Intelligent cleaning method and system for train car body in port based on AI technology

Abstract

The application discloses a port train car side intelligent cleaning operation method and system based on an AI technology, and the method comprises the following steps: S1, acquiring scene environment parameters, accumulated material parameters and car side parameters; S2, extracting environment risk coefficients, accumulated material risk coefficients and cleaning difficulty coefficients; S3, building a comprehensive cleaning model and calculating a comprehensive cleaning index; S4, setting a cleaning threshold to evaluate the comprehensive cleaning index, determining a cleaning mode and cleaning parameters; and S5, collecting car side surface data before and after cleaning, and adjusting weight distribution of the comprehensive cleaning model according to a neural network learning technology until a minimum cleaning error is achieved. The method collects multi-dimensional parameter values such as environment parameters, car side surface parameters and accumulated material parameters in real time, extracts environment risk coefficients, accumulated material risk coefficients and cleaning difficulty coefficients, realizes accurate quantitative evaluation, improves the pertinence and accuracy of cleaning operation, and adaptively updates the comprehensive cleaning model based on the AI learning technology, and improves the system adaptability.

Description

Technical Field

[0001] The present invention relates to the field of intelligent cleaning technology, and specifically to an AI-based intelligent cleaning method and system for the sides of port trains. Background Art

[0002] With the rapid development of global trade and the continued growth of port logistics throughput, the cleaning efficiency of train loading bays, core facilities for bulk cargo transshipment, directly impacts the safety of loading and unloading operations and the service life of transportation equipment. Traditional cleaning methods rely on manual labor and are unable to cope with the complex and changing characteristics of accumulated materials, such as adhesion, particle size differences, and dynamic environmental factors such as temperature, humidity, and dust concentration. This leads to unstable cleaning results and waste of resources.

[0003] Currently, smart cleaning technologies primarily focus on automated cleaning using single sensor monitoring or fixed program control. For example, some systems use lidar or visual sensors to identify accumulated material or dirt, and then use preset programs to drive cleaning equipment for cleaning. Other systems utilize historical data storage and simple model analysis to optimize cleaning parameters. Consequently, the application of existing smart cleaning technologies to train side cleaning equipment still faces significant limitations. These limitations include a single perception dimension of the cleaning environment and insufficient comprehensive evaluation capabilities for multimodal data, making it difficult to comprehensively assess the risk of accumulated material, environmental interference, and cleaning difficulty. Cleaning decisions rely on empirical thresholds, failing to accurately quantify the combined impact of the environment and accumulated material. Cleaning models are simple and static, unable to dynamically adjust parameters based on real-time operating conditions. This can lead to over- or under-cleaning in complex operating conditions. Path planning algorithms fail to incorporate the risk distribution of accumulated material, making them unsuitable for loading bay cleaning equipment. This makes it difficult to prioritize areas at high risk of accumulated material, resulting in low cleaning efficiency. Furthermore, the lack of a closed-loop feedback mechanism prevents online iterative optimization of model weights, making the system difficult to adapt to long-term dynamic factors such as side material aging and changes in accumulated material properties.

[0004] Therefore, it is particularly necessary to develop an AI-based intelligent cleaning method and system for port train carriages to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and provide an AI-based intelligent cleaning method and system for the side of railway cars in port areas to solve the problems raised in the above-mentioned background technology.

[0006] In order to achieve the above object, the technical solution of the present invention is as follows:

[0007] The application discloses a port train car body intelligent cleaning operation method based on an AI technology, and comprises the following steps: S1, acquiring environment parameters, accumulated material parameters and car body parameters of a cleaning scene, identifying accumulated material types and train car body types, wherein the environment parameters comprise temperature, humidity and dust concentration, the accumulated material parameters comprise accumulated material granularity and adhesion, and the car body parameters comprise three-dimensional profile data, an inclination angle, a surface contact pressure and a real-time surface roughness of the car body; S2, preprocessing the environment parameters, the accumulated material parameters and the car body parameters, performing multi-factor grouping fusion and weight calculation, respectively establishing an environment risk model, an accumulated material risk model and a cleaning difficulty model, and generating an environment risk coefficient, an accumulated material risk coefficient and a cleaning difficulty coefficient; S3, establishing a comprehensive cleaning model, importing the environment risk coefficient, the accumulated material risk coefficient, the cleaning difficulty coefficient and the accumulated material granularity and distributing weights, and calculating a comprehensive cleaning index; S4, setting a cleaning threshold to evaluate the comprehensive cleaning index, determining a cleaning mode and cleaning parameters; and S5, collecting car body surface data before and after cleaning, calculating a cleaning error, setting a learning rate, adjusting weight distribution of each coefficient in the comprehensive cleaning model according to an error back propagation mechanism of a neural network, and stopping until a minimum cleaning error is obtained.

