Research And Application Of Deep Learning In The Field Of Thermal Pipeline Engineering Design

A popular application area of artificial intelligence in engineering design is the analysis and processing of engineering design drawings,which involves many professional technologies such as image processing,machine vision,OCR optical character recognition and neural networks.This thesis summarizes the research progress,basic research framework and latest achievements of artificial intelligence technology,especially target detection,and designs CAD drawing recognition and analysis system based on convolutional neural network and object detection technology.The purpose of this thesis is to study how to design and construct an automated and intelligent thermal pipeline engineering drawing review and analysis system.The system should identify the location and type of components from the drawings based on the deep network and object detection technology and extract corresponding information from it.The main work done in this thesis is as follows:Object detection algorithm design and model training stage.In this paper,we use a deep learning network based on InceptionResNetV2 and Faster R-CNN to extract features,locate targets and classificate targets on the design drawings and obtain a object detection network based on the industrial drawing component data set,then the OCR optical recognition technology is used to extract the data from the drawing area located by the object detection model.Object detection algorithm optimization phase.Aiming at the object detection problem of the original algorithm on the thermal pipeline drawing,we adjust the network structure of the Inception ResNetV2 network and optimize the parameter of the RPN network,include increasing the convolution branch of the Inception Res-net module,integrate the output feature map,and modifying the RPN Anchor parameters,to improve the model's ability to extract features and small targets for drawing targets.Design and construct an engineering drawing review and analysis system based on the above object detection algorithm.Elaborating the system design,system work flow and system function modules in detail,and demonstrate the actual working effect of the system.The thesis focuses on the theory and application of object detection technologybased on convolutional neural network and OCR optical character recognition technology.We analysis the insufficiency of the original algorithm in the environment of thermal pipeline engineering drawing to improve and optimize it,At the last,this thesis realizes a engineering drawing recognition and analysis system based on deep learning and object detection.