Image Analysis-based Control System Of Scrap Copper Smelting Processes

Energy conservation is an important issue which should be paid great attentions in scrap copper smelting processes. However, current detection techniques of smelting dust concentration and copper content introduce too much time delay and cause high energy consumption. Based on processing technology analysis, this dissertation proposes smelting dust and copper content estimation image detection methods and image feedback-based smelting process closed-loop control methods, which are summarized as follows:1. The dust concentration is difficult to detect online in scrap copper smelting process. For the problem, an image feature-based dust detection method is proposed. According to correlations between smelting dust images and dust concentration, the flame luminance index and background blurring index is constructed as new dust image features, which achieves low-cost and online detection of smelting dust. Additionally, the image feature extraction method is improved to filter sunlight noises, which is more suitable in varying daylight conditions.2. The open-loop control of de-dusting fan speed leads to high enelgy consumption in the smelting process. For the problem, a new image feature feedback-based closed-loop control method is proposed. Firstly, a fan speed closed-loop control system is designed, which is based on dust image feature feedbacks. Then, an image feature classification-based fan speed controller model is developed according to the image feature distribution characteristics. Finally, the controller model is improved to a regression model, which is a dynamic multi-model construction. The improved controller could adjust fan speed more smoothly than the classification-based controller.3. To solve the problem of copper content offline detection, a color feature-based secondary copper content estimation method is proposed, which achieves rapid and low-cost detection of copper content. Three copper content least squares support regression models are developed, which are RGB-based model, color vector angle-based model and hue intensity-based model. By comparative study, three color based models are effective to detect secondary copper content. Among them, the color vector angle-based model has the best estimation accuracy.4. For the problem that surface defects affect the color characteristics of copper sample images, a new copper content estimation method is proposed, which is based on ROI selection and improved color vector angle. Firstly, a defect detection based ROI (region of interest) selection strategy is proposed to select the modelling ROI before color feature extraction. Then a new color vector angle is constructed according to the unique color distribution of secondary copper. Finally, the improved copper content model is developed based on the least squares support vector regression.5. To solve the problem that current domestic scrap copper smelting enterprises lack available energy-saving control system, a double-loop control system is developed based on image feature feedbacks. The external loop controls the power of smelting furnace according to the image based copper content estimation, while the inner loop controls the de-dusting fan speed using the smelting dust image feature feedback. An actual prototype system has been put into operation and has got excellent economic benefits.Finally, the conclusions and challenges of future studies are illustrated.