| At present, energy shortage and the increasing cost makes many dyeing enterprises have no benefits, and this situation seriously affects the companies' sustainable development. The main reasons resulted in this situation include: weak basic work of energy management, low energy management level, serious waste, inreasing cost etc. To solve these problems, energy detecting and measurement must be emphasized.Because the range and quantity is large while measuring energy consuption in dyeing workshop, the invest cost will be high if all measurements are used instrument. What's more, it is difficult to establish accurate mathematical model. because the process of dyeing production is a typical batch production process with characteristics of non-linear, time-varying input and output etc. To survive from this plight, soft-sensing method was applied to detect real timely the dyeing energy consumption, which provides a new idea. This article used soft-sensing method of artificial neural network in real-time detection of workshop enery consumption. This method mades full use of the strong non-linear approaching and learnning abilities and achieved very good results.This article took the energy consumption of dyeing production process as an example, systematic analysised the major factors influenced energy consumption, combined with the actual situation of some dyeng enterprise workshop and put forward two different solution models.1,Soft-sensor model built by improved BP neural network basing on in-depth understanding of the affectting factors of dye vat energy consumption. The model uses the major factors of energy consumption as inputs and makes the energy consumption of each process outputs. Through the model, the consumption of steam, water and electricity of each stain procedure are determined. Then they are compared with the actual collection of data to amend the model.2. Sample data is divided into different types of value centers by similar guidelines and Fuzzy c-means clustering (FCM) algorithm.Each type has a membership degree with its counterparts. Each classification was built mode based on the RBF network and the soft-sensing outputs were got by model output for membership weighted sum. This method is more suitable for the problems with much sample data and certain measurement accuracy requirements. Soft-sensing model is adjusted by calculating superposition of a bias.Further, two methods of measurement software are analysised and compared. The results showed that the FMM algorithm's results have been greatly increased compared with the simple generalization results of the RBF.Finally, the dyeing energy consumption software system was designed and realized.. Through soft-sensing we can get each dyeing vat's and process's energy consumption and make a pre-warning over the vat energy to control energy consumption in an optimal range of settings. |
