INTRODUCTION
Professionals consider the shape modifications of knitted fabrics during
different steps of fabrication and use as an unavoidable problem. That`s
why, many fundamental researches treated the stability of the knitted
clothes and machine constructors are continually developing systems aiming
to control and reduce the bad impact of the dimensional variations, especially
concerning the length of the absorbed yarn, on the fabrics.
The literature reports some of the parameters that have the main influence
at different levels on the geometric properties of relaxed knitted fabrics,
e.g. stitch length, the yarn physical properties (Chen, 2002).
The effect of some knitting parameters on the mechanism of loop formation
(Knapton, 1966a, b). They proposed some modifications on the cam system
geometry to ameliorate the knitting conditions. Shanahan (1970) established
different relaxation conditions that were more or less drastic and tried
to analyse theoretically the structure of the knitted fabric in terms
of the loop configuration. Some researchers tried to improve the dimensional
stability of the fabrics by chemically treating the fiber (Cortez, 2002).
Even if its known that it is difficult in practice to obtain a perfectly
relaxed fabric (Shanahan, 1970) we tried out to study, by statistical
means the influence of some knitting parameters in the two extreme cases
of relaxation which are dry relaxation and industrial one using the relaxlab.
The fabric is made by a nondyed High Bulk Acrylic yarn.
MATERIALS AND METHODS
The fabrics were made with the 122FF Shima Seiki knitting machine equipped
with DSCS and Parameters variations are done automatically using a PC
which monitors all the machine components.
Dry relaxation consists on conditioning the fabric during one week at
room temperature (25 °C) and constant relative humidity (about 60%).
Industrial relaxation using the relaxlab consists on vibrating the fabric
for 15 min in presence of watersteam. The digital stitch control system helps to knit a prototype or a test
piece of a pattern.
The data for a particular pattern are based on the loop length and the
number of stitches in each row. The interest of the DSCS is to treat each
stitch as a single data, with whole share.
During the knitting of each row, the machine`s computer monitors and
regulates the stitch length, referring to the data sent from the DSCS,
in order to maintain it constant on following rows, with a high degree
of accuracy, whatever are the knitting conditions such as the nature of
the yarn or the external parameters (yarn tension, humidity).
Factorial design: Table 1 shows the factors
which will be taken in account to investigate their influences on the
variables to be studied. These latest are the length of the absorbed yarn
(called LFA) and the basis weight. The different values of the factors,
also called the explanatory variables, used in this study are indicated
in the third column.
Table 1: 
Explanatory variables: names, types and levels 

