The Development of Cotton-yarn-quality Predicting System

---

The Development of Cotton-yarn-quality Predicting System 
Ying Xiao
1 
Tianjin Polytechnic University, School of Textiles 
Tianjin, China 
[e-mail] 
Shulin Zhao
2
Tianjin Polytechnic University, School of Textiles
Tianjin, China 
Abstract: A new cotton-yarn-quality predicting system was 
developed by using the merging programming technique of 
VB and Matlab in the paper. Using this system, the tenacity 
and evenness CV% of the cotton-yarn processing by 
conventional spinning at standard temperature and 
humidity can be predicted through inputting some fiber 
properties into the system, such as the percentage of 
impurities, the principal length, the percentage of short 
fiber, the degree of maturity, fiber strength and the value of 
Micronaire. And also it was verified that the system did the 
good job for predicting yarn tenacity and evenness CV% 
exactly with the relative error of less than 4% after the 
model being trained. The accuracy can meet the demand of 
spinning factories and so the predicting results would be 
useful for guiding the spinning practice.
Key words: BP network; network model; yarn quality 
prediction; merging programming of VB and Matlab 
I. 
I
NTRODUCTION
The quality of yarns is mainly determined by fiber 
properties, spinning processing, spinning equipments and 
the temperature and humidity of workshops, and so on. In 
these factors, fiber properties are the most important one 
for the yarn quality. In order to predict the yarn quality 
through digital simulation, it is essential to analyze the 
relationship between fiber properties and yarn quality 
under any other conditions unchanged, such as processing 
temperature and humidity. In this paper, a BP neural 
network model was set up and based on the model, a new 
yarn-quality predicting system was developed by which 
the tenacity and evenness CV %of cotton yarns by ring 
spinning process at the standard temperature and humidity 
could be predicted through inputting six fiber properties 
into the system. Furthermore, it was validated that the 
system model did a good job for predicting both the yarn 
tenacity and the evenness CV% exactly with the relative 
error of less than 4%.
II. T
HE 
D
ESIGN OF THE 
O
VERALL 
S
TRUCTURE
A system was developed by using the merging 
programming technology of VB and Matlab and the 
system integrated the user-friendly interface of VB and 
the powerful function for data processing of Matlab. The 
overall system structure was shown in Figure 1. 
Figure1.The Overall System structure 
The system mainly consisted of the neural network 
building module, network training module, yarn-quality 
predicating module, the model database and other 
modules. Each module could operate coordinately under 
the domination and regulation of main control program of 
VB and realize many functions, such as data document 
read and writing, network modeling, network model 
training, retention, call, and yarn quality prediction. 
III. BP
N
EURAL 
N
ETWORK 
M
ODEL 
D
EVELOPMENT
1. The Introduction of BP Neural Network 
The BP Neural Network ( short for BP network), also 
known as back propagation network, is a multiple layer 
feed-forward network, which generally includes an input 
layer, an output layer and one or several hidden layers, 
with its operation mainly consisting of two processes: the 
feed-forward transmission of the information and the 
reverse transmission of the error
[1]
. In the former process, 
the information was calculated from the input layer to 
output layer, layer by layer via hidden layer ,with the 
neural unit state of each layer only affecting the next 
one's ;and in output layer, the system would turn to back 
propagation if the desired results were not achieved when 
compared with the target value. Then the error was 
transferred from the output layer to input layer with the 
network algorithmically adjudging its own weights and 
thresholds and the adjusting process was repeated till it 
reached the desired results.