Research On Pipe Orifice Visual Positioning Methods For Power Plant Condenser Cleaning Robot

Large condenser is one of the key heater exchange equipments in heat-engineplant and petrochemical industry. For the unclear cooling water and chemical reactionsduring heat exchange when the condenser is running, the fouling which is notfavorable for heat transfer is accumulated in the inner wall of condenser tube. Thesefouling have produced such harm: severely decreases the heat exchange capacity ofthe condenser, lowers the vacuum degree, increases energy consumption and probablyleads to accidence due to the blockage of condenser tube and followed by erosion andperforation.Firstly, the necessity of fouling cleaning is explained by investigating the causeof fouling accumulation and the influence on heat transfer performance of condenser.The advantages and disadvantages of the ordinary fouling cleaning methods areanalyzed. The developments of the worldwide cleaning robots and their keytechnologies are also summarized.Secondly, a novel strategy of underwater visual positioning system of condensercleaning robot is taken into considering for solving the former tubes positioningproblems caused by coordinates input. In consideration, key matters in the process ofthis strategy implementation are analyzed，and the structure of underwater visualpositioning system is presented in detail.As the visual positioning mainly relies on the image information, the applicationsof image preprocessing technologies, edge detecting methods in nozzle positioningprocess are introduced in this paper. By analyzing of contrast enhancement and noisesuppression methods for pipe orifice images, a new pipe orifice image denoisingmethod based on neural network is proposed. In the methods, image denoising isdivided into two steps. Firstly, each noisy pixel is classified by the BP-net, and thenthe noisy pixels are filtered using a select multi-mold adaptaion noise filtering method.Experimental results show that the proposed image denoising method is better than themean noise suppression method and the median noise suppression method. Thismethod can greatly reduce noisy points in pipe orifice image, and can maintaineffectively the edges of the image. By analyzing of the existing edge detectionalgorithm, a new improved random Hough transform method for pipe orifice positionis proposed. This method can preferably calculate the edge and circle radius of pipe orifice, and this method can help the robot accurately position for pipe orifice.Finally, the main innovations of the thesis are summarized and the fields forfurther research are expected.