signal receiver or Bluetooth module. The STM32 controller outputs a corresponding PWM (pulse width modulation) signal through a selected pin according to the parsed control signal (the pulse period of the PWM signal is 20ms, and the steering angle of the steering gear is controlled by changing the duty cycle). After receiving the PWM signal, the steering gear in the small flame spray device can be directional rotated according to the duty cycle of the signal, and the rotation of the steering gear is transmitted to the control part of the flamethrower through the gear rack transmission mechanism, thereby realizing the switch of the flamethrower and the adjustment of the flame size.

[0036] like Fig. 9 As shown, the leakage simulation unit 300 includes a water tank, a water pump, a relay and a water spraying device, wherein the water pump is connected to the water tank, the control end of the relay is connected to the GPIO pin of the main control board, and the load end of the relay is connected to the power line of the water pump. The water spraying device is composed of a PU air pipe and a variable diameter nozzle, one end of the PU air pipe is connected to the water outlet of the water pump, and the other end is connected to the variable diameter nozzle, and the variable diameter nozzle is installed in three places of the gas generator, fuel booster pump and oxidant booster pump of the liquid rocket engine.

[0037] In this embodiment, a control signal can be sent to the main control board through a wireless device (such as a wireless handle or a mobile phone). After receiving the control signal, the main control board parses the signal through its built-in signal receiver or Bluetooth module. The main control board outputs the control signal to the relay through the selected GPIO pin according to the parsed control signal. After receiving the control signal, the relay controls the circuit switch of the water pump. When the relay is closed, the water pump starts; when the relay is disconnected, the water pump stops. After the water pump starts, the high-pressure liquid is transported to the variable-diameter nozzle through the PU air pipe. The variable-diameter nozzle adjusts the injection speed and direction of the liquid according to the designed flow adjustment structure, so that liquid leakage under different pressures can be simulated.

[0038] like Figures 6 to 8 As shown, the sensor abnormal detachment simulation unit 400 includes an iron block 20, a de-energized electromagnet 21, an electromagnet fixing bracket 22 and an engine model wall 23, wherein the engine model wall 23 is fixed to the liquid rocket engine by screws to provide a flat surface, the iron block 20 is attached to the engine model wall 23, the de-energized electromagnet 21 is fixed to the electromagnet fixing bracket 22 by screws, and the electromagnet fixing bracket 22 is fixed to the inner cavity of the fuel booster pump and the oxidizer booster pump of the liquid rocket engine by strong glue, in addition, the live wire of the de-energized electromagnet 21 is connected to the output pin of the STM32 controller through a switch, and the power supply and ground wire of the de-energized electromagnet 21 are connected to the 24V power supply and ground wire respectively. A control signal can be sent to the STM32 controller through a wireless device (such as a wireless handle or a mobile phone). After receiving the control signal, the STM32 controller parses the signal through its built-in signal receiver or Bluetooth module. The STM32 controller outputs the control signal to the switch of the power-off type electromagnet 21 through the selected GPIO pin according to the parsed control signal. The live wire of the power-off type electromagnet 21 is connected to the output pin of the STM32 controller through the switch. When the STM32 controller outputs a high level, the power-off type electromagnet 21 is energized and generates a magnetic field to adsorb the iron block 20, so that the iron block 20 is firmly attached to the engine model wall 23; when the STM32 controller outputs a low level, the power-off type electromagnet 21 is powered off, and the magnetic field disappears at this time, and the iron block 20 loses its adsorption force and falls off from the engine model wall 23, thereby simulating an abnormal sensor falling-off failure.

[0039] With the sensor abnormal detachment simulation unit 400 having the above structure, the operator can remotely control the on and off of the electromagnet through a wireless device, and can trigger the sensor detachment fault without contacting the engine, thereby improving the safety and flexibility of the experiment. In addition, since the triggering of the sensor detachment fault is completely dependent on the on and off of the electromagnet, this design makes it very easy to reproduce the fault, which is conducive to repeated experimental comparison and fault analysis, thereby improving the accuracy and reliability of the liquid rocket engine fault simulation experiment.