[0008] Further, the environment risk model outputs an environment risk coefficient , which comprehensively considers the influences of temperature, humidity and dust concentration:

[0009]

[0010] wherein, , , ; , , , , , are environment weight coefficients of each item, + + + + + =1; T is a real-time collected temperature value, is a temperature mean value, is a temperature standard deviation; is a real-time collected humidity value, is a humidity mean value, is a humidity standard deviation; is a real-time collected dust concentration value, is a dust concentration mean value, is a dust concentration standard deviation.

[0011] Further, the accumulated material risk model outputs an accumulated material risk coefficient , the influence of the deformation of the vehicle body, the pressure generated by the accumulated material on the surface of the vehicle body, and the adhesion between the accumulated material and the vehicle body are comprehensively considered:

[0012]

[0013] wherein, , and are the weight coefficients of each item of the vehicle body, and ; , S is the real-time collected three-dimensional profile data of the vehicle body, is the standard profile data of the vehicle body in the empty state, is the maximum deformation of the vehicle body material or the vehicle body structure; P is the real-time surface contact pressure of the vehicle body at the node, is the average value of the surface contact pressure of the vehicle body collected by the pressure sensor array, is the standard contact pressure threshold, is the standard deviation of the surface contact pressure; is the adhesion between the accumulated material and the surface of the vehicle body, is the reference adhesion.

[0014] Further, the cleaning difficulty model outputs a cleaning difficulty coefficient , which comprehensively considers the influence of the environment, the accumulated material, and the roughness of the surface of the vehicle body on the cleaning operation:

[0015]

[0016] wherein, is the real-time surface roughness of the vehicle body, is the reference roughness, is the maximum roughness of the surface of the vehicle body, is the inclination angle.

[0017] Further, the comprehensive cleaning model outputs a comprehensive cleaning index :

[0018]

[0019] wherein, , , and are the cleaning weight coefficients of each item, and + + + =1; is the maximum particle size within the particle size distribution of the accumulated material whose concentration ratio reaches 50%, is the maximum particle size within the particle size distribution of the accumulated material whose concentration ratio reaches 90%.

[0020] Further, the cleaning model is adaptively adjusted in weight based on the cleaning error, which includes the following steps:

[0021] 1) define the loss function: wherein, is the cleaning error, which is the difference between the residual material thickness after cleaning and the expected residual material thickness after cleaning;

[0022] 2) calculate the gradient of the loss function to the weights , , and ;

[0023] 3) set the learning rate, update the temporary weights in the opposite direction of the gradient by the gradient descent method, and dynamically gradient attenuate each weight with the iteration number to obtain the temporary weight under the minimum error value;

[0024] 4) normalize each temporary weight to obtain the final weight value and update the comprehensive cleaning model.

[0025] Further, step S4 further includes priority cleaning path planning based on the improved dijkstra algorithm, wherein the key nodes are assigned according to the wheel contour, the material risk coefficient of each key node is calculated, and the distance of the sub-path between adjacent key nodes is represented by the edge weight , and the calculation formula of the edge weight is:

[0026]

[0027] wherein, is the Euclidean distance of the sub-path between the key nodes, is the material risk coefficient of the visited key node, is the material risk threshold, is the super-threshold enhancement coefficient;

[0028] The material risk coefficient of the cleaning starting point is initialized to 0, and the cleaning path is planned from the cleaning starting point according to the path target and the node coverage requirement based on the edge weight .

