| The currrent situation of air pollution from ships in China's ports is serious.By the end of 2019,China's inland,coastal and ocean-going transport had handled 7.472 billion tons of cargo and handled 10.396304 billion ton-kilometers of cargo turnover,according to data released by the Ministry of Transport.The "White Paper on Air Pol-lution Prevention and Control of China's Ships and Ports" shows that the sulfur content of marine fuel oil is 100 to 3,500 times that of automotive diesel.In order to compre-hensively control the air pollution from ships in ports,China has established a ship air pollutant emission control area and implemented the "Marine Fuel Sulfur Limit Order",requiring all ships to not use marine fuel with a sulfur content greater than 0.1%in the inland river emission control area.International ships are not allowed to use marine fuel oil with a sulfur content greater than 0.5%in all domestic waters.This has brought great challenges to the maritime department to implement regulatory responsibilities.The traditional supervision method of boarding ships to extract oil samples for third-party inspection has large randomness and narrow coverage.However,the method of pre-screening vessels suspected of using excessive fuel oil by using exhaust ship mon-itoring technology can improve the supervision efficiency and be gradually applied in maritime supervision.The atmospheric pollution measurement lidar based on the principle of differential absorption is one of the effective means to monitor the emission of pollution sources from ships.It has the advantages of a wider monitoring range and high accuracy,and is not restricted by wind conditions.However,it also has the problem of not being able to quickly capture and accurately track the remote mobile emission sources.These two shortcomings limit the popularization and popularization of lidar,and also put for-ward the requirements for further improving the tracking control technology for lidar.To be specific,the current lidar for atmospheric pollution measurement has the follow-ing technical problems.Firstly,the existing working mode of lidar can not meet the requirements of fast positioning of smoke outlet.Secondly,small disturbances in the external environment will produce large deviations at the target.Thirdly,the existing controllers are difficult to meet the needs of fast target tracking in actual lidar systems.Fourthly,the saturation characteristic of the actuator weakens the control performance and threatens the stability of the system.Therefore,in view of the above problems,this dissertation comprehensively con-siders two aspects of smoke outlet detection and tracking controller design of ship mov-ing pollution sources,and carries out research on the tracking control problem of air pollution measurement lidar for remote mobile pollution sources monitoring.The main research contents are as follows.(1)Aiming at the problem of data drift between the actual port image and the train-ing set image affected by the variable environment,a robust unilateral alignment domain adaptation-based port ship image exhaust smoke vent detection method is proposed.By combining the classical deep learning object detetection network with the feature-based domain adaptive method in transfer learning,a robust unilateral alignment method is proposed to suppress the data distribution discrepancy between the target domain and source domain,on the basis of generating candidate boxes by RPN network and extract-ing the features of candidate regions by convolutional neural network.The features of two domains with data drift are mapped to the same distributed high-dimensional space by the same mapping function.It can improve the classification performance of the proposed mathed for candidate regions,solve the data drift problem between the actual scene images and the training set images,and enhance the adaptability of the proposed object detection method to images in different scenes.(2)To solve the problem that the external environment and the uncertainty of the lidar affect the tracking and monitoring accuracy,an integral sliding mode tracking control method based on the disturbance compensation of the time delay observer is proposed.Firstly,considering the robustness of integral sliding mode control to pa-rameter changes and the characteristics of suppressing sliding mode chattering,an anti-disturbance tracking control algorithm based on integral sliding mode is proposed.On this basis,considering the problem that the upper bound of the lumped disturbance of environmental disturbance and model uncertainty is difficult to be accurately known,according to the related theory of time delay control,a time delay observer is designed.The robustness of integral sliding mode rejection controller to continuous disturbance signals and uncertain parameters is increased.The simulation results show that the method can effectively compensate for the environmental disturbance and the uncer-tainty of the lidar model,and achieve accurate tracking control effects.(3)In order to solve the problem that it is difficult for the existing control algo-rithms to achieve fast tracking in the presence of disturbances,this dissertation proposes a self-defined convergence time fast tracking control algorithm based on a non-singular fixed-time terminal sliding mode.Firstly,a novel fixed time terminal sliding mode sur-face is designed,which can realize that the upper bound of the convergence time of sliding mode phase is a constant dependent on only one parameter.Then,a nonsingular fixed time terminal sliding mode control law is designed by introducing the fixed-time reaching law,so that the total convergence time of reaching phase and sliding mode phase has an upper bound that can be designed independently,so that the control target of the lidar scanning mechanism can quickly track the given rotation trajectory.(4)To deal with the actuator input saturation problem of lidar scanning mecha-nism in the presence of uncertain interference,an input saturation compensation track-ing control algorithm based on the auxiliary system and extreme learning machine is proposed.Firstly,an auxiliary system is introduced to compensate the saturation of the actuator.On the basic of the auxiliary system,the saturation compensation control method is proposed.Then,considering the influence of system lumped disturbance on tracking control,the algorithm of Gaussian kernel function based ELM is introduced,and a tracking control method of lidar input saturation compensation based on the aux-iliary system and extreme learning machine is designed.On the one hand,the lumped disturbance of the system can be estimated quickly by G-ELM,which is the explicit parameter learning method.And on the other hand,the saturation characteristics of the actuator are compensated to achieve the accurate tracking of the system in the presence of disturbance and saturation characteristics.To sum up,this dissertation builds a tracking control system of lidar with" finding emission sources according to images and fast and accurate tracking by controller",which provides theoretical support for the tracking and monitoring of remote mobile pollution sources and early technical guarantee for the realization and popularization of differential absorption lidar technology. |
PDF Full Text Request