基于RBM和T-S模型的料位软测量方法的研究

球磨机是火力发电厂的基础设备，可靠测量料位是实现系统优化的关键。针对球磨机音频信号中存在强噪声、非线性等问题，结合受限玻尔兹曼机(RBM)、减法聚类和T-S模糊模型，提出了一种软测量方法。首先采用微调后的玻尔兹曼机提取特征，去除存在的噪声，然后使用减法聚类辨识模糊模型的初始结构，最后采用T-S模糊模型预测球磨机料位。通过在球磨机运行数据上进行模型验证，验证了该方法的实用性和可行性。

Study on Soft Sensor Approach for Fill Level based on RBM and T-S fuzzy model

Ball mill is a basic equipment in thermal power plant, which is a key factor for the mill system optimization to measure the fill level accurately. The acoustic frequency spectrum of ball mill has strong noise and nonlinearity, which reduces the measurement accuracy. To solve the problems, a soft sensor method is proposed, which combines restricted boltzmann machine (RBM), subtractive clustering and Takagi-Sugeno fuzzy model. Firstly RBM having been fine-tuned is employed to extract the features and remove the existing noises. Then subtractive clustering is used for fuzzy system structure Identification. At last the fill level is predicted by the T-S model. The results based on the collected data of ball mill validate the feasibility and practicability.