A leak-proof fuel tank and intelligent detection method thereof

Abstract

The present invention relates to a fuel tank anti-leakage technology, and in particular to an anti-leakage fuel tank and an intelligent detection method thereof, comprising a base, the outer side of the base is fixedly connected with a support rod, the top of the support rod is fixedly connected with a fuel tank body, the top of the fuel tank body is fixedly connected with an oil inlet pipe, the bottom of the fuel tank body is fixedly connected with a guide groove, a lifting groove is opened in the inner wall of the base, the lifting groove is slidably connected with a slide rod, the outer side of the slide rod is fixedly connected with a storage box, a sponge is installed inside the storage box, the bottom of the slide rod is fixedly connected with a No. 1 reset spring, and the outer side of the storage box is fixedly connected with a moving rod. The present invention has a simple structure and is easy to use. When a fuel tank leaks, it can be detected and processed in time and an alarm reminder can be issued, and the leaked fuel can be recycled.

Description

Technical Field

[0001] The invention relates to an anti-leakage fuel tank and an intelligent detection method thereof, belonging to the technical field of anti-leakage of fuel tanks. Background Art

[0002] A fuel tank is a container for oil. It refers specifically to a device for storing fuel oil on a machine driven by a diesel or gasoline engine, especially an auxiliary tank or a disposable tank that can be used to increase the range or carry condensed oil. A fuel tank is a special container for storing hydraulic oil or hydraulic fluid in a hydraulic system. Its Chinese name is: fuel tank. In addition to storing oil in a hydraulic system, a fuel tank also plays a role in heat dissipation, separating bubbles in the oil, and precipitating impurities in the fuel tank.

[0003] There are still some problems with the existing fuel tanks during use. The fuel tanks are generally placed inside the car. When the car is running for a long time, bumps and collisions may cause leakage at the bottom of the fuel tank, and personnel cannot discover it in time, and the wasted fuel cannot be recycled, resulting in fuel waste. Summary of the invention

[0004] The purpose of the present invention is to provide a leak-proof fuel tank and an intelligent detection method thereof. The present invention has a simple structure and is easy to use. When a fuel tank leaks, it can be detected and processed in time and an alarm can be issued. At the same time, the leaked fuel can be recovered and processed to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A leakage-proof fuel tank and an intelligent detection method thereof, comprising a base, the outer side of the base is fixedly connected with a support rod, the top of the support rod is fixedly connected with a fuel tank body, the top of the fuel tank body is fixedly connected with an oil inlet pipe, the bottom of the fuel tank body is fixedly connected with a guide groove, a lifting groove is opened in the inner wall of the base, the lifting groove is slidably connected with a sliding rod, the outer side of the sliding rod is fixedly connected with a storage box, a sponge is installed inside the storage box, the bottom of the sliding rod is fixedly connected with a No. 1 return spring, the outer side of the storage box is fixedly connected with a moving rod, the bottom of the moving rod is fixedly connected with an extrusion cover, the outer side of the base is fixedly connected with a power box, the outer side of the power box is fixedly connected with a switch, the outer side of the power box is fixedly connected with a connecting plate, the top of the connecting plate is fixedly connected with a No. 2 return spring, and the outer side of the power box is fixedly connected with an alarm light.

[0007] Furthermore, an electric telescopic rod is fixedly connected to the outer side of the base, a pushing rod is fixedly connected to the outer side of the electric telescopic rod, an extrusion plate is fixedly connected to the bottom of the pushing rod, a conduit is fixedly connected to the outer side of the storage box, a collecting box is fixedly connected to the outer side of the conduit, a filter plate is installed inside the collecting box, an oil outlet pipe is fixedly connected to the outer side of the collecting box, an extraction pump is fixedly connected to the outer side of the oil outlet pipe, a return pipe is fixedly connected to the top of the extraction pump, and a controller is fixedly connected to the outer side of the base.

[0008] Furthermore, the support rods are symmetrically distributed on the outside of the base, and the fuel tank body is fixedly connected to the top of the base through the support rods, and the fuel tank body is fixedly connected to the top of the base through the support rods, so that the fuel tank body is located directly above the base.

[0009] Furthermore, the guide grooves are symmetrically distributed at the bottom of the fuel tank body, and the lifting grooves are symmetrically distributed in the inner wall of the base, and are fixedly connected to the bottom of the fuel tank body through the guide grooves. When leakage occurs at the bottom of the fuel tank body, the leakage will enter the interior of the sponge body through the guide grooves.

[0010] Furthermore, the storage box is installed inside the base through a sliding rod, and the bottom of the No. 1 return spring is fixedly connected to the inner bottom wall of the lifting slot. The sliding rod and the No. 1 return spring are used in coordination, so that the storage box is installed inside the base.

[0011] Furthermore, the sponge body is located directly below the guide groove, the extrusion cover is located directly above the switch, and the top of the No. 2 reset spring is fixedly connected below the switch. When the storage box moves downward inside the base, it will drive the moving rod and the extrusion cover to move.

