Locomotive excrement collector intelligent control system and method

Abstract

The invention relates to the technical field of intelligent control, and discloses an intelligent control system and method for a locomotive excrement collector, the system comprises a multi-source sensing module, an edge calculation module, a decision optimization module and an execution control module, by constructing a multi-source sensing fusion mechanism, three-dimensional point cloud and chemical distribution characteristics are integrated during sewage treatment flow control, and a multi-source sensing fusion mechanism is established; a dirt state dynamic model is established, the accuracy of dirt feature recognition under different working conditions is guaranteed, meanwhile, environment parameters and dirt phase states are subjected to real-time correlation analysis, the problems of solid layering and vacuum degree abnormity in a dirt box can be sensed in real time, the flushing mode misjudgment risk is reduced, the thoroughness of flushing operation is guaranteed, and water resource consumption is reduced. The reliability of the excrement collection system is improved, and when an excrement collection operation decision is made, the dirt fermentation dynamic characteristics are analyzed, and the microbial activity change trend is dynamically calculated, so that the system can pre-judge the dirt accumulation critical state, and the equipment fault caused by overload of the dirt box is effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, specifically to an intelligent control system and method for locomotive toilets. Background Technology

[0002] Intelligent control is a control method with intelligent information processing, intelligent information feedback and intelligent control decision-making. It is an advanced stage in the development of control theory and is mainly used to solve control problems of complex systems that are difficult to solve using traditional methods. The main characteristics of intelligent control research objects are mathematical models with uncertainty, high nonlinearity and complex task requirements.

[0003] The intelligent control system and method for locomotive toilets is a highly efficient management system integrating sensing technology, intelligent decision-making, and automatic control. It aims to optimize the process of locomotive waste treatment and improve the overall sanitation level. This is one of the core components of the system, responsible for real-time monitoring of the waste tank status and coordinating the operation of various actuators. A high-precision sensor network can acquire key data such as waste phase distribution, liquid level, and environmental parameters to ensure accurate identification of waste status changes and system anomalies during the treatment process. To ensure the accuracy of control commands and the stability of system operation, the system is equipped with a dynamic decision engine and a biomimetic optimization algorithm to ensure the efficiency and reliability of locomotive toilet operation.

[0004] Currently, traditional waste collection systems rely on discrete function switches and simple control logic. When controlling the waste treatment process, valve actions are triggered by level and pressure switches, making it impossible to perceive the fermentation state and distribution characteristics of waste in real time. When there is solid-liquid mixing and stratification in the waste tank or abnormal vacuum levels, the flushing operation mode will be selected inappropriately, resulting in incomplete cleaning and water waste. At the same time, during the waste collection process, no waste state evolution model has been established, making it difficult to dynamically predict waste accumulation trends and microbial activity. This can lead to waste tank overload and frequent start-stop of vacuum valves. Furthermore, the system cannot adaptively adjust strategies when system parameters are abnormal. In the fault diagnosis stage, existing devices mainly rely on the experience of maintenance personnel to judge faulty components, lacking a multi-parameter fusion diagnostic mechanism for solenoid valves, sensors, and power supply status. This results in the need to replace components one by one until the function is restored during maintenance, increasing operation and maintenance costs and reducing system availability. In addition, the hardware architecture and communication protocols of waste collector controllers in different vehicle models are not uniform, resulting in the inability to universalize the core control module, further increasing manufacturing and maintenance costs.

[0005] Therefore, an intelligent control system and method for locomotive toilets are proposed to solve the above problems. Summary of the Invention

[0006] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an intelligent control system and method for locomotive toilets, solving the problems mentioned in the background section.

[0007] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: an intelligent control system and method for locomotive toilets, the method comprising the following steps: S1. Collect internal state data and environmental parameters of the waste bin through a multi-source sensor array to generate a raw dataset of waste status. S2. Perform multimodal feature fusion processing on the original dataset of the dirt state to generate a dirt state fusion feature matrix; S3. Based on the improved YOLOv7 algorithm, target segmentation processing is performed on the image inside the sludge bin to generate a heat map of sludge distribution and liquid level contour data; S4. Construct a waste state evolution prediction model, combine historical operation data to conduct time series analysis on waste fermentation state, and generate a waste fermentation activity index. S5. Input the sludge state fusion feature matrix, sludge distribution heat map, liquid level contour data and sludge fermentation activity index into the dynamic decision engine to generate the optimal operation mode instruction set. S6. Based on the biological behavior simulation algorithm, adaptively optimize the operation mode instruction set to generate biomimetic optimized operation parameters; S7. The actuator control module drives the toilet valve, flushing pump and deodorization device to work together to complete the closed-loop control of the entire sewage treatment process.

