本文選題：窯爐控制 + 機器視覺　； 參考：《杭州電子科技大學(xué)》2016年碩士論文

【摘要】：窯爐控制系統(tǒng)的研究在國內(nèi)起步較晚,發(fā)展并不完善,多數(shù)是基于工控機,缺少智能化技術(shù)。目前,嵌入式控制系統(tǒng)已經(jīng)成為工業(yè)控制的主要應(yīng)用,而ePLC(embedded PLC)更是嵌入式控制系統(tǒng)的最新技術(shù)。本文在基于ePLC的窯爐控制系統(tǒng)的基礎(chǔ)上,研究通過機器視覺對坯體入窯密度的自動檢測,實現(xiàn)對窯爐的智能化控制,這對提高我國窯爐控制的技術(shù)水平具有較大的現(xiàn)實意義。論文研究了在嵌入式硬件平臺上,實現(xiàn)視覺算法與機器學(xué)習(xí)算法。在視覺模塊上集成了多個圖像算法,用構(gòu)件化的形式封裝在CASS機器視覺平臺上。主要為采集的圖像做圖像預(yù)處理,包括采集圖片的有效信息位置剪裁、圖像灰度立方圖均衡化、圖像閾值二值化以及圖像仿射變換。對處理后的圖像信息,運用改進的M-ary SVM(Support Vector Machine)算法進行分類。對于分類編碼,在信息碼中加入了糾錯編碼,增強了其泛化性。在圖像的特征提取方面,提出了整數(shù)權(quán)值的特征提取法,在0-1矩陣圖像里,直接提取16位二進制數(shù)作為一個整數(shù)成為一個特征值,適用于嵌入式平臺上的機器視覺實現(xiàn)。最后,在整窯的入窯密度檢測方面,通過對經(jīng)由SVM分類的各層的入窯密度,進行權(quán)值疊加得到。根據(jù)窯爐內(nèi)部上下部分的加熱環(huán)境不同,在反饋給窯爐控制時,加入權(quán)值計算總體的熱力需求,再選擇合適的曲線進行產(chǎn)品燒制。整個窯爐控制系統(tǒng)都是基于ePLC開發(fā),是智能化、自動化更高的控制系統(tǒng)。該系統(tǒng)經(jīng)實驗測試,能有效通過機器視覺對坯體入窯密度進行自動檢測,能實現(xiàn)對窯爐的數(shù)碼控制。
[Abstract]:The research of kiln control system starts late in our country, and its development is not perfect. Most of the research is based on industrial control computer and lack of intelligent technology. At present, embedded control system has become the main application of industrial control, and ePLC embedded PLC is the latest technology of embedded control system. Based on the control system of kiln based on ePLC, this paper studies how to realize intelligent control of kiln by machine vision, which is of great practical significance to improve the technical level of kiln control in China. This paper studies how to realize vision algorithm and machine learning algorithm on embedded hardware platform. Several image algorithms are integrated into the visual module and encapsulated in the Cass machine vision platform in the form of component. Image preprocessing is mainly done for the collected image, including the effective information position cutting, the image gray cube image equalization, the image threshold binarization and the image affine transformation. The improved M-ary SVM support Vector Machine algorithm is used to classify the processed image information. For classification coding, error correction coding is added to the information code, which enhances its generalization. In the aspect of image feature extraction, an integer weight feature extraction method is proposed. In 0-1 matrix image, 16-bit binary number is directly extracted as an integer to become a feature value, which is suitable for machine vision implementation on embedded platform. Finally, in the whole kiln density detection, through the SVM classification of each layer of the kiln density, the weight of the superposition. According to the different heating environment of the upper and lower parts of the kiln, when feedback is given to the kiln, the weight value is added to calculate the total thermal demand, and then the appropriate curve is selected for the product firing. The whole kiln control system is based on ePLC. It is an intelligent and automatic control system. The system can detect the density of billet into kiln by machine vision and realize the digital control of kiln.
【學(xué)位授予單位】：杭州電子科技大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2016
【分類號】：TP273

【相似文獻】

相關(guān)碩士學(xué)位論文 前1條

1 鄭維凱;基于ePLC的窯爐數(shù)碼控制系統(tǒng)的研究與應(yīng)用[D];杭州電子科技大學(xué);2016年

，

本文編號：2009676