[0012] Furthermore, the alarm light is electrically connected to the power box, and the alarm light is located on the outside of the base. When the switch is squeezed by the squeezing cover, the power box is started, and the alarm light on the outside of the power box will sound.

[0013] Furthermore, the controller is electrically connected to the extraction pump, and one end of the return pipe away from the extraction pump is fixedly connected to the outside of the fuel tank body. The extraction pump is started by the controller, so that the extraction pump extracts the fuel inside the collection tank.

[0014] Furthermore, the intelligent detection method for the anti-leakage fuel tank comprises the following specific steps:

[0015] S1. When leakage occurs at the bottom of the fuel tank body, the fuel enters the sponge body through the guide groove at the bottom of the fuel tank body, thereby increasing the gravity of the storage box;

[0016] S2, the movement of the storage box drives the extrusion cover at the bottom of the moving rod to move, thereby squeezing the switch;

[0017] S3. When the power supply box is powered on, the alarm light on the outside of the power supply box will light up because the power supply box is electrically connected to the alarm light through the PLC control device, thereby achieving the effect of intelligent alarm and detection;

[0018] S4. Inspection and acceptance: After processing, the workpiece is manually inspected and accepted.

[0019] Furthermore, in the intelligent detection method for the anti-leakage fuel tank, in S3, the detection of the anti-leakage fuel tank is realized by an intelligent control algorithm, and the specific steps are as follows:

[0020] Step 1: Multimodal data collection;

[0021] Pressure sensor: installed at the top and bottom of the fuel tank body (3) to monitor the pressure change of the internal fuel;

[0022] Temperature sensor: installed at the fuel inlet pipe (4) and the fuel outlet pipe (24) of the fuel tank to monitor the fuel temperature;

[0023] Humidity sensor: placed inside the storage box (8), monitoring the humidity change of the sponge (9) to detect whether there is fuel leakage adsorption;

[0024] Acoustic wave sensor: installed on the outside of the fuel tank body (3), used to detect the integrity of the tank body through acoustic wave echo;

[0025] Optical sensor: Installed on the outside of the fuel tank to capture images of the fuel tank surface to detect cracks or abnormal deformations on the surface;

[0026] Step 2: Construction and training of deep learning model;

[0027] Convolutional neural network (CNN) is used to process optical image data, and long short-term memory (LSTM) is used to process pressure, temperature, sound wave and humidity data. Convolutional neural network is used for image data processing. It extracts local features in the image through convolutional layers. The basic convolution operation can be expressed as:

[0028] ;

[0029] in, The output feature map is at position The value at is the input image, is the convolution kernel, Represents the input image With convolution kernel The convolution operation at position The result of Convolution kernel Sum all the elements of The input image is at position The value at The convolution kernel is at position The value at and Output the row and column indices of the feature map, and The row and column indices of the convolution kernel and the input image;

[0030] Convolutional Layer:

[0031] The first convolution layer: 64 3x3 convolution kernels, stride 1, using ReLU activation function;

[0032] The second convolution layer: 128 3x3 convolution kernels, stride 1, using ReLU activation function;

[0033] The third convolution layer: 256 3x3 convolution kernels, stride 1, using ReLU activation function;

[0034] Pooling layer:

[0035] Each convolutional layer is followed by a 2x2 max pooling layer to reduce the size of the feature map;

[0036] LSTM is used to process and predict time series data;

[0037] The forget gate is used to control the forgetting of part of the information in the memory cell of the previous time step:

[0038] ;

[0039] in is the output of the forget gate, is the weight matrix of the forget gate, For the previous time step The hidden state of is the current time step Input, is the bias term of the forget gate, is the Sigmoid activation function;

[0040] The input gate controls the amount of new information that enters the memory cell:

[0041] ;

[0042] in is the input gate output, is the weight matrix of the input gate, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the input gate, is the Sigmoid activation function;

[0043] The new information is used to update the contents of the memory cell:

[0044] ;

[0045] in For new information, To calculate the weight matrix of new information, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the new information, tanh is the hyperbolic tangent function, and is the activation function of the hidden layer;

[0046] Update memory cells and calculate new memory cell states:

[0047] ;

[0048] in is the state of memory cells, is the output of the forget gate, is the memory cell state at the previous time step, is the input gate output, For new information;

[0049] The output gate controls the information in the hidden state to be passed to the next time step:

[0050] ;

[0051] in is the output gate output, is the weight matrix of the output gate, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the output gate;

[0052] Hide status updates:

[0053] ;

[0054] is the hidden state at the current time step, is the output gate output, is the memory cell state at the current time step, tanh is the hyperbolic tangent function, and is the activation function of the hidden layer;

[0055] First layer LSTM: number of units: 64, activation function: tanh;

[0056] Second layer LSTM: number of units: 64, activation function: tanh;

[0057] The third layer LSTM: number of units: 32, activation function: tanh;

[0058] Fully connected layer:

[0059] Fully connected layer 1: Number of neurons: 128, activation function: ReLU;

[0060] Step 3: CNN and LSTM fusion

[0061] The fusion of CNN and LSTM is achieved through a fully connected layer, the output of which is calculated by the following formula:

[0062] ;

[0063] in, is the final prediction output, is the activation function ReLU, is the weight matrix of the fully connected layer, is the bias, is the output feature vector of CNN, is the output feature vector of LSTM, represents the concatenation of two eigenvectors;

[0064] Fully connected layer:

[0065] The first fully connected layer has 1024 neurons and uses the ReLU activation function.