[0008] Preferably, S1 includes: S11. Collect three-dimensional point cloud data inside the waste bin using a 3D-ToF sensor array to generate a spatial distribution model of waste volume. S12. Use a multispectral imager to obtain the distribution map of chemical composition on the surface of the dirt and generate an ammonia nitrogen concentration gradient matrix. S13. Deploy a temperature and humidity composite sensor to monitor the microenvironment parameters inside the chamber in real time and generate an environmental state vector. ,in Indicates the temperature value. Indicates humidity value. This indicates the air pressure value.

[0009] Preferably, S2 includes: S21. Establish the waste characteristic tensor ,in Represents 3D point cloud data. Let E represent the ammonia nitrogen concentration gradient matrix, and E represent the environmental state vector. S22. Perform cross-modal correlation analysis on feature tensors using a convolutional feature pyramid network to generate a fused feature matrix. in, For convolution operations, For feature fusion operators, This refers to the attention mechanism module; the meanings of the remaining characters are the same as above.

[0010] Preferably, S3 includes: S31. Construct a dual-branch segmentation network: Spatial branching uses dilated convolution to extract dirt boundary features; Semantic branches parse the dirt material properties using a Transformer encoder; S32, Output Fouling Phase Distribution Diagram and liquid level height time series curve ; in, This is an area containing solid waste. This area contains liquid contaminants. This area contains gaseous contaminants. It is a function of liquid level height.

[0011] Preferably, S4 includes: S41. Establish the equation for the evolution of the state of pollutants: Where A represents the fermentation activity index. Indicates the concentration of organic matter. , , The biochemical reaction coefficient, This is the temperature value. This is the humidity value; S42. Predicting the future using LSTM networks Activity index during the period .

[0012] Preferably, S5 includes: S51. Define the job mode decision matrix: in, Parameters for controlling the intensity of the flushing water flow, Parameters for controlling the degree of valve opening, The power level of the deodorization device, These are the decision variables, corresponding to the specific values ​​of flushing intensity, valve opening, and deodorization level, respectively. S52. Generating decision weights based on a fuzzy rule base: in This is the sensitivity coefficient. For feature threshold, Let i be the i-th feature of the fused feature matrix.

[0013] Preferably, S6 includes: S61. Initialize the osprey population and optimize its location. ; S62, Location Update During Exploration Phase: in It is a random number. , The current optimal position represents the optimal combination of decision parameters in the algorithm's history; S63. Development Phase Location Optimization: in The convergence factor is For the number of iterations, The location boundaries represent the upper and lower limits of the decision parameters.

[0014] Preferably, S6 further includes: S64. Construct the fitness function: in, These are weighting coefficients, corresponding to the relative importance of energy consumption, operation time, and cleanliness, respectively. This is an energy consumption indicator, representing the energy consumed during rinsing and deodorization. This is a task time indicator, representing the time spent on a single task. As an indicator of cleanliness, it quantifies the effectiveness of dirt removal; S65, when satisfied Output optimal parameters at the time .

[0015] Preferred, including: Multi-source sensing module, including 3D-ToF sensor array, multispectral imager and environmental sensor; The edge computing module, equipped with an FPGA chip, enables real-time feature fusion; The decision optimization module includes a dynamic decision engine and a biomimetic optimization processor; The execution control module drives the solenoid valve group, variable frequency water pump and plasma deodorization device.

[0016] Preferably, the edge computing module adopts a heterogeneous computing architecture: Neural network acceleration units process image segmentation tasks; The matrix coprocessor performs F* feature fusion; The time series prediction unit runs the LSTM model independently; Each computing unit exchanges data via the NoC on-chip network.