[0040] In another embodiment, the image acquisition and fault identification module includes: an inspection robot 5, wherein the inspection robot 5 includes a robot body, an AI visual mechanical arm 3 is arranged at the top of the robot body, and the AI ​​visual mechanical arm 3 includes a bearing 18 and a bearing frame 17, the bearing 18 is fixed to the mechanical arm assembly hole at the top of the robot body through the bearing frame 17, and the mechanical arm assembly hole is connected to the robot body through a support column (such as Figure 3As shown, there are 4 support columns between the assembly hole and the robot body) to connect and form a steering gear slot. A third high-voltage bus steering gear 19 is arranged at the bottom end of the bearing 18, and the third high-voltage bus steering gear 19 is located in the steering gear slot and is responsible for the overall horizontal rotation of the robot arm; a second high-voltage bus steering gear 16 is arranged on the bearing frame 17, and the second high-voltage bus steering gear 16 is connected to one end of the second aluminum alloy bracket 15, and the second high-voltage bus steering gear 16 is responsible for the rotation of the middle section of the robot arm; the other end of the second aluminum alloy bracket 15 is connected to the first high-voltage bus steering gear 13, and the second aluminum alloy bracket 15 is used to transmit the rotation action, and the first high-voltage bus steering gear 13 is responsible for the vertical rotation of the robot arm; the first high-voltage bus steering gear 13 is also movably connected to the first aluminum alloy bracket 14, and the first aluminum alloy bracket 14 is provided with a camera mounting bracket 12, and the camera mounting bracket 12 is fixedly connected with a high-definition camera 11, and the high-definition camera 11 is connected to the main control board through the CSI interface, and the high-definition camera 11 is used to collect image data of the liquid rocket engine during the test run to perform visual line finding and fault detection. The main control board generates a corresponding PWM signal according to the received operation signal or the result of the visual algorithm. The PWM signal is output to the first high-voltage bus servo 13, the second high-voltage bus servo 16, and the third high-voltage bus servo 19 through the selected GPIO pins to control the AI ​​visual robot arm 3 to rotate horizontally, vertically, and in the middle section, respectively, so as to flexibly adjust the angle of the high-definition camera 11 to ensure comprehensive monitoring of all parts of the liquid rocket engine. The high-definition camera 11 processes the collected image data through a visual module (such as OpenCV) to identify long red lines (the patrol line color can be set). At the same time, a PID controller is used for path tracking to ensure that the inspection robot 5 can move accurately along the line. At the same time, the high-definition camera 11 can also analyze the collected image data in real time through the built-in target detection algorithm to identify fault features such as abnormal flames, leakage points, and falling sensors in the image. Once any fault feature is detected, the Raspberry Pi 4B controller will control the inspection robot 5 to stop moving and stay at the current position for a period of time (such as 10 seconds) to obtain detailed fault image data, and record fault information and generate a test report at the same time. The AI ​​vision robot arm 3 works in conjunction with the luminous ultrasonic sensor 7 to ensure that the inspection robot can flexibly adjust the shooting angle during the inspection process while avoiding collisions.

[0041] In the image acquisition and fault identification module, the third high-voltage bus servo 19 is responsible for the overall horizontal rotation of the AI ​​visual mechanical arm 3, so that the AI ​​visual mechanical arm 3 can move freely on the horizontal plane, thereby expanding the coverage of the mechanical arm. The second high-voltage bus servo 16 is responsible for the rotation of the middle section of the AI ​​visual mechanical arm 3, so that the AI ​​visual mechanical arm 3 can be bent and adjusted in the middle part, thereby increasing the flexibility of the AI ​​visual mechanical arm 3. The first high-voltage bus servo is responsible for the vertical rotation of the AI ​​visual mechanical arm 3, so that the AI ​​visual mechanical arm 3 can move freely in the vertical direction, ensuring that the camera can shoot from different angles. In summary, the three high-voltage bus servos work together to enable the AI ​​visual mechanical arm 3 to achieve complex motion trajectories, so that the camera can freely and flexibly aim at the target area, and can be quickly in place to shorten the adjustment time and inspection time. In addition, this multi-degree-of-freedom design can also make the AI ​​visual mechanical arm 3 more stable during movement to reduce monitoring errors caused by vibration or impact.