[0029] An intelligent cleaning operation system includes a multi-modal data acquisition terminal, an intelligent processing terminal, a cloud database, an intelligent execution control terminal, and an automatic cleaning device; wherein,

[0030] The cloud database creates a historical database, a model parameter database, and a benchmark database by deploying a distributed relational database cluster;

[0031] The multi-modal data acquisition terminal uploads each parameter signal collected in real time to the historical database for storage;

[0032] The intelligent processing terminal respectively calls corresponding parameter values from the historical database, the model parameter database and the benchmark database, pre-processes and analyzes each parameter data, extracts the environmental risk coefficient Er, the accumulated material risk coefficient Ar and the cleaning difficulty coefficient Dc, and imports the comprehensive cleaning model to generate a comprehensive cleaning index; the intelligent processing terminal uploads each coefficient and the comprehensive cleaning index to the cloud database for storage, and simultaneously sends them to the intelligent execution control terminal;

[0033] The intelligent execution control terminal is electrically connected to the automatic cleaning device, which dynamically adjusts the cleaning mode and cleaning parameters according to the comprehensive cleaning index, plans a cleaning path, and sends control instructions and scheduling instructions to the automatic cleaning device, which executes cleaning work according to the instructions.

[0034] Further, the multi-modal data acquisition terminal includes a three-dimensional laser scanner, a pressure sensor array, an inclination sensor, a roughness sensor, a particle size detection sensor, an adhesion force sensor, a temperature and humidity sensor, and a dust concentration sensor.

[0035] Compared with the prior art, the port train car side intelligent cleaning operation method and system based on AI technology have the following beneficial effects:

[0036] 1. The cleaning operation method comprehensively considers multi-dimensional influencing factors, the system can collect environmental parameters, car side surface parameters and accumulated material parameters and other multi-dimensional parameter values in real time, pre-processes the collected multi-dimensional parameters by grouping, noise reduction filtering, and assigns weights for fusion, respectively extracts environmental risk coefficients, accumulated material risk coefficients and cleaning difficulty coefficients, realizes accurate quantitative evaluation of the cleaning operation scene, avoids the problems of strong subjectivity of traditional manual judgment and insufficient perception dimension of automatic equipment, and significantly improves the pertinence and accuracy of cleaning operation;

[0037] 2. By collecting car side parameters before and after cleaning, the cleaning error of the actual cleaning effect deviating from the expected cleaning effect is determined, the AI neural network learning technology is used to online iteratively optimize the comprehensive cleaning model through an adaptive weight updating method, and the benchmark parameter data is combined to support the system to adapt to different car side models, accumulated material types and environmental changes;

[0038] 3. The intelligent execution control terminal of the system dynamically adjusts the negative pressure and cleaning frequency of the cleaning device based on the comprehensive cleaning index, realizes accurate and effective cleaning, and allocates cleaning capacity in a targeted manner; based on the improved Dijkstra algorithm, a priority cleaning path is planned according to the accumulated material risk coefficient, high-risk accumulated material areas are preferentially processed, and "on-demand cleaning" and efficient operation are realized. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1The flow chart of the intelligent cleaning operation method disclosed in the present application;

[0040] Figure 2 The composition schematic diagram of the intelligent cleaning operation system disclosed in the present application. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present application will be clearly and completely described in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only the best embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0042] Reference to "embodiments" in this document means that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive or alternative to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0043] Train transportation has various types of materials, and the transportation car hopper needs to be effectively cleaned to avoid excessive cleaning or insufficient cleaning due to mismatched cleaning degree. Since the existing intelligent cleaning technology has single perception and cannot perceive complex environment, it lacks comprehensive evaluation ability of accumulated materials, and cannot be applied to accumulated material cleaning operation in complex environment of train car hopper. For example, for coal accumulated materials with large adhesion force in humid environment, it is difficult to effectively remove the adhered accumulated materials. For sandy materials prone to dust in dry and high temperature environment, insufficient suction force leads to serious dust. Therefore, the present example embodiment provides a port train car hopper intelligent cleaning operation method based on AI technology, which comprehensively evaluates the cleaning index by fusing multi-modal perception data, calls the corresponding cleaning mode for targeted cleaning, and improves the cleaning quality. Specifically, as shown in the figure, it includes the following steps: Figure 1