[0017] (III) Beneficial Effects Compared with the prior art, the present invention provides an intelligent control system and method for locomotive toilets, which has the following beneficial effects: 1. In this invention, by constructing a multi-source sensing fusion mechanism, three-dimensional point cloud and chemical distribution features are integrated to establish a dynamic model of waste state when controlling the waste treatment process. This ensures the accuracy of waste feature identification under different working conditions. At the same time, environmental parameters and waste phase state are correlated and analyzed in real time. This enables real-time perception of solid stratification and vacuum abnormalities in the waste tank, reduces the risk of misjudgment of flushing mode, ensures the thoroughness of flushing operations, reduces water consumption, and improves the reliability of the waste collection system.

[0018] 2. In this invention, by designing a sludge activity prediction module, when making decisions on sludge collection operations, the system analyzes the kinetic characteristics of sludge fermentation and dynamically calculates the trend of microbial activity changes. This enables the system to predict the critical state of sludge accumulation, effectively avoiding equipment failures caused by sludge tank overload. Furthermore, when the vacuum valve pressure fluctuates abnormally, the system can automatically adjust the flushing frequency and valve action sequence based on a real-time evolution model, achieving adaptive closed-loop adjustment of abnormal system parameters, reducing the need for manual intervention and extending the service life of key components.

[0019] 3. In this invention, by adopting a biomimetic optimization decision architecture, the flushing intensity, valve opening and deodorization level are simultaneously optimized by simulating biological behavior patterns when optimizing the control system parameters. This enables collaborative control of multiple actuators and dynamically matches differentiated operation strategies according to the distribution characteristics of different phases of contaminants. This allows the system to accurately suppress the vacuum suppression effect caused by solid-liquid mixing and stratification, reduce the problem of incomplete cleaning caused by parameter adjustment lag in traditional methods, improve hygiene levels and reduce operation and maintenance costs. Attached Figure Description