[0042] Furthermore, it should be noted that the high-definition camera 11 performs real-time analysis of the image through a built-in target detection algorithm, which specifically includes the following steps:

[0043] Step 1: Preprocess the image;

[0044] Image preprocessing includes denoising and contrast enhancement, where Gaussian filters can be used Denoising, using histogram equalization To enhance contrast, the preprocessing can be done with the following composite function: To express:

[0045]

[0046] Step 2: Feature extraction;

[0047] Specifically, convolutional neural networks can be used to extract image features and output feature maps , among which, Fig.10 As shown, the convolutional neural network includes:

[0048] The multi-scale feature extraction module includes a multi-resolution input layer, a multi-scale convolution layer and a feature pyramid network (FPN), wherein the multi-resolution input layer is used to scale the pre-processed image to different scales to generate images with multiple resolutions; for example, the original resolution is H×W×5, and different scaling ratios are set, such as 0.5, 0.75, and 1.0, then the scaled resolutions are 0.5 H×0.5W×5, 0.75H×0.75W×5, and H×W×5. The multi-scale convolution layer is used to perform independent convolution processing on images with different resolutions, and the multi-scale convolution layer includes a first convolution layer (convolution kernel size 3×3, stride 1, padding 1, and output channel number 64), a second convolution layer (convolution kernel size 5×5, stride 1, padding 2, and output channel number 64), and a third convolution layer (convolution kernel size 7×7, stride 1, padding 3, and output channel number 64). The feature pyramid network (FPN) fuses feature maps of different scales through upsampling and downsampling operations, wherein the upsampling layer uses bilinear interpolation to upsample the low-resolution feature map to high resolution; the downsampling layer uses maximum pooling to downsample the high-resolution feature map to low resolution; the fusion layer generates a multi-scale feature map by element-by-element addition.

[0049] The multimodal feature fusion module includes an attention calculation layer and a weight adjustment layer, wherein the attention calculation layer uses a dot product attention mechanism to calculate the similarity between different modal features; and the weight adjustment layer is used to adjust the weight of each modal feature according to the similarity.

[0050] The adaptive feature enhancement module includes an adaptive normalization layer and a conditional random field. The adaptive normalization layer is used to normalize the fused features output by the multimodal feature fusion module; the conditional random field is used to enhance the spatial consistency of features by modeling the dependency relationship between pixels.

[0051] The lightweight network structure includes a deep convolution layer (convolution kernel size 3×3, stride 1, padding 1, and 64 output channels), a point convolution layer (convolution kernel size 1×1, stride 1, padding 0, and 64 output channels), and a ReLU activation function. In addition, a residual connection is added to the network, that is, the output of the lightweight network structure is added to the original input feature map.

[0052] After the above neural network processing, the final feature map F can be obtained:

[0053]

[0054] in, Represents a neural network.

[0055] In the convolutional neural network shown above, in the multi-scale feature extraction module, the multi-resolution input layer generates images of multiple resolutions by scaling the input image at different scales, which helps to capture local details and global information in the image. This multi-scale input enables the network to better understand features at different scales, thereby improving the recognition ability and robustness of the neural network for images. The multi-scale convolution layer uses convolution kernels of different sizes (3×3, 5×5, 7×7) to perform independent convolution processing on images of different resolutions, and can extract multi-scale features. Small convolution kernels are suitable for capturing fine-grained features, while large convolution kernels are suitable for capturing coarse-grained features. This multi-scale convolution helps to improve the model's sensitivity to features of different scales.

[0056] The feature pyramid network (FPN) fuses feature maps of different scales through upsampling and downsampling to generate multi-scale feature maps. The upsampling layer uses bilinear interpolation to upsample the low-resolution feature map to high resolution, and the downsampling layer uses maximum pooling to downsample the high-resolution feature map to low resolution. This fusion method can make full use of feature information at different levels and enhance the richness and diversity of features.

