基于非参数回归的发酵过程软测量建模

为了对发酵过程进行检测并分析发酵中产物的浓度，提出一种基于非参数回归的发酵过程软测量方法。该方法直接利用未经处理过的历史数据，采用多维窗宽选择方法确定窗宽，通过计算待估计点与历史数据库中其他时刻点数据的距离来确定该时刻点的权值，从而达到软测量建模的目的，并将其用于青霉素发酵过程产物浓度的测量。实验研究表明，所得软测量模型精度较高，能够有效地实现对青霉素发酵产物浓度的估计。

A non-parametric soft sensor modeling method for use in fermentation processes

A soft sensor for use in fermentation processes which allows the process to be followed and product concentration to be analyzed has been simulated. The soft sensor is based on nonparametric regression which leverages untreated historical data directly, uses a multidimensional window width selection method to confirm window width, calculates the distance between the estimation point and other moment points in historical data to confirm the weight of the point at the given moment, in order to simulate the soft sensor and measure the product concentration during a penicillin fermentation process. The results showed that the simulation model is highly precise and it succeeds in estimating the product concentration from penicillin fermentation, which proves the accuracy and effectiveness of the method.