[0020] Figure 1 This is a flowchart of the intelligent control method for locomotive toilets according to the present invention; Figure 2 This is a schematic diagram of the intelligent control system for locomotive toilets of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 The intelligent control system and method for the locomotive toilet system includes the following steps: S1. Collect internal state data and environmental parameters of the waste bin through a multi-source sensor array to generate a raw dataset of waste status. S2. Perform multimodal feature fusion processing on the original dataset of dirt state to generate a dirt state fusion feature matrix; S3. Based on the improved YOLOv7 algorithm, target segmentation processing is performed on the image inside the sludge bin to generate a heat map of sludge distribution and liquid level contour data; S4. Construct a waste state evolution prediction model, combine historical operation data to conduct time series analysis on waste fermentation state, and generate a waste fermentation activity index. S5. Input the sludge state fusion feature matrix, sludge distribution heat map, liquid level profile data and sludge fermentation activity index into the dynamic decision engine to generate the optimal operation mode instruction set. S6. Based on the biological behavior simulation algorithm, adaptively optimize the operation mode instruction set to generate biomimetic optimized operation parameters; S7. The actuator control module drives the toilet valve, flushing pump and deodorization device to work together to complete the closed-loop control of the entire sewage treatment process. S1 includes: S11. Collect three-dimensional point cloud data inside the waste bin using a 3D-ToF sensor array to generate a spatial distribution model of waste volume. S12. Use a multispectral imager to obtain the distribution map of chemical composition on the surface of the dirt and generate an ammonia nitrogen concentration gradient matrix. S13. Deploy a temperature and humidity composite sensor to monitor the microenvironment parameters inside the chamber in real time and generate an environmental state vector. ,in Indicates the temperature value. Indicates humidity value. Indicates air pressure value; S2 includes: S21. Establish the waste characteristic tensor ,in Represents 3D point cloud data. Let E represent the ammonia nitrogen concentration gradient matrix, and E represent the environmental state vector. S22. Perform cross-modal correlation analysis on feature tensors using a convolutional feature pyramid network to generate a fused feature matrix. in, For convolution operations, For feature fusion operators, This refers to the attention mechanism module; the meanings of the remaining characters are the same as above. S3 includes: S31. Construct a dual-branch segmentation network: Spatial branching uses dilated convolution to extract dirt boundary features; Semantic branches parse the dirt material properties using a Transformer encoder; S32, Output Fouling Phase Distribution Diagram and liquid level height time series curve ; in, This is an area containing solid waste. This area contains liquid contaminants. This area contains gaseous contaminants. It is a function of liquid level height; S4 includes: S41. Establish the equation for the evolution of the state of pollutants: Where A represents the fermentation activity index. Indicates the concentration of organic matter. , , The biochemical reaction coefficient, This is the temperature value. This is the humidity value; S42. Predicting the future using LSTM networks Activity index during the period ; S5 includes: S51. Define the job mode decision matrix: in, Parameters for controlling the intensity of the flushing water flow, Parameters for controlling the degree of valve opening, The power level of the deodorization device, These are the decision variables, corresponding to the specific values ​​of flushing intensity, valve opening, and deodorization level, respectively. S52. Generating decision weights based on a fuzzy rule base: in This is the sensitivity coefficient. For feature threshold, This represents the i-th feature of the fused feature matrix; S6 includes: S61. Initialize the osprey population and optimize its location. ; S62, Location Update During Exploration Phase: in It is a random number. , The current optimal position represents the optimal combination of decision parameters in the algorithm's history; S63. Development Phase Location Optimization: in The convergence factor is For the number of iterations, The location boundaries represent the upper and lower limits of the decision parameters; S6 also includes: S64. Construct the fitness function: in, These are weighting coefficients, corresponding to the relative importance of energy consumption, operation time, and cleanliness, respectively. This is an energy consumption indicator, representing the energy consumed during rinsing and deodorization. This is a task time indicator, representing the time spent on a single task. As an indicator of cleanliness, it quantifies the effectiveness of dirt removal; S65, when satisfied Output optimal parameters at the time ; include: Multi-source sensing module, including 3D-ToF sensor array, multispectral imager and environmental sensor; The edge computing module, equipped with an FPGA chip, enables real-time feature fusion; The decision optimization module includes a dynamic decision engine and a biomimetic optimization processor; The execution control module drives the solenoid valve group, the variable frequency water pump, and the plasma deodorization device; The edge computing module adopts a heterogeneous computing architecture: Neural network acceleration units process image segmentation tasks; The matrix coprocessor performs F* feature fusion; The time series prediction unit runs the LSTM model independently; Each computing unit exchanges data via the NoC on-chip network.

[0023] Example 1: Detailed Implementation Process of Intelligent Control System and Method for Locomotive Toilet System A multi-source sensing module is installed at the bottom of the locomotive carriage. This module consists of a 3D-ToF sensor array, a multispectral imager, and a temperature and humidity composite sensor. When the toilet starts operating, the 3D-ToF sensor first scans the internal structure of the waste tank, generates a spatial distribution model of waste volume, and marks the solid deposition area. At the same time, the multispectral imager performs chemical analysis on the waste surface and outputs ammonia nitrogen concentration gradient data in real time. The temperature and humidity composite sensor synchronously collects micro-environmental parameters inside the tank, including temperature, humidity, and air pressure values, forming an environmental state vector. All raw data is transmitted to the FPGA chip of the edge computing module via the CAN bus.

[0024] The edge computing module uses a heterogeneous architecture to process data: the neural network acceleration unit segments the multispectral image, identifies liquid and gaseous regions in the waste, and outputs a waste phase distribution map; the matrix coprocessor fuses 3D point cloud data, chemical distribution matrix, and environmental vectors to construct a waste feature tensor; the time series prediction unit runs an LSTM model based on historical operation data, analyzes the trend of waste fermentation activity changes, and predicts the microbial activity index for future periods; the processed fused features generate an initial instruction set through a dynamic decision engine, matching a high-pressure flushing mode for areas with high solid deposition and activating a strong deodorization level for areas with high ammonia nitrogen concentration.

[0025] The decision optimization module adopts a biomimetic optimization strategy: the initialization of the osprey population position simulates different combinations of operating parameters. In the exploration phase, the algorithm simulates the osprey search behavior, randomly adjusts the parameter combination and evaluates its performance. In the development phase, the algorithm converges to the optimal solution, generates biomimetic optimization parameters that take into account energy consumption, operation time and cleanliness, and finally outputs instructions to drive the actuator. The solenoid valve group controls the sewage discharge path according to the optimized opening degree, the variable frequency water pump dynamically adjusts the flushing water pressure according to the solid zone distribution, and the plasma deodorization device starts in different areas according to the chemical concentration gradient.