[0044] S1, obtaining the environment parameters, accumulated material parameters and car hopper parameters of the cleaning scene through multi-source perception elements, the environment parameters including temperature, humidity and dust concentration, the accumulated material parameters including particle size and adhesion force of the accumulated material, and the car hopper parameters including three-dimensional profile data, inclination angle, surface contact pressure and real-time surface roughness of the car hopper;

[0045] ​The above real-time collected multi-source parameter data is uploaded to a cloud database for storage, and based on the real-time data, type data of the accumulated material and different models and materials of the train car body are recognized, and corresponding reference data is called out, the environmental reference data including mean value and standard deviation of temperature, humidity and dust concentration are calculated by long-term collection of environmental data of the train car body in an empty state, and the reference data of the train car body includes reference value of roughness of the train car body surface in an ideal state and standard contour data of the empty train car body, wherein the reference value of roughness is measured by a laser roughness measuring instrument at three representative points on the surface of the new train car body which is not damaged and polluted, such as the middle, edge and corner of the train car body, and the reference roughness and standard contour data are bound with the information of different models and materials of the train car body and stored in the reference database; the reference adhesion between different accumulated materials and different train car bodies is obtained by adhesion test, the material of the adhesion test device is consistent with the material and roughness of the train car body, in a standard environment, the contact state of the train car body and the accumulated material is simulated, a standard pressure is applied and maintained for a certain period of time, the natural deposition process of the accumulated material on the surface of the train car body is simulated, and then the adhesion test device is vertically pulled at a uniform speed, the maximum pulling force value at the moment of separation is recorded, the average value of 5 repeated measurements is taken, and the average value is the reference adhesion, finally the reference adhesion is bound and stored with the information of the train car body model, measurement point and accumulated material type;

[0046] In addition, the surface contact pressure of the train car body includes the surface contact pressure of the measurement point and the average surface contact pressure , the average surface contact pressure is obtained by averaging the surface contact pressure collected by the pressure sensor array of all measurement points, and the standard deviation of the surface contact pressure is calculated , the pressure sensor array is uniformly installed on the inner surface of the train car body, and the installation density is not less than 10 / ㎡;

[0047] S2, call corresponding parameter reference value, pretreat environment parameter, accumulated material parameter and train car body parameter, and carry out multi-factor grouping fusion and weight calculation, respectively build environment risk model, accumulated material risk model and cleaning difficulty model, generate environment risk coefficient, accumulated material risk coefficient and cleaning difficulty coefficient; specifically,

[0048] (1) considering the influence of environment on cleaning degree, introducing environment risk coefficient for evaluation, fusing environmental temperature, humidity and dust concentration to construct environment risk model and calculate environment risk coefficient :

[0049]

[0050] Among them, , , ; , , 、 、 、 is the environmental weight coefficient of each environmental factor, which is in , + + + + + =1; T is the real-time collected temperature value, is the temperature mean value, is the standard deviation of temperature; is the real-time collected humidity value, is the humidity mean value, is the standard deviation of humidity; is the real-time collected dust concentration value, is the concentration mean value, is the standard deviation of concentration;

[0051] (2) The deformation of the car body, the pressure generated on the surface of the car body by the accumulated material, and the adhesion between the accumulated material and the car body are the key factors affecting the material deposition and the generation of accumulated material. The accumulated material risk model is constructed by integrating the three factors, and the accumulated material risk coefficient is calculated:

[0052]

[0053] wherein, , and are the car body weight coefficients of each factor, which are in , and ; the accumulated material thickness , S is the real-time collected three-dimensional profile data of the car body, is the standard profile data of the car body in the empty state, is the maximum deformation of the car body material or structure; P is the real-time surface contact pressure of the car body at the node, is the average value of the car body surface contact pressure collected by the pressure sensor array, is the standard contact pressure threshold, is the standard deviation of the surface contact pressure; is the adhesion between the accumulated material and the car body surface, is the reference adhesion;

[0054] (3) The cleaning difficulty needs to consider the surface state of the car body, combine the roughness and inclination, and integrate the environmental influence and the accumulated material risk to construct the cleaning difficulty model for comprehensive evaluation of the accumulated material cleaning difficulty, and output the cleaning difficulty coefficient :

[0055]

[0056] wherein, is the real-time surface roughness of the carbody, is the reference roughness, is the maximum roughness of the carbody surface, is the carbody inclination angle.