[0066] The second fully connected layer has 512 neurons and uses the ReLU activation function.

[0067] Step 4: Anomaly detection algorithm;

[0068] Isolation Forest detects anomalies by comparing live data with the baseline model:

[0069] ;

[0070] ;

[0071] ;

[0072] ;

[0073] in, It is a data point The anomaly score, is the sample size, It is a data point Path length in an isolation forest The average value of is the normalization factor, is the number of trees in the isolation forest, It is a data point In the The path length in an isolated tree, It is A harmonic number, It is A harmonic number, is the index variable to be summed, From 1 to of The sum of the reciprocals of ;

[0074] Step 5: Adaptive adjustment;

[0075] Adaptive learning of the model is achieved through online gradient descent, allowing the model to continuously update its weights and biases based on new data to adapt to new operating conditions and environmental changes:

[0076] ;

[0077] in, is at the time step The model parameters when is at the time step The updated model parameters, is the learning rate, which controls the step size of parameter updates, is the loss function About parameters The gradient is calculated based on the data points at the current time step , and Represent the model prediction output and the true output respectively, is the loss function, expressed in terms of parameters Next, for input And the real output The loss value is Model Output About parameters The gradient of

[0078] The incoming data is continuously analyzed and once an anomaly is detected, the alarm system is triggered, activating the alarm light (17) and notifying the operator.

[0079] The beneficial effects of the present invention are:

[0080] (I) The present invention is provided with a base, and the fuel tank body is fixedly connected to the top of the base through the support rod on the outside of the base. When leakage occurs at the bottom of the fuel tank body, the leaked fuel enters the sponge body inside the storage box through the guide groove at the bottom of the fuel tank body. When more and more fuel leaks, the weight of the sponge body inside the storage box will also increase. At this time, with the cooperation of the slide rod and the No. 1 return spring, the storage box moves downward inside the base, and the movement of the storage box drives the moving rod to move, so that the moving rod drives the extrusion cover to move. When it moves to the bottom, it will squeeze the switch on the outside of the power box, so that the power box is powered on, and then the alarm light on the outside of the base will sound, achieving the effect of intelligent detection and alarm.

[0081] (ii) The present invention is provided with a base, and a storage box is installed inside the base. When the alarm light sends out an alarm reminder, the electric telescopic rod is started through the controller on the outside of the base, so that the electric telescopic rod drives the push rod to move. At this time, the push rod drives the extrusion plate to move to the top of the sponge body, so that the extrusion plate squeezes the sponge body inside the storage box. When the extrusion plate squeezes the sponge body, the fuel entering the sponge body will be squeezed out and then enter the inside of the collection box through the conduit. The extraction pump is started through the controller on the outside of the base. At this time, the extraction pump will extract the fuel inside the collection box through the oil outlet pipe, and a filter plate is fixedly connected to the inside of the collection box, so as to filter the fuel, and the extracted fuel will then enter the inside of the fuel tank again through the reflux pipe, thereby achieving a recycling effect and avoiding fuel waste.

[0082] (III) Through real-time data analysis and machine learning models, the system is able to predict potential leaks and issue an alarm before an actual leak occurs. This not only reduces environmental pollution, but also avoids potential safety risks. Emergency measures are automatically initiated at the early stage of leak detection, such as starting the extraction pump and adjusting the diversion groove. These automated response measures greatly reduce human delays and improve processing efficiency. Through precise data models and algorithms, the intelligent control system can effectively distinguish normal changes from abnormal leaks, reduce the occurrence of false alarms and missed alarms, and improve the reliability of the system. Intelligent algorithms can continuously learn and optimize based on historical data and real-time feedback, so that the system can adapt to changing operating conditions and environments and maintain efficient monitoring capabilities. The intelligent monitoring system reduces reliance on manual regular inspections and reduces labor costs. The system's automatic diagnosis and problem reporting functions help the maintenance team quickly locate problems and develop maintenance plans, shorten maintenance time, and improve maintenance efficiency. Timely leak detection and treatment help prevent fuel from leaking into the external environment and protect the ecological environment from pollution. By preventing and reducing leaks, the system helps companies meet stricter environmental regulations and standards and avoid possible legal and financial risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the specific embodiments of the present invention, but do not constitute a limitation of the present invention.