The softwares used are SPSS 9 and Minitab 14, 18,
22 and 26 are equivalent respectively to 0.925, 1.02 and 1.11 cm/loop.
RESULTS AND DISCUSSION
Length of the absorbed yarn: According to the statistical tests,
the results below are obtained:
• 
p>α = 0.05, where p is the value given by the
normality test of Anderson Darling 
• 
Average of the residues = 0.000 
• 
Test of Durbin Watson, D.W. = 1.65: a value which can be statistically
considered close to 2. 
So the data follow a normal law and the coefficients of the explanatory
parameters can be estimated by the ordinary least squares method.
In addition, Fig. 1 shows that the loop length is the
most influent factor. The carriage speed and the interaction (carriage
speed/DSCS) are influencing the considered answer, too.
Further we are trying to find out the operation state of the DSCS in
which the speed has the greatest influence on the LFA?
According to Table 2, we can say that, whatever is
the type of relaxation applied to the knitted fabric, we obtain the following
results:
• 
Without using the DSCS, the parameters that influence
the stitch length, by descending order and by accepting a risk of
error of 5% in the estimation of the explanatory variables coefficients,
are the loop length then the speed of the carriage. The variation
of one of these two factors is proportional to that of the stitch
length. 
• 
By using the DSCS, the factor speed looses its significance (t test
< 1.96) and the takingdown remains always a non significant factor. 
That is due to the fact that the DSCS regularizes the positions of the
stitch cams in each passage of the carriage, in order to ensure a constant
consumption of the yarn whatever are the yarn tension around the spool
and the flow of the yarn.
In the other case, when the operation state of the DSCS is off, it can
be said that the carriage is directly related to the spool and in our
experimental conditions, a considerable increase in its speed generates
a greater delivery of the yarn. Consequently, the quantity of yarn absorbed
by needles will be greater.
Basis weight: The same steps of study were carried out for the
answer basis weight and confirmed that the values of the variables follow
a normal law and the validity of estimating the coefficients of the explanatory
factors by the ordinary least squares.
The diagram of Pareto showed that the basis weight is influenced primarily
by the following factors, arranged in descending order: Relaxation, loop
length, Interaction (relaxation/taking down) and Interaction (DSCS/carriage
speed/relaxation).
The basis weight is largely influenced by the industrial relaxation and
the loop length. In fact, during the industrial relaxation, the acrylic
high bulk yarn, in view of its structure, shrinks in a remarkable way
in the directions of column and row. So that the stitches` density will
be greater, which explain the considerable increase of the surface mass
of the knitted fabric.
The more the loop length increases, the more the basis weight decreases (negative
coefficient of regression of the loop length). That`s due to the fact that the
quantity of yarn consumed, per unit of area, by a loose knitted fabric is less
than that consumed by a sleazy one.
The effect of the takingdown is meaningful only when the knitted fabric
is relaxed dry (p < 0.05). In the other case, the factor taking down
loses an important share of its significance. Therefore, the relaxation
using the relaxlab could hide the information which the variable takingdown
can carry about the basis weight. That`s because the steam and the vibrations
in which the fabric was subjected during its relaxation on the relaxlab
cancelled all the mechanical interactions and stresses within the stitches.
These internal interactions represent the effect brought by the takingdown
(Table 3).
In the previous results concerning the stitch length, we found that the
carriage speed is proportional to this response when the DSCS is not in
operation and considering that the increase of absorbed yarn generates
the reduction of the basis weight, it can be said that speed is inversely
proportional to the basis weight. However, according to the statistical
results showed on the Table 3, in the case of DSCS off,
the increase of the stitch length generated by the increase of the carriage
speed doesn`t have a significant effect when the knitted fabric is not
relaxed by the relaxlab (p>0.05). In the other case, the variation
of the carriage speed can have a significant influence with 5% errors
risk.

Fig. 1: 
Pareto diagram of the main effects using α = 0.05 
Table 2: 
The results estimation of the explicative model for
each combination (DSCS, relaxation); stitch length = f (loop length,
takingdown, carriage speed) 

*Significant to 10% ; **Significant to 5%; ***Significant
to 1% 
Table 3: 
The results estimation of the explicative model for
each combination (DSCS, relaxation); basis weight = f (loop length,
takingdown, carriage speed) 

*Significant to 10% ;**Significant to 5%; ***Significant
to 1% 
This result can be interpreted as follows: since dry relaxed knitted
fabric shrinks little in the columns and rows directions and considering
the increase of the stitch length per unit of area, generated by the carriage
speed, is not so important to influence significantly the basis weight,
there will not be a significant variation of our dependant variable. Whereas,
in the case of an industrial relaxation, the important shrinking of the
fabric, generates an accumulation of the amount of yarn added by the carriage
speed effect. Thus, this accumulation will have a significant effect which
is inversely proportional to the basis weight. This last result confirms
the results obtained for the first output variable.
CONCLUSION
Obtained results show the contribution of the digital stitch control
system to the stability of the knitted fabrics and their regularity. In
addition, the factorial design allowed us to establish the relationship
between all treated parameters and their interactions either in contribution
model or in estimating one. We showed in that research only the first
models.
Nevertheless, the results obtained in this study can be followed by controlling
other parameters during knitting operation like the tension of the yarn,
relative humidity and temperature of the knitting room, rugosity of the
yarns.