Unmanned vessel-based maintenance decision-making method for diagnosing sedimentation in drainage pipelines and providing early warning for waterlogging

An unmanned vessel characterized in that the unmanned vessel is consisted of an integrated vessel body inside a pipeline and a manhole box;

• the integrated vessel body inside the pipeline comprises a Doppler ultrasound module, a three-axis gyroscope, high-definition infrared cameras, a main controller, a deep learning development board, baffles, a cabin cover, a cabin, fins, towing rings, and a navigation interface;
• the Doppler ultrasound module, the three-axis gyroscope, the high-definition infrared cameras, the navigation interface, and the deep learning development board are each connected to the main controller;
• the fins, the towing rings, the cabin cover, and the baffles are connected to the cabin;
• the main controller is set on the cabin, and the deep learning development board is loaded with a sedimentation diagnosis model for drainage pipelines, a waterlogging warning model, and an optimized maintenance decision-making model based on multi-objective optimization and waterlogging damage assessment methods;
• the manhole box comprises a floating towing cable, lithium batteries, a non-contact liquid level gauge for pipelines, a wireless communication module, whip-like transmitting antennas, a box body, a box cover, and fixing claws;
• the floating towing cable, the lithium batteries, the wireless communication module, the non-contact liquid level gauge for pipelines, and the whip-like transmitting antennas are all fixed inside the box body;
• the whip-like transmitting antennas are connected to the wireless communication module;
• the fixing claws and the box cover are connected to the box body;
• the non-contact liquid level gauge for pipelines, the lithium batteries, and the wireless communication module which is connected to an external server) are connected to the floating towing cable linked with the navigation interface;
• wherein, a deep learning algorithm in the drainage pipeline sedimentation disease diagnosis model is consisted of Gramian Angular Summation Field, convolutional neural networks, bidirectional long short-term memory neural networks, and attention mechanisms;
• physics knowledge-guided deep learning algorithms are embedded in the pipeline sedimentation disease diagnosis model, and the physics knowledge is extracted from a hydraulic models of changes in drainage pipeline inlet and outlet flow at different levels of sedimentation;
• wherein, the optimized maintenance decision-making model based on multi-objective optimization and waterlogging damage assessment methods comprises:
• screening infrared images of drainage pipeline inlets and outlets captured by the unmanned vessel under the criteria for screening images, which require that the internal walls on both sides horizontally in the drainage pipeline are displayed within the image frame, with the interface between air and water clear;
• the images comprises light intensity, noise intensity, and factors causing missing points in the capture, and must be taken from multiple shooting angles;
• by using image annotation tools to annotate the screened infrared images, obtaining infrared images of drainage pipeline inlets and outlets with annotated interfaces between air and water;
• based on the infrared images of drainage pipeline inlets and outlets with annotated interfaces between air and water, combining pre-measured Doppler ultrasound data and data about the inlet and outlet flow and sedimentation of drainage pipelines, and constructing training, validation, and test sets for the pipeline sedimentation diagnosis model;
• and
• training the drainage pipeline sedimentation diagnosis model by using its training, validation, and test sets, wherein tuning hyperparameters involve decoupled weight decay magnitude, batch size, maximum iterations, learning rates, and momentum coefficients;
• after completion of training, obtaining the optimal drainage pipeline sedimentation diagnosis model and outputting the diagnosis results of sedimentation in the drainage pipelines in a study area;
• wherein, a deep learning algorithm in the drainage pipeline sedimentation disease diagnosis model is consisted of Gramian Angular Summation Field, convolutional neural networks, bidirectional long short-term memory neural networks, and attention mechanisms;
• physics knowledge-guided deep learning algorithms are embedded in the pipeline sedimentation disease diagnosis model, and the physics knowledge is extracted from a hydraulic models of changes in drainage pipeline inlet and outlet flow at different levels of sedimentation;
• based on diagnosis results of drainage pipeline sedimentation diseases in a study area output by a drainage pipeline sedimentation disease diagnosis model, building a rainfall-runoff numerical model of drainage pipeline sedimentation to obtain simulated data about waterlogging;
• constructing training, validation, and test sets of the waterlogging warning model by using the diagnosis results of drainage pipeline sedimentation diseases in a study area output by a drainage pipeline sedimentation disease diagnosis model and the simulated data about waterlogging;
• by training the waterlogging warning model with the training, validation, and test sets, obtaining an optimal waterlogging warning model;
• wherein, a deep learning algorithm in the waterlogging warning model is consisted of one-dimensional and two-dimensional convolutional neural networks, long short-term memory networks, and attention mechanisms, physics knowledge-guided deep learning algorithms are embedded in the waterlogging warning model, and the physical knowledge is extracted from a hydraulic model of changes in drainage pipeline inlet and outlet flow at different levels of sedimentation and improved coupled one-dimensional and two-dimensional urban surface and underground connection models;
• by using the outputs of the optimal drainage pipeline sedimentation disease diagnosis model and the optimal waterlogging warning model, building an optimal pipeline maintenance decision-making model based on multi-objective optimization and waterlogging damage assessment methods consisted of triangular type-2 fuzzy sets and dynamic proportional substitution methods;
• when deploying the unmanned surface vehicles at inlets and outlets of drainage pipelines in a study area, collecting data and uploading an output of the drainage pipeline sedimentation disease diagnosis model to the external server by each of the unmanned surface vehicle deployed at the inlet or outlet of drainage pipelines, and generating a pipeline sedimentation database by the external server from the output of the drainage pipeline sedimentation disease diagnosis model of each of the unmanned surface vehicle deployed at the inlet or outlet of drainage pipelines in a study area;
• and
• by selected unmanned surface vehicles deployed at the inlets and outlets of drainage pipelines closest to the server, from the generated pipeline sedimentation database as an input, outputting waterlogging warning results and optimal decisions respectively by using the optimal waterlogging warning model and the optimal pipeline maintenance decision-making model based on multi-objective optimization and waterlogging damage assessment methods.