[0084] Figure 1 It is a structural schematic diagram of a leak-proof fuel tank of the present invention;

[0085] Figure 2 It is a structural schematic diagram of the back side of a leak-proof fuel tank of the present invention;

[0086] Figure 3 It is a structural schematic diagram of the outer side of a fuel tank body of a leak-proof fuel tank of the present invention;

[0087] Figure 4 It is a schematic diagram of the structure inside a leak-proof fuel tank base of the present invention;

[0088] Figure 5 It is a structural schematic diagram of the outer side of a leak-proof fuel tank storage box and a power box of the present invention;

[0089] Figure 6 It is a structural schematic diagram of the interior of a collection box of a leak-proof fuel tank and the outside of an extraction pump of the present invention;

[0090] Figure 7 The invention discloses an intelligent algorithm control flow chart of an anti-leakage fuel tank and an intelligent detection method thereof.

[0091] Numbers in the figure: 1. Base; 2. Support rod; 3. Fuel tank body; 4. Oil inlet pipe; 5. Guide groove; 6. Lifting groove; 7. Sliding rod; 8. Storage box; 9. Sponge; 10. No. 1 reset spring; 11. Moving rod; 12. Extrusion cover; 13. Power box; 14. Switch; 15. Connecting plate; 16. No. 2 reset spring; 17. Warning light; 18. Electric telescopic rod; 19. Push rod; 20. Extrusion plate; 21. Conduit; 22. Collecting box; 23. Filter plate; 24. Oil outlet pipe; 25. Extraction pump; 26. Return pipe; 27. Controller. DETAILED DESCRIPTION

[0092] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0093] Example 1 Please refer to Figure 1-Figure 6, the present invention provides a technical solution:

[0094] A leak-proof fuel tank and an intelligent detection method thereof, comprising a base 1, a support rod 2 is fixedly connected to the outside of the base 1, a fuel tank body 3 is fixedly connected to the top of the support rod 2, an oil inlet pipe 4 is fixedly connected to the top of the fuel tank body 3, a guide groove 5 is fixedly connected to the bottom of the fuel tank body 3, a lifting groove 6 is opened in the inner wall of the base 1, a slide rod 7 is slidably connected inside the lifting groove 6, a storage box 8 is fixedly connected to the outside of the slide rod 7, a sponge 9 is installed inside the storage box 8, a No. 1 return spring 10 is fixedly connected to the bottom of the slide rod 7, a moving rod 11 is fixedly connected to the outside of the storage box 8, an extrusion cover 12 is fixedly connected to the bottom of the moving rod 11, a power box 13 is fixedly connected to the outside of the base 1, a switch 14 is fixedly connected to the outside of the power box 13, a connecting plate 15 is fixedly connected to the outside of the power box 13, a No. 2 return spring 16 is fixedly connected to the top of the connecting plate 15, and an alarm light 17 is fixedly connected to the outside of the power box 13.

[0095] Specifically, Figure 1 As shown, the support rods 2 are symmetrically distributed on the outside of the base 1, the fuel tank body 3 is fixedly connected to the top of the base 1 through the support rods 2, the guide grooves 5 are symmetrically distributed on the bottom of the fuel tank body 3, the lifting grooves 6 are symmetrically distributed in the inner wall of the base 1, the storage box 8 is installed inside the base 1 through the slide rod 7, the bottom of the No. 1 return spring 10 is fixedly connected to the inner bottom wall of the lifting groove 6, the sponge 9 is located directly below the guide groove 5, the extrusion cover 12 is located directly above the switch 14, and the top of the No. 2 return spring 16 is fixedly connected below the switch 14.

[0096] Specifically: the fuel tank body 3 is fixedly connected to the top of the base 1 by the support rod 2, so that the fuel tank body 3 is located directly above the base 1, and is fixedly connected to the bottom of the fuel tank body 3 by the guide groove 5. When leakage occurs at the bottom of the fuel tank body 3, the leakage will enter the interior of the sponge body 9 through the guide groove 5. By using the slide rod 7 in conjunction with the No. 1 return spring 10, the storage box 8 is installed inside the base 1. When the storage box 8 moves downward inside the base 1, it will drive the moving rod 11 and the extrusion cover 12 to move.

[0097] In this embodiment: a base 1 is provided, and the fuel tank body 3 is fixedly connected to the top of the base 1 through the support rod 2 on the outside of the base 1. When leakage occurs at the bottom of the fuel tank body 3, the leaked fuel is allowed to enter the sponge 9 inside the storage box 8 through the guide groove 5 at the bottom of the fuel tank body 3. When more and more fuel is leaked, the gravity of the sponge 9 inside the storage box 8 will also increase. At this time, with the cooperation of the slide bar 7 and the No. 1 return spring 10, the storage box 8 is moved downward inside the base 1. The movement of the storage box 8 drives the moving rod 11 to move, so that the moving rod 11 drives the extrusion cover 12 to move. When it moves to the bottom, the switch 14 on the outside of the power box 13 is squeezed, so that the power box 13 is energized, and then the alarm light 17 on the outside of the base 1 sounds, achieving the effect of intelligent detection and alarm.