[0026] The system's closed-loop control is achieved through real-time feedback: During operation, the multi-source sensing module continuously monitors changes in the state of contaminants. When the liquid level profile data rises after rinsing, the edge computing module immediately triggers secondary analysis to determine whether vacuum suppression is caused by solid-liquid mixing and stratification. Based on this, the decision engine calls the backup strategy, extending the rinsing time and increasing the valve opening. When the temperature and humidity sensor detects an increase in the air pressure inside the chamber, the system automatically reduces the rinsing intensity and increases the deodorization level to prevent gaseous contaminants from splashing. The fault diagnosis unit simultaneously scans the status of the actuators. When the solenoid valve is stuck or the water pump current exceeds the limit, it automatically switches to redundant equipment and pushes maintenance alarms to the vehicle monitoring terminal.

[0027] Example 2: Implementation process of the intelligent control system for locomotive toilets in high-speed train sets. After the train entered the maintenance station, the operators activated the intelligent control program for the toilet. The 3D-ToF sensor array deployed on the top of the waste tank performed a panoramic scan of the space inside the tank, accurately capturing the three-dimensional distribution of solid sediments. The multispectral imager installed on the side wall was activated simultaneously, using infrared spectral analysis to identify the urea crystallization area and the concentration of undecomposed organic matter on the waste surface in real time. Embedded temperature and humidity sensors continuously monitored the environmental conditions inside the tank. When the temperature exceeded 35°C and the humidity reached 90%, the system automatically marked it as a high-risk fermentation condition. All sensor data was transmitted to the edge computing gateway via industrial Ethernet. The heterogeneous computing chipset built into the gateway processed the data stream in parallel. The neural network processor performed semantic segmentation on the multispectral image and output a real-time solid-liquid-gas three-phase distribution thermal map. The matrix coprocessor fused three-dimensional point clouds and chemical features to construct a dynamic waste state tensor. The time-series prediction engine called up the most recent 30 operation records and used a long short-term memory network to predict the trend of waste fermentation activity index changes within the next 15 minutes.

[0028] After receiving the fused feature data, the dynamic decision engine generates a basic operating strategy based on the current state of the contaminants: when a solidified area is detected at the rear of the tank, the main control unit increases the flushing intensity to the rated value; for the liquid area with high ammonia nitrogen concentration in the middle, the secondary deodorization program is activated and the valve opening time is extended; when the prediction shows that the microbial activity will reach its peak within 10 minutes, the auxiliary stirring device is started in advance to prevent biogas accumulation. The decision parameters are then imported into the biomimetic optimizer. The algorithm initializes 50 candidate schemes and iterates multiple times by simulating the foraging behavior of ospreys. In the exploration phase, the parameter combinations are randomly adjusted and the cleaning efficiency is evaluated. In the development phase, the algorithm converges to the equilibrium point with the lowest energy consumption. The final output optimized parameters are flushing intensity 105%, valve opening 85%, and deodorization level 3.

[0029] The actuators respond in concert: the pneumatic control unit drives the arc gate to open and close precisely at 85% opening; the variable frequency water pump generates a high-pressure rotating water flow in the rear area of ​​the tank to impact the sludge; the plasma deodorization module starts the level 3 power mode in area 2 where the ammonia nitrogen concentration exceeds the standard; during the operation, the multispectral imager monitors the dissolution status of the sludge in real time; when the volume reduction of the sludge in the sludge-filled area is less than the expected value, the system automatically triggers the dynamic compensation mechanism to increase the flushing pressure and extend the operation time in that area; after the main operation cycle is completed, the pressure sensor detects a decrease in the vacuum at the front of the tank; the intelligent diagnostic module immediately determines that residual liquid sludge is blocking the pipeline, and then starts the reverse flushing program and pushes "pipeline maintenance suggestions" to the crew's handheld terminal.

