The cotton ginning work was carried out in eight factories of the main cotton companies operating in Côte d’Ivoire (COIC, SECO, IVOIRE COTON and CIDT) at a rate of two factories per cotton company ( Figure 1 ). These factories are generally equipped with Lummus-type gins with 170 saws.

The plant material used in this study consisted of seed cotton of Gossypium hirsutum L. specie, that is the only one cotton specie grown in Côte d’Ivoire. The seed cotton collected from the fields and sent to the factories from the tree production generations (G3, R1 and R2):

G3: Cotton given by research as created verity;

R1: First cultivation year;

R2: Second cultivation year.

The technical equipment used in this study consisted of 170-saw gins at the mills ( Figure 2(a) ) and an HVI 1000/1000 integrated measurement chain ( Figure 2(b) ) for determining the technological characteristics of the fibres.

The data collected at the factories were made possible by ginning tests [6] , carried out on loads of seed cotton arriving in ginning factories, for ginning. During the tests, waste rates and fibre yields were determined according to Formulas (1) and (2) respectively.

Formula (1): $\boxed{\text{Fibre}\text{content}\left(\text{\%}\right)=\frac{\text{Fibre}\text{weight}\left(\text{kg}\right)}{\text{Seed}\text{cotton}\text{weigth}\left(\text{kg}\right)}\times 100}$

Formula (2): $\boxed{\text{Waste rate}\left(\text{\%}\right)=\frac{\text{Waste}\text{weigth}}{\text{Seed}\text{cotton}\text{weigth}\left(\text{kg}\right)}\times 100}$

The fibre samples collected after ginning were analysed on an HVI 1000/1000 integrated measurement chain [6] . This fully automated device has three modules for determining agro-industrial quality from a sample of 100 g of fibre. The main technological characteristics assessed were fibre length (UHML), fibre tenacity, short fibre content and fibre length uniformity.

In order to assess the impact of waste on the agro-industrial quality of cotton, linear regressions were used to study its relevance by predicting technological characteristics as a function of these impurities.

Principal component analysis (PCA) was used to observe the relevance of the technological parameters studied. All these analyses were carried out using Excel 2016 and STATISTICA version 7.1 software.

The rate of waste contained in seed cotton increased with the R² generation, with a value of 5.33% ( Figure 3 ). On the other hand, the lowest rate was obtained with generation G3 (2.96%).

Figure 4 shows that a significant prediction (p = 0.002) could also be made between waste rate and cotton fibre yield. The accuracy was approximately 62% (R² = 0.6273). The equation of the regression line showed that it is also negative (y = −0.6024x + 45.376).

The relevance of the variables studied is shown in Figure 5 . Most of the information was contained on axes 1 and 2 (84%). On these axes, the most explanatory variables were the quantity of seed cotton (material), the rate of waste (total losses) and the rate of cotton fibre (fibre yield). For these variables, the correlations were negative between the quantity of seed cotton and the rate of cotton fibre on the one hand, and the rate of waste on the other.

The regression of cotton fibre length as a function of cotton waste content is shown in Figure 6 . The regression was negative (y = −0.4408x + 28.991) with a precision of only 39% but was still significant (p = 0.005). The more waste, the shorter the cotton fibres.

According to the linear regression of the graph in Figure 7 , the level of waste contained in the cotton had no influence (p = 0.441) on the tenacity of the cotton fibres.

The influence of seed cotton waste on short staple fibre content was highly significant (p = 0.000) and positive ( Figure 8 ). The coefficient of determination was 0.5833. In fact, the more impurities there are in the seed cotton, the more short fibres are generated by the ginning process in the mass of fibres obtained.

A negative influence (y = −0.5963x + 81.078), significant (p = 0.001) and with a coefficient of determination R² = 0.5185 was observed ( Figure 9 ). So, the more waste there is in the harvested cotton, the less uniform the fibre lengths will be.