[0057] In the multimodal feature fusion module, the attention calculation layer uses the dot product attention mechanism to calculate the similarity between different modal features, and can dynamically adjust the importance weight of each modal feature. This mechanism enables the model to automatically focus on the most relevant modal information according to actual needs, improving the effectiveness and accuracy of feature fusion. The weight adjustment layer adjusts the weight of each modal feature according to the similarity, so that the fused features highlight important modal information while suppressing noise and irrelevant information. This adaptive weight adjustment helps to improve the generalization ability and robustness of the model.

[0058] In the adaptive feature enhancement module, the adaptive normalization layer normalizes the fused features output by the multimodal feature fusion module, which can make the feature distribution more stable and reduce the differences between different features. This normalization helps to improve the training speed and convergence of the model. The conditional random field enhances the spatial consistency of features by modeling the dependencies between pixels. CRF can capture the long-distance dependencies between pixels, so that adjacent pixels in the feature map have stronger correlation, thereby improving the coherence and integrity of the features.

[0059] In the lightweight network structure, using deep separable convolution instead of standard convolution can significantly reduce the number of model parameters and computational complexity, and improve the computational efficiency of the model. The deep convolution layer focuses on feature extraction of each channel, while the point convolution layer is responsible for cross-channel information fusion. The point convolution layer performs cross-channel information fusion through a 1×1 convolution kernel, which can reduce the number of parameters while maintaining the richness of features. The point convolution layer helps to improve the expressive power of the model. In addition, adding the output of the lightweight network structure to the original input feature map through a residual connection can alleviate the gradient vanishing problem and improve the training stability of the model. The residual connection enables the model to learn residual mapping more easily and improve the performance of the model.

[0060] Step 3: Fault diagnosis;

[0061] Specific classifiers can be used For feature maps Classify and output the failure probability , expressed as:

[0062]

[0063] Among them, the classifier It can be a deep learning model such as YOLOv5 or other object detection models.

[0064] Step 4: Fault record;

[0065] According to the probability of failure Determine whether there is a fault and record the fault information. This paper designs a fault recording function and sets the fault probability threshold to , if the probability of failure of a certain category exceeds , it is considered that a fault is detected. Specifically, the fault recording function It is expressed as follows:

[0066]

[0067] If a fault is detected, the fault information is recorded :

[0068]

[0069] Combining the above steps, we can get a comprehensive fault detection function:

[0070]

[0071] Furthermore, the present disclosure provides an exemplary description of fault detection based on the above model:

[0072] Based on the inspection robot, an image of a liquid rocket engine with a dimension of 1024×768 (width×height) is collected. After feature extraction, a feature map with a dimension of 256×192×512 (width×height×number of channels) is obtained. ; For the feature map Classification, including the following categories and corresponding probabilities:

[0073] Abnormal flame: 0.85

[0074] Leakage point: 0.30

[0075] Sensor fall off: 0.92

[0076] Assuming a failure probability threshold =0.7, further use the fault recording function Determine whether there is a fault and record the fault information. In the above example, the probability of abnormal flame is 0.85, and the probability of sensor detachment is 0.92, both exceeding the threshold of 0.7. Therefore, a fault is detected, and the fault information is recorded as follows:

[0077]

[0078] Based on the above examples, it can be seen that the use of the above model can accurately detect faults in liquid rocket engines under various fault simulations and environmental interference simulation environments, thereby ensuring that potential faults are discovered and recorded in a timely manner during the liquid rocket test process, thereby improving the safety and reliability of the engine.

[0079] In another exemplary embodiment, the inspection robot 5 further includes a box 10, which is fixed to the robot body by bolts and is used to accommodate the main control board 4, the lithium battery 6 and the luminous ultrasonic sensor 7, wherein the main control board adopts the Raspberry Pi 4B (4G) controller, which is used to receive the operation signal in the manual mode or process and judge the signal and image information transmitted by the sensor and the camera to control the movement of the inspection robot 5. The sensor adopts the luminous ultrasonic sensor 7, which is installed at the lower front end of the robot body, and the Trig and Echo pins of the luminous ultrasonic sensor 7 are respectively connected to different GPIO pins of the Raspberry Pi controller. The luminous ultrasonic sensor 7 can be used to detect obstacles in front in real time. The main control board 4 uses the obstacle avoidance algorithm to adjust the movement direction of the inspection robot according to the distance information returned by the luminous ultrasonic sensor 7, thereby realizing automatic obstacle avoidance.

