【作者】 王旭；

【导师】 戴华平；

【作者基本信息】 浙江大学 ， 控制理论与控制工程， 2013， 硕士

【摘要】 铜金属具有很多优良特性,在生产和生活各个方面有着广泛的应用。随着经济的高速发展,传统铜矿日益枯竭,废杂铜资源显得愈发重要。废杂铜来源广泛成分含量参差不齐,加上废杂铜冶炼工艺和参数的差异,致使冶炼得到的再生铜的铜成分含量不一。再生铜中不同的铜成分含量代表了不同的品质,因而检测再生铜的铜成分含量是废杂铜冶炼自动化过程的关键技术之一。目前现有的方法无法有效实现准确快速地检测。为此,本文针对现有再生铜品质检测方法存在的问题,在研究了光谱成像技术的基础上,提出了两种基于高光谱特征的铜成分回归建模方法,通过建立的回归模型实现再生铜品质的检测。本文的研究工作主要包括以下几个方面：1)由于成像光谱仪成套系统价格昂贵,在分析了成像光谱仪的物理原理和实现原理的基础上,结合本文再生铜品质检测的应用背景,进行了硬件平台的部件选型和组装搭建。根据硬件平台特性和应用需求,开发了一款配套的成像光谱仪铜品质检测软件。2)提出了再生铜样本高光谱特征的提取方法,在分析讨论了样本数据处理、核函数选取以及建模参数优化的基础上,提出了基于高光谱特征铜成分的LSSVR回归建模方法,设计了LSSVR回归建模的详细步骤。根据详细的建模步骤,进行了实验研究,实验结果表明通过建立的回归模型可以实现再生铜品质的准确快速地检测。3)在研究了非负矩阵算法(NMF)和粒子群算法(PSO)理论的基础上提出了基于粒子群的非平滑非负矩阵算法(PSO-nsNMF)。结合本文再生铜品质检测的应用背景,利用PSO-nsNMF算法提取局部非负信息的特性,提出了基于高光谱特征的铜成分PSO-nsNMF-LSSVR回归建模方法,设计了PSO-nsNMF-LSSVR回归建模的详细步骤。根据详细的建模步骤,进行了实验研究,实验结果表明通过建立的回归模型可以实现再生铜品质准确快速地检测。最后,对本文工作进行了总结并给出了进一步的研究展望。更多还原

【Abstract】 Copper has many excellent features and has been widely used in the production and life. With the rapid development of the economy, the increasing depletion of traditional copper mine, scrap copper resources is increasingly important. The content’s differences in scrap copper and smelting process differences lead to differences of the recycled copper in content, but different content represents a different quality, and therefore it is important to detect the copper’s content of the recycled copper. At present, existing methods cannot effectively achieve online accurate detection.In order to solve the problems of the existing regeneration copper quality detection methods, we studied the spectral imaging technology and put forward copper-based ingredients regression modeling methods based on hyperspectral characteristics of copper, through the model to achieve the quality detection of the recycled copper. The main research work includes the following aspects:1. Because of the high price of complete systems for imaging spectrometer, Combined with the application background of recycled copper quality detection, we analyzed the basis of physical principles and realization of the principle of the imaging spectrometer and then took parts selection and assembly of the hardware platform. According to the characteristics of the hardware platform and application requirements, we developed a copper’s quality testing software supporting the imaging spectrometer.2. After putting forward an extraction method of hyperspectral characteristics of the recycled copper sample, discussing sample data processing and kernel function selection and optimization modeling parameters, we put forward a LSSVR regression modeling method based on hyperspectral characteristics. Based on detailed modeling steps, we conduct an experiment to verify the feasibility of the LSSVR regression model.3. After studying the theory of non-negative matrix algorithm (NMF) and particle swarm optimization (PSO), we put forward a new algorithm called non-smooth non-negative matrix particle swarm algorithm (PSO-nsNMF). Combined with the application background of recycled copper quality detection, using the PSO-nsNMF algorithm to extract local non-negative nature of the information, we put forward a PSO-nsNMF-LSSVR regression modeling method based on hyperspectral characteristics. Based on detailed modeling steps, we conduct an experiment to verify the feasibility of the PSO-nsNMF-LSSVR regression model.In the end, future work is pointed out and research is still ongoing.更多还原