[0030] Example 3: Implementation of the Intelligent Control System for Locomotive Toilet System in a Plateau Freight Locomotive When a freight locomotive enters a high-altitude, low-temperature region, the system automatically activates its cold-resistant mode. A 3D-ToF sensor deployed on the side wall of the waste bin initiates a high-frequency scanning mode to capture the three-dimensional accumulation morphology of high-viscosity solid waste in real time. Simultaneously activated, a multispectral imager analyzes the distribution of ice crystals and unfrozen organic matter on the waste surface using near-infrared spectroscopy. Embedded temperature and humidity sensors continuously monitor environmental parameters inside the bin. When the temperature is detected to be below -15℃ and humidity reaches 70%, the system automatically marks it as a high-risk freezing condition. All sensor data is transmitted to the edge computing gateway via a redundant CAN bus. The gateway's embedded heterogeneous processor group executes multiple tasks in parallel: a neural network acceleration unit performs ice-water-solid three-phase segmentation on the spectral image, generating a real-time phase distribution heatmap; a matrix coprocessor fuses point cloud data and temperature gradients to construct a dynamic waste viscosity model; and a time-series prediction engine calls historical high-altitude cold-condition data, using a long short-term memory network to predict the freezing rate of waste and the changing trend of microbial activity within the next 20 minutes.

[0031] After receiving the fused data, the dynamic decision engine generates a customized operation strategy: when a high-viscosity solid area is detected at the bottom of the tank, the main control unit starts the preheating mode and raises the flushing water temperature to 40°C through the electric heating film. For the area with excessive ammonia nitrogen concentration in the middle, the three-stage deodorization program is activated and the positive pressure flushing time is extended. When the prediction shows that an ice film will form on the surface of the dirt within 15 minutes, the ultrasonic vibration device is turned on in advance to prevent ice from forming. The decision parameters are imported into the bionic optimizer. The algorithm initializes 30 candidate schemes and iterates multiple times by simulating the hunting behavior of Arctic seals. In the exploration phase, the parameter combination is randomly adjusted and the ice-breaking efficiency is evaluated. In the development phase, it converges to the optimal balance point between energy consumption and antifreeze effect. The final output optimized parameters are flushing pressure 150%, valve opening 70%, and water temperature 38°C.

[0032] The actuator responds to the following commands: the pneumatic control unit drives the cold-resistant alloy gate to open and close precisely at 70% opening; the variable frequency water pump generates a high-pressure vortex water flow at the bottom of the tank to impact and freeze the dirt; the plasma deodorization module starts at level 4 power mode in zone B where ammonia nitrogen exceeds the standard; during operation, the multispectral imager monitors the ice melting status in real time; when the volume reduction of the frozen zone is less than expected, the system automatically triggers the dynamic compensation mechanism, increases the flushing pressure and injects antifreeze electrolyte; after completing the main operation cycle, the air pressure sensor detects an abnormal fluctuation in the vacuum degree at the top of the tank to 0.3 kPa; the intelligent diagnostic module immediately determines that the pipeline is partially blocked due to the condensation of gaseous dirt, and then starts the hot air drying program and pushes the "pipeline insulation layer check" alarm to the human-machine interface in the driver's cab.

[0033] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0034] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent control of a locomotive toilet, characterized in that: The method includes the following steps: S1. Collect internal state data and environmental parameters of the waste bin through a multi-source sensor array to generate a raw dataset of waste status. S2. Perform multimodal feature fusion processing on the original dataset of the dirt state to generate a dirt state fusion feature matrix; S3. Based on the improved YOLOv7 algorithm, target segmentation processing is performed on the image inside the sludge bin to generate a heat map of sludge distribution and liquid level contour data; S4. Construct a waste state evolution prediction model, combine historical operation data to conduct time series analysis on waste fermentation state, and generate a waste fermentation activity index. S5. Input the sludge state fusion feature matrix, sludge distribution heat map, liquid level contour data and sludge fermentation activity index into the dynamic decision engine to generate the optimal operation mode instruction set. S6. Based on the biological behavior simulation algorithm, adaptively optimize the operation mode instruction set to generate biomimetic optimized operation parameters; S7. The actuator control module drives the toilet valve, flushing pump and deodorization device to work together to complete the closed-loop control of the entire sewage treatment process.