[0080] Six mechanical feet 9 are arranged at the bottom of the robot body, each of which is fixed to the robot body by bolts, and each of the mechanical feet 9 is provided with three high-voltage bus servos, and the signal pins of the three high-voltage bus servos are connected to the GPIO pins of the main control board to respectively control the three degrees of freedom (hip joint, knee joint and ankle joint) of the mechanical foot.

[0081] The inspection robot 5 is also provided with a pan-tilt 2, on which an industrial camera 1 is provided. The pan-tilt 2 has two axial rotation capabilities, namely, horizontal (azimuth) and vertical (pitch) rotation, which enables the equipment (such as industrial cameras) installed thereon to flexibly adjust the shooting angle to cover a wider field of view. The pan-tilt can receive instructions from the main control board 4 through communication with the main control board, and accurately adjust to a preset position to ensure that it can accurately align with the target area when needed.

[0082] The inspection robot 5 is also provided with an emergency stop switch 8, which is fixed to the internal bracket above the front end of the inspection robot 5 by screws. The circuit module at the tail of the emergency stop switch 8 is connected to the power bus. When the operator presses the emergency stop switch 8, the internal contacts are disconnected, cutting off the current of the power bus, thereby immediately stopping all movements of the inspection robot 5. In addition, the emergency stop switch 8 usually needs to be reset manually, that is, after the emergency situation is resolved, the operator needs to manually reset the emergency stop switch 8 to resume the normal operation of the inspection robot 5. The emergency stop switch 8 can not only quickly cut off the power supply to ensure the safety of the operator and equipment, but also provide effective emergency measures when the inspection robot 5 fails or loses control. The design and installation position of the emergency stop switch 8 ensure that the operator can quickly reach it, improving the safety and reliability of the inspection task.

[0083] In another embodiment, the environmental interference simulation module includes a high-brightness lighting lamp 500, a water mist machine 600 and a vibration motor 700, wherein the high-brightness lighting lamp 500 is arranged in the nozzle of the liquid rocket engine to simulate the strong light interference received when the nozzle is working; the water mist machine 600 is arranged on the outside of the liquid rocket engine to simulate the water mist environment when the liquid rocket engine is working; the vibration motor 700 is arranged on the outside of the liquid rocket engine to simulate the vibration of the liquid rocket engine.

[0084] In this embodiment, the high-brightness lighting lamp 500, the water mist machine 600 and the vibration motor 700 are all connected to the same STM32 controller. The operator can send a control signal to the STM32 controller through a wireless device (such as a wireless handle or a mobile phone). After receiving the control signal, the STM32 controller parses the signal through its built-in signal receiver or Bluetooth module. The STM32 controller outputs the control signal to the switch of the high-brightness lighting lamp 500 through the selected GPIO pin according to the parsed control signal. After receiving the control signal, the high-brightness lighting lamp 500 is turned on or off to simulate the strong light interference when the nozzle is working. The STM32 controller outputs the control signal to the switch of the water mist machine 600 through the selected GPIO pin according to the parsed control signal. After receiving the control signal, the water mist machine 600 is turned on or off to simulate the water mist environment during operation. The STM32 controller outputs the control signal to the switch of the vibration motor 700 through the selected GPIO pin according to the parsed control signal. After receiving the control signal, the vibration motor 700 is turned on or off to simulate the engine vibration during operation.

[0085] The environmental interference simulation module realizes remote simulation of strong light interference, water mist environment and engine vibration during the test of liquid rocket engine by integrating components such as high-brightness lighting lamp 500, water mist machine 600 and vibration motor 700. These devices can not only accurately control the opening and closing of environmental interference, but also work stably in complex test environments, providing effective support for fault diagnosis and performance testing of liquid rocket engines.

[0086] In another embodiment, the present disclosure also provides a storage medium, which includes instructions, and when the instructions are executed on a computer, the computer executes the method as described above.

[0087] In another embodiment, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the aforementioned method when executing the program.