[0057] S3, a comprehensive cleaning model is built, environmental risk coefficient, accumulated material risk coefficient, cleaning difficulty coefficient and accumulated material particle size are imported and weighted, and a comprehensive cleaning index is calculated:

[0058] The comprehensive cleaning model outputs a comprehensive cleaning index :

[0059]

[0060] wherein, , , and are the cleaning weight coefficients of each influencing factor, all in , and + + + =1; is the maximum particle size in the accumulated material particle size distribution with a concentration ratio of 50%, is the maximum particle size in the accumulated material particle size distribution with a concentration ratio of 90%.

[0061] S4, according to the preset cleaning threshold, the size relationship between the comprehensive cleaning index and the preset threshold is compared, and the corresponding cleaning mode and cleaning parameter are determined, such as using the standard mode when it is lower than the preset threshold, setting the vibration frequency as the standard frequency, and the negative pressure as H , H is a negative pressure factor, a preset constant; when it exceeds the preset threshold, the high-frequency vibration mode is triggered, the negative pressure is H , the vibration frequency is f+v , f is a basic frequency, and v is a frequency factor.

[0062] According to the accumulated material risk coefficients of each key node on the carbody profile, the priority cleaning path of the cleaning operation is planned, the key nodes with high accumulated material risk are preferentially processed, the logic of "the higher the risk, the lower the weight, and the higher the priority" is adopted, the accumulated material risk coefficient is introduced, the priority cleaning path is planned based on the improved dijkstra algorithm, the shortest path from the cleaning starting point to other key nodes is calculated, the accumulated material risk coefficient of the cleaning starting point (the first key node) is initialized to 0, the accumulated material risk coefficients of other key nodes are calculated respectively, the path distance between adjacent key nodes of the planned path is represented by edge weight, the edge weight between key nodes that cannot be planned is infinite, and the edge weight , is the Euclidean distance between the key nodes, is the risk coefficient of accessing the key node, and the shortest path from the cleaning starting point to access all other key nodes is calculated;

[0063] For example: assuming that the wheel profile has five key nodes A (0, 0, 0), B (2, 0, 0), C (3, 2, 0), D (1, 3, 0), and E (0, 2, 0), and the preset risk threshold of the accumulated material is 0.5, where Ar of A is 0.3, Ar of B is 0.6, Ar of C is 0.4, Ar of D is 0.7, and Ar of E is 0.4, the path distance between adjacent key nodes is defined by the edge weight W, and the edge weight W is calculated as follows:

[0064]

[0065] wherein, is the Euclidean distance between the key nodes, is the risk coefficient of accessing the key node, is the risk threshold of the accumulated material, is the super-threshold enhancement coefficient;

[0066] The risk coefficient of the cleaning starting point is initialized to 0, and according to the known data, the Euclidean distance of A-B is d A-B = 2, the edge weight W A-B of accessing B is 0.88; the Euclidean distance of A-E is d A-E ≈ 2, the edge weight W A-E of accessing E is 1.2; the Euclidean distance of E-D is d E-D ≈ 1.414, the edge weight W E-D of accessing D from E is 0.509; the Euclidean distance of B-C is d B-C ≈ 2.236, the edge weight W B-C of accessing C from B is 1.342, the Euclidean distance of D-C is d D-C ≈ 2.236, the edge weight W D-C of accessing C is 1.342;

[0067] If the path target is set to a single continuous path and covers all nodes, the nearest unvisited node to the cleaning starting point is selected each time, the shortest path of its adjacent nodes is updated according to the edge weight, and the process is repeated until all nodes are visited. The improved Dijkstra algorithm path planning process is as follows:

[0068] The cleaning starting point A priority queue is initialized: {A: 0}, indicating that the weight of the cleaning starting point A is 0; ​​

[0069] First step: take out A, expand neighbor B weight as 0.88 and E weight as 1.2, queue update as {B: 0.8, E: 1.2};