[0098] Example 2 Please refer to Figure 1 , Figure 2 and Figure 6 The difference between this embodiment and embodiment 1 is that: an electric telescopic rod 18 is fixedly connected to the outer side of the base 1, a pushing rod 19 is fixedly connected to the outer side of the electric telescopic rod 18, an extrusion plate 20 is fixedly connected to the bottom of the pushing rod 19, a conduit 21 is fixedly connected to the outer side of the storage box 8, a collecting box 22 is fixedly connected to the outer side of the conduit 21, a filter plate 23 is installed inside the collecting box 22, an oil outlet pipe 24 is fixedly connected to the outer side of the collecting box 22, an extraction pump 25 is fixedly connected to the outer side of the oil outlet pipe 24, a return pipe 26 is fixedly connected to the top of the extraction pump 25, and a controller 27 is fixedly connected to the outer side of the base 1.

[0099] Specifically, Figure 1-6 As shown, the warning light 17 is electrically connected to the power box 13 and is located on the outside of the base 1 , the controller 27 is electrically connected to the extraction pump 25 , and the end of the return pipe 26 away from the extraction pump 25 is fixedly connected to the outside of the fuel tank body 3 .

[0100] Specifically, when the switch 14 is squeezed by the squeezing cover 12, the power box 13 is started, and the alarm light 17 on the outside of the power box 13 sounds, and the extraction pump 25 is started through the controller 27, so that the extraction pump 25 extracts the fuel inside the collection box 22.

[0101] In this embodiment: by setting up a base 1, a storage box 8 is installed inside the base 1. When the alarm light 17 sends out an alarm reminder, the electric telescopic rod 18 is started through the controller 27 outside the base 1, so that the electric telescopic rod 18 drives the push rod 19 to move. At this time, the push rod 19 drives the extrusion plate 20 to move to the top of the sponge 9, so that the extrusion plate 20 squeezes the sponge 9 inside the storage box 8. When the extrusion plate 20 squeezes the sponge 9, the fuel entering the sponge 9 will be squeezed out, and then enter the inside of the collection box 22 through the conduit 21. The extraction pump 25 is started by the controller 27 outside the base 1. At this time, the extraction pump 25 will extract the fuel inside the collection box 22 through the oil outlet pipe 24, and the filter plate 23 is fixedly connected to the inside of the collection box 22, so as to filter the fuel, and then the extracted fuel will enter the inside of the fuel tank body 3 again through the return pipe 26, thereby achieving a recycling effect and avoiding fuel waste.

[0102] A leak-proof fuel tank and an intelligent detection method thereof, the specific steps are as follows:

[0103] S1. When leakage occurs at the bottom of the fuel tank body 3, the fuel enters the sponge body 9 through the guide groove 5 at the bottom of the fuel tank body 3, thereby increasing the gravity of the storage box 8;

[0104] S2, the movement of the storage box 8 drives the extrusion cover 12 at the bottom of the moving rod 11 to move, thereby squeezing the switch 14;

[0105] S3. When the power supply box 13 is powered on, the power supply box 13 is electrically connected to the alarm light 17 through the PLC control device, so that the alarm light 17 on the outside of the power supply box 13 lights up, thereby achieving the effect of intelligent alarm and detection;

[0106] S4. Inspection and acceptance: After processing, the workpiece is manually inspected and accepted.

[0107] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0108] In this embodiment: a leak-proof fuel tank and an intelligent detection method thereof, the detection of the leak-proof fuel tank is realized through an intelligent control algorithm, and the specific steps are as follows:

[0109] Step 1: Multimodal data collection;

[0110] Pressure sensor: installed at the top and bottom of the fuel tank body (3) to monitor the pressure change of the internal fuel;

[0111] Temperature sensor: installed at the fuel inlet pipe (4) and the fuel outlet pipe (24) of the fuel tank to monitor the fuel temperature;

[0112] Humidity sensor: placed inside the storage box (8), monitoring the humidity change of the sponge (9) to detect whether there is fuel leakage adsorption;

[0113] Acoustic wave sensor: installed on the outside of the fuel tank body (3), used to detect the integrity of the tank body through acoustic wave echo;

[0114] Optical sensor: Installed on the outside of the fuel tank to capture images of the fuel tank surface to detect cracks or abnormal deformations on the surface;

[0115] Step 2: Construction and training of deep learning model;

[0116] Convolutional neural network (CNN) is used to process optical image data, and long short-term memory (LSTM) is used to process pressure, temperature, sound wave and humidity data. Convolutional neural network is used for image data processing. It extracts local features in the image through convolutional layers. The basic convolution operation can be expressed as:

[0117] ;

[0118] in, The output feature map is at position The value at is the input image, is the convolution kernel, Represents the input image With convolution kernel The convolution operation at position The result of Convolution kernel Sum all the elements of The input image is at position The value at The convolution kernel is at position The value at and Output the row and column indices of the feature map, and The row and column indices of the convolution kernel and the input image;

[0119] Convolutional Layer:

[0120] The first convolution layer: 64 3x3 convolution kernels, stride 1, using ReLU activation function;

[0121] The second convolution layer: 128 3x3 convolution kernels, stride 1, using ReLU activation function;

[0122] The third convolution layer: 256 3x3 convolution kernels, stride 1, using ReLU activation function;

[0123] Pooling layer:

[0124] Each convolutional layer is followed by a 2x2 max pooling layer to reduce the size of the feature map;

[0125] LSTM is used to process and predict time series data;

[0126] The forget gate is used to control the forgetting of part of the information in the memory cell of the previous time step:

[0127] ;

[0128] in is the output of the forget gate, is the weight matrix of the forget gate, For the previous time step The hidden state of is the current time step Input, is the bias term of the forget gate, is the Sigmoid activation function;

[0129] The input gate controls the amount of new information that enters the memory cell:

[0130] ;

[0131] in is the input gate output, is the weight matrix of the input gate, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the input gate, is the Sigmoid activation function;

[0132] The new information is used to update the contents of the memory cell:

[0133] ;

[0134] in For new information, To calculate the weight matrix of new information, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the new information, tanh is the hyperbolic tangent function, and is the activation function of the hidden layer;

[0135] Update memory cells and calculate new memory cell states:

[0136] ;

[0137] in is the state of memory cells, is the output of the forget gate, is the memory cell state at the previous time step, is the input gate output, For new information;

[0138] The output gate controls the information in the hidden state to be passed to the next time step:

[0139] ;

[0140] in is the output gate output, is the weight matrix of the output gate, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the output gate;

[0141] Hide status updates:

[0142] ;

[0143] is the hidden state at the current time step, is the output gate output, is the memory cell state at the current time step, tanh is the hyperbolic tangent function, and is the activation function of the hidden layer;

[0144] First layer LSTM: number of units: 64, activation function: tanh;

[0145] Second layer LSTM: number of units: 64, activation function: tanh;

[0146] The third layer LSTM: number of units: 32, activation function: tanh;

[0147] Fully connected layer:

[0148] Fully connected layer 1: Number of neurons: 128, activation function: ReLU;

[0149] Step 3: CNN and LSTM fusion

[0150] The fusion of CNN and LSTM is achieved through a fully connected layer, the output of which is calculated by the following formula:

[0151] ;

[0152] in, is the final prediction output, is the activation function ReLU, is the weight matrix of the fully connected layer, is the bias, is the output feature vector of CNN, is the output feature vector of LSTM, represents the concatenation of two eigenvectors;

[0153] Fully connected layer:

[0154] The first fully connected layer has 1024 neurons and uses the ReLU activation function.

[0155] The second fully connected layer has 512 neurons and uses the ReLU activation function.

[0156] Step 4: Anomaly detection algorithm;

[0157] Isolation Forest detects anomalies by comparing live data with the baseline model:

[0158] ;

[0159] ;

[0160] ;

[0161] ;

[0162] in, It is a data point The anomaly score, is the sample size, It is a data point Path length in an isolation forest The average value of is the normalization factor, is the number of trees in the isolation forest, It is a data point In the The path length in an isolated tree, It is A harmonic number, It is A harmonic number, is the index variable to be summed, From 1 to of The sum of the reciprocals of ;

[0163] Step 5: Adaptive adjustment;

[0164] Adaptive learning of the model is achieved through online gradient descent, allowing the model to continuously update its weights and biases based on new data to adapt to new operating conditions and environmental changes:

[0165] ;

[0166] in, is at the time step The model parameters when is at the time step The updated model parameters, is the learning rate, which controls the step size of parameter updates, is the loss function About parameters The gradient is calculated based on the data points at the current time step , and Represent the model prediction output and the true output respectively, is the loss function, expressed in terms of parameters Next, for input And the real output The loss value is Model Output About parameters The gradient of

[0167] The incoming data is continuously analyzed and once an anomaly is detected, the alarm system is triggered, activating the alarm light (17) and notifying the operator.

[0168] Numerical case analysis:

[0169] During a monitoring cycle, the sensor collects the following data:

[0170] Fuel tank top and bottom pressure sensors: Normal range is 1.5 to 2.5 atm. A sharp drop will be seen in case of a leak.

[0171] Oil inlet and outlet pipe temperature sensors: The normal operating temperature range is 25 to 35°C. There will be abnormal fluctuations when there is a leak.

[0172] Humidity sensor inside the glove box: Normal humidity is below 50% RH. In the event of a leak, humidity can increase significantly as fuel enters the sponge.

