The Fuzzy Recognition System Of Rough Diesel Engine Cylinder Head Based On Neural Network

Currently the eligibility test of diesel engine cylinder head can only be achieved by airflow test after machining. If test failed, the cylinder head need to recast recycled, resulting great losses in time and cost. For this problem we designed the fuzzy recognition system of rough diesel engine cylinder head based on neural network. Determine the eligibility of the cylinder head before machining. System includes two modules. The first module is the three-dimensional data acquisition module. Use the program of line structured light to get three dimensional data on the airway edge points of the cylinder head. Then fit the point cloud to calculate the morphology information of the airway. The second module is the neural network module. Create a neural network and train it by three-dimensional data to fuzzy recognize the rough cylinder head and Judge the Eligibility. This study mainly includes the following sides:1. Build the hardware environment which required by three-dimensional data acquisition modules. Focus on solving issues of the selection, display, communication and connected with each other during the various components, and joined the SCM-mediated light control module.2. Create the software environment of system, use MFC single document mode as main framework and add OPENCV visual library. Write the driver and adjust operation interface system required according to the function of the camera and motor controller provided. An image algorithm applicable to the system is proposed according to the Contradiction problem of several different displays and handling mode of image. Design the specific operational procedures of calibration and three-dimensional scanning. Make a reasonable choose to the parameter setting and operation mode of each module according to the algorithm.3. Establish the neural network module. Put the data which calculated from three-dimensional data acquisition module as input of neural network. Put the eligibility of the cylinder head as output of neural network. Choose the appropriate algorithm and set the network parameters according to the system requirements. Finally, train and test the network.