[0070] Second step: preferentially take out the minimum weight B, mark as "processed", expand neighbor C weight as 0.88+1.342=2.222, queue update as {E: 1.2, C: 2.222};

[0071] Third step: take out E, expand neighbor D weight as 1.2+0.509=1.709, queue update as {D: 1.709, C: 2.222};

[0072] Fourth step: take out D, mark as "processed", expand neighbor C weight as 1.709+1.342=3.054, queue update as {C: 2.222};

[0073] In the above path planning process, high-risk area B is processed earlier than D, which embodies the logic of "the higher the risk, the lower the weight, and the priority of access". If the goal is to "cover all high-risk areas", the shortest path and its processing nodes are A-B-E-D-C, and the priority cleaning path planning processes nodes B and D preferentially;

[0074] If the goal is set as an efficient path and covers high-risk nodes, the efficient path to access high-risk nodes B and D is to execute path A-E-D, and process A-B separately. This place will not be elaborated on this planning process;

[0075] If the goal is set as a single continuous path and only covers high-risk nodes, the optimal path is A-B-C-D, and this place will not be elaborated on this planning process.

[0076] S5, generate the three-dimensional point cloud data of the car side surface before and after cleaning by scanning with a three-dimensional laser scanner, and determine the residual material thickness of each key node of the car side surface , then according to the expected cleaning material thickness , calculate the cleaning error :

[0077] = ;

[0078] Based on the AI neural network deep learning technology, the comprehensive cleaning model is updated to reduce the cleaning error through the error back propagation mechanism. Specifically, it includes the following steps:

[0079] 1) Define the loss function of the error: ;

[0080] 2) Calculate the loss function of the cleaning weight coefficient , , and gradient of:

[0081]

[0082] wherein, ;

[0083] Similarly:

[0084]

[0085] 3) If > , indicating insufficient cleaning, then is a positive value, the gradient is positive, a step size of 0.01 is used, i.e. the learning rate is 0.01, the cleaning weight coefficient is decreased along the negative gradient, i.e. the weight value is reduced, and each cleaning weight coefficient is updated iteratively;

[0086]

[0087] wherein, is a temporary weight, is a learning rate, is an initial weight value or a weight value of the last gradient;

[0088] Similarly, the other temporary weights are updated in the same way, and each weight is dynamically gradient-decayed with the number of iterations, while is constantly reduced until it approaches a minimum value close to 0, obtaining the temporary weight under the minimum error value; the last updated temporary weight values are normalized to obtain the final weight value and update the comprehensive cleaning model;

[0089] Understandably, if < , indicating over-cleaning, then is a negative value, the gradient is negative, a step size of 0.01 is used, the weight is decreased along the positive gradient, i.e. the weight value is increased, and each cleaning weight coefficient is updated iteratively.

[0090] The present example embodiment also provides an intelligent cleaning operation system, as shown in Figure 2 based on the operation method of the above embodiment, which includes:

[0091] Multi-modal data acquisition terminal for real-time acquisition of environmental parameters, car parameters and material parameters, which includes a three-dimensional laser scanner, a pressure sensor array, an inclination sensor, a roughness sensor, a particle size detection sensor, an adhesion sensor, a temperature and humidity sensor and a dust concentration sensor, wherein the three-dimensional laser scanner is installed on the top or both sides of the loading building, and the three-dimensional profile data S(x, y, z) of the car is collected by emitting laser, the pressure sensor array is used to directly collect the contact pressure distribution of the material on the car during the loading process, which is uniformly installed on the inner surface of the train car, the installation density is ≥10 / ㎡, the inclination sensor is at least 2, covering the main inclined area, and the average inclination θ of the car is collected in real time, which is at least installed at the center position of the side plate and the bottom plate of the car, the roughness sensor collects the current car surface roughness R a in real time; the adhesion sensor collects the adhesion force F a of the current material and the car surface at each key node in real time, the particle size detection sensor collects the particle size value and its distribution, and the temperature and humidity sensor and the dust concentration sensor collect the environmental parameters of the scene in real time.