[0173] Fuel tank outer acoustic sensor: Under normal circumstances, the echo is stable. Leakage will cause the echo pattern to change because fuel leakage affects the propagation of sound waves.

[0174] Fuel tank outer optical sensor: Normally there is no crack or deformation. Leakage may cause deformation or cracks on the fuel tank surface.

[0175] During a specific monitoring cycle, the sensor readings are as follows:

[0176] Pressure sensor: 1.9 atm at the top, 1.8 atm at the bottom;

[0177] Temperature sensor: oil inlet pipe 30°C, oil outlet pipe 29°C;

[0178] Humidity sensor: 55% RH, indicating a leak;

[0179] Acoustic sensor: echo response time increased by 5%, indicating a possible leak;

[0180] Optical sensor: Image analysis showed no obvious cracks or deformation;

[0181] CNN is used to extract features from the image, LSTM is used to process the time series data, and the prediction model outputs a high probability of leakage, and the alarm light (17) is activated, notifying the operator to perform further inspection and maintenance.

Claims

1. An intelligent detection method for a leak-proof fuel tank, the fuel tank comprising a base (1), Features: The outer side of the base (1) is fixedly connected to a support rod (2), the top of the support rod (2) is fixedly connected to a fuel tank body (3), the top of the fuel tank body (3) is fixedly connected to an oil inlet pipe (4), the bottom of the fuel tank body (3) is fixedly connected to a guide groove (5), a lifting groove (6) is provided in the inner wall of the base (1), a sliding rod (7) is slidably connected inside the lifting groove (6), a storage box (8) is fixedly connected to the outer side of the sliding rod (7), a sponge (9) is installed inside the storage box (8), and the sliding rod (7) is fixedly connected to the outer side of the storage box (8). A No. 1 return spring (10) is fixedly connected to the bottom, a moving rod (11) is fixedly connected to the outside of the storage box (8), an extrusion cover (12) is fixedly connected to the bottom of the moving rod (11), a power box (13) is fixedly connected to the outside of the base (1), a switch (14) is fixedly connected to the outside of the power box (13), a connecting plate (15) is fixedly connected to the outside of the power box (13), a No. 2 return spring (16) is fixedly connected to the top of the connecting plate (15), and an alarm light (17) is fixedly connected to the outside of the power box (13); The outer side of the base (1) is fixedly connected to an electric telescopic rod (18), the outer side of the electric telescopic rod (18) is fixedly connected to a push rod (19), the bottom of the push rod (19) is fixedly connected to an extrusion plate (20), the outer side of the storage box (8) is fixedly connected to a guide tube (21), the outer side of the guide tube (21) is fixedly connected to a collection box (22), a filter plate (23) is installed inside the collection box (22), the outer side of the collection box (22) is fixedly connected to an oil outlet pipe (24), the outer side of the oil outlet pipe (24) is fixedly connected to an extraction pump (25), the top of the extraction pump (25) is fixedly connected to a return pipe (26), and the outer side of the base (1) is fixedly connected to a controller (27); The specific steps of the method are as follows: S1. When leakage occurs at the bottom of the fuel tank body (3), the fuel enters the interior of the sponge body (9) through the guide groove (5) at the bottom of the fuel tank body (3), thereby increasing the gravity of the storage box (8); S2, the movement of the storage box (8) drives the extrusion cover (12) at the bottom of the moving rod (11) to move, thereby squeezing the switch (14); S3, when the power supply box (13) is powered on, since the power supply box (13) is electrically connected to the alarm light (17) through the PLC control device, the alarm light (17) on the outside of the power supply box (13) lights up, thereby achieving the effect of intelligent alarm and detection; S4, inspection and acceptance. After the processing is completed, the workpiece is manually inspected and accepted; S3 also includes the detection of leak-proof fuel tanks through intelligent control algorithms. The specific steps are as follows: Step 1: Multimodal data collection; Pressure sensor: installed at the top and bottom of the fuel tank body (3) to monitor the pressure change of the internal fuel; Temperature sensor: installed at the fuel inlet pipe (4) and the fuel outlet pipe (24) of the fuel tank to monitor the fuel temperature; Humidity sensor: placed inside the storage box (8), monitoring the humidity change of the sponge (9) to detect whether there is fuel leakage adsorption; Acoustic wave sensor: installed on the outside of the fuel tank body (3), used to detect the integrity of the tank body through acoustic wave echo; Optical sensor: Installed on the outside of the fuel tank to capture images of the fuel tank surface to detect cracks or abnormal deformations on the surface; Step 2: Construction and training of deep learning model; Convolutional neural network (CNN) is used to process optical image data, and long short-term memory (LSTM) is used to process pressure, temperature, sound wave and humidity data. Convolutional neural network is used for image data processing. It extracts local features in the image through convolutional layers. The basic convolution operation is expressed as: ； in, The output feature map is at position The value at is the input image, is the convolution kernel, Represents the input image With convolution kernel The convolution operation at position The result of Convolution kernel Sum all the elements of The input image is at position The value at The convolution kernel is at position The value at and Output the row and column indices of the feature map, and The row and column indices of the convolution kernel and the input image; Convolutional Layer: The