首页|基于GMM和GA-LSTM的稀土熔盐电解过程原料含量状态识别模型

基于GMM和GA-LSTM的稀土熔盐电解过程原料含量状态识别模型

扫码查看
在高温高风险的稀土熔盐电解工艺中,为了实现稀土熔盐电解过程原料含量状态的智能识别,提出了一种基于混合高斯背景建模(GMM)和遗传算法优化的长短期记忆神经网络(GA-LSTM)的分类模型.模型通过GMM算法、R通道自适应滤波和中值滤波准确提取图像的火焰前景和特征,以量化熔盐电解反应的剧烈程度,进而判断稀土熔盐电解处于原料含量过多或含量正常状态;然后利用GA-LSTM神经网络建立熔盐表面火焰特征和稀土熔盐电解过程原料含量状态的非线性映射关系.结果表明:模型的识别精度高达99.79%,具有较好的泛化性,为实现稀土熔盐电解工艺自动化提供了一定的参考价值.
Raw material content status identification model in rare earth molten salt electrolysis process based on GMM and GA-LSTM
In the high-temperature and high-risk rare earth molten salt electrolysis process,in order to achieve intelligent identification of raw material content status during the rare earth molten salt electrolysis process,a classification model based on Gaussian mixed background modeling(GMM)and genetic algorithm optimized long short-term memory neural network(GA-LSTM)was proposed.The model accurately extracted the flame foreground and characteristics of the image through the GMM algorithm,R channel adaptive filtering and median filtering to quantify the intensity of the molten salt electrolysis reaction and then determine whether the rare earth molten salt electrolysis was in a state of excessive raw material content or normal content.Then the GA-LSTM neural network was used to establish the nonlinear mapping relationship between the flame characteristics of the molten salt surface and the raw material content state during the rare earth molten salt electrolysis process.The results show that the recognition accuracy of the model is as high as 99.79%,and it has good generalization,providing a certain reference value for realizing the automation of rare earth molten salt electrolysis process.

rare earth molten saltflamefeature extractionGaussian mixture model(GMM)long short-term memory(LSTM)networkgenetic algorithm(GA)

张震、朱尚琳、伍昕宇、刘飞飞、何鑫凤、王家超

展开 >

江西理工大学 电气工程与自动化学院,赣州 341000

中南大学 自动化学院,长沙 410083

江西理工大学 机电工程学院,赣州 341000

稀土熔盐 火焰 特征 混合高斯模型 长短期记忆神经网络 遗传算法

江西省重点创新研发平台计划

20181BCD40009

2024

中国有色金属学报
中国有色金属学会

中国有色金属学报

CSTPCD北大核心
影响因子:1.108
ISSN:1004-0609
年,卷(期):2024.34(5)
  • 24