[0092] Cloud database, the cloud database is deployed by distributed relational database cluster, respectively creates historical database, model parameter database and benchmark database; the historical database stores the time series data of car parameters, material parameters and environmental parameters uploaded by multi-modal data acquisition terminal and the type data of different models, materials and material types of car, which adopts time slicing storage strategy for historical data and implements data life cycle management; the model parameter database stores the weight coefficient and version iteration record of the comprehensive cleaning model; the benchmark database stores the benchmark statistical value of environmental temperature, humidity and dust concentration, material benchmark parameter, empty car standard three-dimensional profile and material benchmark parameter;

[0093] The intelligent processing terminal retrieves real-time car parameters, material parameters, environmental parameters and pressure parameters, denoising and filtering pretreatment, combines with the benchmark parameters, calculates the environmental risk coefficient E r , the material risk coefficient A r and the cleaning difficulty coefficient D c , and imports the above coefficients into the comprehensive cleaning model to calculate and generate the comprehensive cleaning index I c , and uploads the result data to the cloud database for storage;

[0094] The intelligent execution control terminal is connected with the intelligent processing terminal, dynamically adjusts the cleaning mode and cleaning parameters of the automatic cleaning device according to the comprehensive cleaning index I c sent by the intelligent processing terminal, such as negative pressure and vibration frequency, and adjusts the cleaning mode and cleaning parameters of the automatic cleaning device according to the car three-dimensional profile and the material risk coefficient A rDistribution, by improving Dijkstra algorithm planning priority cleaning path, priority scheduling automatic cleaning device to high A r Accumulation area to clean up the work.

[0095] Those skilled in the art can understand that all or part of the step processes of the above-mentioned embodiment methods can be completed by instructing related hardware through a software program, which can be applied to related devices and equipment and can be stored in a computer readable storage medium.

[0096] Although embodiments of the present application have been shown and described, it is to be understood that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AI-based intelligent cleaning method for railway carriages in port areas, characterized in that: The following steps are involved: S1: Acquire environmental parameters, material accumulation parameters, and train side parameters of the cleaning scene, and identify the material accumulation type and train side type. The environmental parameters include temperature, humidity, and dust concentration. The material accumulation parameters include the particle size and adhesion of the material accumulation. The train side parameters include the three-dimensional profile data, inclination angle, surface contact pressure, and real-time surface roughness of the train side. S2: Preprocessing the environmental parameters, the material accumulation parameters, and the vehicle side parameters, performing multi-factor grouping fusion and weight calculation, respectively building an environmental risk model, a material accumulation risk model, and a cleaning difficulty model, and generating an environmental risk coefficient, a material accumulation risk coefficient, and a cleaning difficulty coefficient; S3: Building a comprehensive cleaning model, importing the environmental risk factor, the material accumulation risk factor, the cleaning difficulty coefficient, and the material accumulation particle size, assigning weights, and calculating a comprehensive cleaning index; S4: Setting a cleaning threshold to evaluate the comprehensive cleaning index, and determining the cleaning mode and cleaning parameters; S5: Collecting vehicle side surface data before and after cleaning, calculating the cleaning error, setting a learning rate, and adjusting the weight distribution of each coefficient in the comprehensive cleaning model according to the error back propagation mechanism of the neural network until the minimum cleaning error is achieved.

2. The AI-based intelligent cleaning method for railway carriages in port areas according to claim 1 is characterized by: The environmental risk model outputs the environmental risk coefficient , considering the influence of the temperature, humidity and dust concentration: ； in, 、 、 ; 、 、 、 、 、 are the environmental weight coefficients of each item, + + + + + =1; T is the temperature value collected in real time, is the mean temperature, is the standard deviation of temperature; is the humidity value collected in real time. is the mean humidity, is the standard deviation of humidity; is the real-time dust concentration value, is the mean dust concentration, is the standard deviation of dust concentration.

3. The AI-based intelligent cleaning method for railway carriages in port areas according to claim 2 is characterized by: The accumulation risk model outputs the accumulation risk coefficient , taking into account the influence of the deformation of the side of the car, the pressure generated by the accumulated material on the surface of the side of the car, and the adhesion between the accumulated material and the side of the car: ； in, 、 and are the weight coefficients of each item, and ; , S is the three-dimensional profile data collected in real time, It is the standard profile data of the side of the vehicle in the no-load state. is the maximum deformation of the side material or side structure; P is the real-time surface contact pressure of the side at the node, is the average value of the surface contact pressure collected by the pressure sensor array, is the standard contact pressure threshold, is the standard deviation of surface contact pressure; is the adhesion between the accumulated material and the surface of the vehicle side, is the baseline adhesion.