[0088] The above are only preferred embodiments of the present disclosure, and are not intended to limit the patent scope of the present disclosure. Any equivalent structure or equivalent process transformation made using the contents of the present disclosure and the drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present disclosure.

Claims

1. A liquid rocket engine fault simulation system, characterized in that: The system comprises: Fault simulation module, environmental interference simulation module and image acquisition and fault identification module, among which, The fault simulation module is used to simulate various fault conditions that may occur during the test run of a liquid rocket engine; The environmental infection simulation module is used to simulate various environmental disturbances that occur during the test run of a liquid rocket engine; The image acquisition and fault identification module is used to collect video image data of the liquid rocket engine during fault simulation and environmental interference simulation, so as to identify and diagnose the faults of the liquid rocket engine in real time.

2. The system according to claim 1, characterized in that Preferably, the fault simulation module includes: Abnormal flame simulation unit, leakage simulation unit and sensor shedding simulation unit, in, The abnormal flame simulation unit is used to simulate abnormal flame failures that occur during the test run of a liquid rocket engine; The leakage simulation unit is used to simulate the high-pressure liquid leakage failure that occurs during the test run of the liquid rocket engine; The sensor shedding simulation unit is used to simulate abnormal sensor shedding failure that occurs during the test run of a liquid rocket engine.

3. The system according to claim 2, characterized in that The sensor shedding simulation unit comprises: Iron block, de-energized electromagnet, electromagnet fixing bracket and engine model wall, in, The engine model wall is fixed on the liquid rocket engine, the iron block is attached to the engine model wall, the electromagnet fixing bracket is fixed in the liquid rocket engine, and the de-energized electromagnet is fixed by the electromagnet fixing bracket.

4. The system according to claim 1, characterized in that The image acquisition and fault identification module includes: The inspection robot comprises a robot body, and an AI visual mechanical arm is arranged on the top of the robot body, wherein: The inspection robot body is used to automatically inspect the liquid rocket engine during fault simulation and environmental interference simulation; The AI ​​vision robotic arm is used to dynamically collect image data of the liquid rocket engine during the automatic inspection process along with the inspection robot body, so as to identify faults of the liquid rocket engine in real time.

5. The system according to claim 4, characterized in that The AI ​​visual robotic arm comprises: A bearing and a bearing frame, wherein the bearing is fixed to a mechanical arm assembly hole at the top of the robot body through the bearing frame, and the mechanical arm assembly hole is connected to the robot body through a support column to form a steering gear slot; A third high-voltage bus servo is fixed to the bottom end of the bearing, and the third high-voltage bus servo is located in the servo slot. A second high-voltage bus servo is arranged on the bearing frame, and the second high-voltage bus servo is connected to one end of the second aluminum alloy bracket, and the other end of the second aluminum alloy bracket is connected to the first high-voltage bus servo; the first high-voltage bus servo is also movably connected to the first aluminum alloy bracket, and a camera mounting bracket is arranged on the first aluminum alloy bracket, and a high-definition camera is fixedly connected to the camera mounting bracket.

6. The system according to claim 1, characterized in that The inspection robot is also provided with an emergency stop switch.

7. The system according to claim 1, characterized in that The environmental interference simulation module includes: High brightness lighting, water mist machine and vibration motor, among which, The high-brightness lighting lamp is arranged in the nozzle of the liquid rocket engine to simulate the strong light interference when the nozzle is working; The water mist machine is arranged outside the liquid rocket engine to simulate the water mist environment when the liquid rocket engine is working; The vibration motor is arranged on the outside of the liquid rocket engine and is used to simulate the vibration of the liquid rocket engine.

8. A liquid rocket engine fault simulation method, characterized in that: The method comprises: Conduct fault simulation and environmental interference simulation on liquid rocket engines, and collect image data of liquid rocket engines during the simulation process; Preprocessing the image data; The pre-processed image data is analyzed to identify the fault characteristics of the liquid rocket engine and record the fault information, and a fault detection report is generated.

9. A storage medium, characterized in that: The method comprises instructions which, when executed on a computer, cause the computer to perform the method of claim 8 .

10. An electronic device, characterized in that: The electronic device comprises: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to claim 8 is implemented.

Images