2. The intelligent control method for locomotive toilets according to claim 1, characterized in that: S1 includes: S11. Collect three-dimensional point cloud data inside the waste bin using a 3D-ToF sensor array to generate a spatial distribution model of waste volume. S12. Use a multispectral imager to obtain the distribution map of chemical composition on the surface of the dirt and generate an ammonia nitrogen concentration gradient matrix. S13. Deploy a temperature and humidity composite sensor to monitor the microenvironment parameters inside the chamber in real time and generate an environmental state vector. ,in Indicates the temperature value. Indicates humidity value. This indicates the air pressure value.

3. The intelligent control method for locomotive toilets according to claim 1, characterized in that: S2 includes: S21. Establish the waste characteristic tensor ,in Represents 3D point cloud data. Let E represent the ammonia nitrogen concentration gradient matrix, and E represent the environmental state vector. S22. Perform cross-modal correlation analysis on feature tensors using a convolutional feature pyramid network to generate a fused feature matrix. in, For convolution operations, For feature fusion operators, This refers to the attention mechanism module; the meanings of the remaining characters are the same as above.

4. The intelligent control method for locomotive toilets according to claim 1, characterized in that: S3 includes: S31. Construct a dual-branch segmentation network: Spatial branching uses dilated convolution to extract dirt boundary features; Semantic branches parse the dirt material properties using a Transformer encoder; S32, Output Fouling Phase Distribution Diagram and liquid level height time series curve ; in, This is an area containing solid waste. This area contains liquid contaminants. This area contains gaseous contaminants. It is a function of liquid level height.

5. The intelligent control method for locomotive toilets according to claim 1, characterized in that: S4 includes: S41. Establish the equation for the evolution of the state of pollutants: Where A represents the fermentation activity index. Indicates the concentration of organic matter. , , The biochemical reaction coefficient, This is the temperature value. This is the humidity value; S42. Predicting the future using LSTM networks Activity index during the period .

6. The intelligent control method for locomotive toilets according to claim 1, characterized in that: S5 includes: S51. Define the job mode decision matrix: in, Parameters for controlling the intensity of the flushing water flow, Parameters for controlling the degree of valve opening, The power level of the deodorization device, These are the decision variables, corresponding to the specific values ​​of flushing intensity, valve opening, and deodorization level, respectively. S52. Generating decision weights based on a fuzzy rule base: in This is the sensitivity coefficient. For feature threshold, Let i be the i-th feature of the fused feature matrix.

7. The intelligent control method for locomotive toilets according to claim 1, characterized in that: S6 includes: S61. Initialize the osprey population and optimize its location. ; S62, Location Update During Exploration Phase: in It is a random number. , The current optimal position represents the optimal combination of decision parameters in the algorithm's history; S63. Development Phase Location Optimization: in The convergence factor is For the number of iterations, The location boundaries represent the upper and lower limits of the decision parameters.

8. The intelligent control method for locomotive toilets according to claim 1, characterized in that: S6 further includes: S64. Construct the fitness function: in, These are weighting coefficients, corresponding to the relative importance of energy consumption, operation time, and cleanliness, respectively. This is an energy consumption indicator, representing the energy consumed during rinsing and deodorization. This is a task time indicator, representing the time spent on a single task. As an indicator of cleanliness, it quantifies the effectiveness of dirt removal; S65, when satisfied Output optimal parameters at the time .

9. A smart control system for a locomotive toilet, implementing the method described in any one of claims 1-8, characterized in that: include: Multi-source sensing module, including 3D-ToF sensor array, multispectral imager and environmental sensor; The edge computing module, equipped with an FPGA chip, enables real-time feature fusion; The decision optimization module includes a dynamic decision engine and a biomimetic optimization processor; The execution control module drives the solenoid valve group, variable frequency water pump and plasma deodorization device.

10. The intelligent control system for locomotive toilets according to claim 9, characterized in that: The edge computing module adopts a heterogeneous computing architecture: Neural network acceleration units process image segmentation tasks; The matrix coprocessor performs F* feature fusion; The time series prediction unit runs the LSTM model independently; Each computing unit exchanges data via the NoC on-chip network.

Images