4. The AI-based intelligent cleaning method for railway carriages in port areas according to claim 3 is characterized by: The cleaning difficulty model outputs the cleaning difficulty coefficient , considering the impact of environmental influence, material accumulation and vehicle side surface roughness on cleaning operations: ； in, is the real-time surface roughness of the vehicle side, is the baseline roughness, is the maximum roughness of the car side surface, is the inclination angle.

5. The AI-based intelligent cleaning method for railway carriages in port areas according to claim 4 is characterized by: The comprehensive cleaning model outputs the comprehensive cleaning index : ； in, 、 、 and is the cleaning weight coefficient of each item, and + + + =1; The maximum particle size in the particle size distribution of the accumulated material with a concentration of 50% is It is the maximum particle size within the particle size distribution of the accumulated material with a concentration of 90%.

6. The AI-based intelligent cleaning method for railway carriages in port areas according to claim 5 is characterized by: The weight of the cleaning model is adaptively adjusted based on the cleaning error, which includes the following steps: 1) Define the loss function: ,in, The cleaning error is the difference between the thickness of the residual material after cleaning and the expected thickness of the material after cleaning; 2) Calculate the loss function for weights 、 、 and gradient; 3) setting the learning rate, and updating each temporary weight in the opposite direction of the gradient by the gradient descent method, wherein each weight is dynamically gradient-decayed with the number of iterations, and obtaining the temporary weight under the minimum error value; 4) Normalizing each temporary weight to obtain a final weight value and updating the comprehensive cleaning model.

7. The AI-based intelligent cleaning method for railway carriages in port areas according to claim 6 is characterized by: Step S4 also includes cleaning path planning based on the improved Dijkstra algorithm, in which key nodes are allocated according to the side contour, the accumulation risk coefficient of each key node is calculated, and the distance between adjacent key nodes can be planned as a sub-path with edge weights. Indicates that the edge weight The calculation formula is: ； in, is the Euclidean distance of the subpath between the key nodes, is the material accumulation risk coefficient for accessing key nodes, is the accumulation risk threshold, is the superthreshold enhancement coefficient; The accumulation risk coefficient at the cleaning starting point is initialized to 0, based on the edge weight , the cleaning path is planned from the cleaning starting point according to the path target and node coverage requirements.

8. An intelligent cleaning system, characterized by: It includes multimodal data acquisition terminal, intelligent processing terminal, cloud database, intelligent execution control terminal and automatic cleaning device; among them, The cloud database creates a history database, a model parameter database and a benchmark database by deploying a distributed relational database cluster; The multimodal data acquisition terminal uploads the parameter signals collected in real time to the historical database for storage; The intelligent processing terminal retrieves corresponding parameter values ​​from the historical database, the model parameter database, and the benchmark database, performs preprocessing analysis on each parameter data, extracts the environmental risk coefficient Er, the accumulation risk coefficient Ar, and the cleaning difficulty coefficient Dc, and imports them into the comprehensive cleaning model to generate a comprehensive cleaning index; the intelligent processing terminal uploads each coefficient and the comprehensive cleaning index to the cloud database for storage, and simultaneously sends them to the intelligent execution control terminal; The intelligent execution control terminal is electrically connected to the automatic cleaning device. It dynamically adjusts the cleaning mode and cleaning parameters according to the comprehensive cleaning index, plans the cleaning path, and sends control instructions and scheduling instructions to the automatic cleaning device. The automatic cleaning device performs the cleaning operation according to the instructions.

9. The intelligent cleaning system according to claim 8, characterized in that: The multimodal data acquisition terminal includes a three-dimensional laser scanner, a pressure sensor array, an inclination sensor, a roughness sensor, a particle size detection sensor, an adhesion sensor, a temperature and humidity sensor, and a dust concentration sensor.

Images