first convolution layer: 64 3x3 convolution kernels, stride 1, using ReLU activation function; The second convolution layer: 128 3x3 convolution kernels, stride 1, using ReLU activation function; The third convolution layer: 256 3x3 convolution kernels, stride 1, using ReLU activation function; Pooling layer: Each convolutional layer is followed by a 2x2 max pooling layer to reduce the size of the feature map; LSTM is used to process and predict time series data; The forget gate is used to control the forgetting of part of the information in the memory cell of the previous time step: ； in is the output of the forget gate, is the weight matrix of the forget gate, For the previous time step The hidden state of is the current time step Input, is the bias term of the forget gate, σ is the Sigmoid activation function; The input gate controls the amount of new information that enters the memory cell: ； in is the input gate output, is the weight matrix of the input gate, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the input gate, σ is the Sigmoid activation function; The new information is used to update the contents of the memory cell: ； in For new information, To calculate the weight matrix of new information, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the new information, tanh is the hyperbolic tangent function, and is the activation function of the hidden layer; Update memory cells and calculate new memory cell states: ； in is the state of memory cells, is the output of the forget gate, is the memory cell state at the previous time step, is the input gate output, For new information; The output gate controls the information in the hidden state to be passed to the next time step: ； in is the output gate output, is the weight matrix of the output gate, is the hidden state of the previous time step, is the input of the current time step, is the bias term of the output gate; Hide status updates: ； is the hidden state at the current time step, is the output gate output, is the memory cell state at the current time step, tanh is the hyperbolic tangent function, and is the activation function of the hidden layer; First layer LSTM: number of units: 64, activation function: tanh; Second layer LSTM: number of units: 64, activation function: tanh; The third layer LSTM: number of units: 32, activation function: tanh; Fully connected layer: Fully connected layer 1: Number of neurons: 128, activation function: ReLU; Step 3: CNN and LSTM fusion The fusion of CNN and LSTM is achieved through a fully connected layer, the output of which is calculated by the following formula: ； in, is the final prediction output, is the activation function ReLU, is the weight matrix of the fully connected layer, is the bias, is the output feature vector of CNN, is the output feature vector of LSTM, represents the concatenation of two eigenvectors; Fully connected layer: The first fully connected layer has 1024 neurons and uses the ReLU activation function. The second fully connected layer has 512 neurons and uses the ReLU activation function. Step 4: Anomaly detection algorithm; Isolation Forest detects anomalies by comparing live data with the baseline model: ； ； ； ； in, It is a data point The anomaly score, is the sample size, It is a data point Path length in an isolation forest The average value of is the normalization factor, is the number of trees in the isolation forest, It is a data point In the The path length in an isolated tree, It is A harmonic number, It is A harmonic number, is the index variable to be summed, From 1 to of The sum of the reciprocals of ; Step 5: Adaptive adjustment; Adaptive learning of the model is achieved through online gradient descent, allowing the model to continuously update its weights and biases based on new data to adapt to new operating conditions and environmental changes: ； ； ； in, is at the time step The model parameters when is at the time step The updated model parameters, is the learning rate, which controls the step size of parameter updates, is the loss function About parameters The gradient is calculated based on the data points at the current time step , and Represent the model prediction output and the true output respectively, is the loss function, expressed in terms of parameters Next, for input And the real output The loss value is Model Output About parameters The gradient of The incoming data is continuously analyzed and once an anomaly is detected, the alarm system is triggered, activating the alarm light (17) and notifying the operator.

2. The method according to claim 1, Features: The support rods (2) are symmetrically distributed on the outside of the base (1), and the fuel tank body (3) is fixedly connected to the top of the base (1) via the support rods (2).

3. The method according to claim 1, Features: The guide grooves (5) are symmetrically distributed on the bottom of the fuel tank body (3), and the lifting grooves (6) are symmetrically distributed on the inner wall of the base (1).

4. The method according to claim 1, Features: The storage box (8) is installed inside the base (1) via a slide rod (7), and the bottom of the No. 1 return spring (10) is fixedly connected to the inner bottom wall of the lifting groove (6).

5. The method according to claim 1, Features: The sponge body (9) is located directly below the guide groove (5), the extrusion cover (12) is located directly above the switch (14), and the top of the No. 2 return spring (16) is fixedly connected to the bottom of the switch (14).

6. The method according to claim 1, Features: The warning light (17) is electrically connected to the power box (13), and the warning light (17) is located outside the base (1).

7. The method according to claim 1, Features: The controller (27) is electrically connected to the extraction pump (25), and one end of the return pipe (26) away from the extraction pump (25) is fixedly connected to the outside of the fuel tank body (3).

Images