Since 1971, Unifi Manufacturing, Inc. has been a leading producer of multi-filament polyester and nylon textured yarns. The company’s manufacturing processes turn raw and recycled materials and fibers into synthetic filament yarns that behave like natural yarns, but have superior performance. Unifi’s yarns are found in many products manufactured by the world’s leading retailers—from clothing and industrial applications to home furnishings. Unifi’s customers can specify the yarn characteristics they need—such as dyeability and strength—then Unifi produces the customized yarn by adjusting machine settings using a process known as false-twist texturing. In Unifi’s quest to deliver high-quality synthetic yarns, the company turned to Minitab Statistical Software to help optimize and improve its false-twist texturing process.
To comply with guidelines surrounding manufacturing quality standards and the long-term storage of data, Unifi already collected and analyzed a vast amount of data for their false-twist texturing process. Using this data, process optimization took place daily through multiple statistical methods. These methods, including designed experiments, were used to build various types of causal models, which allowed technicians to predict yarn properties and select process parameters. However, the models were not easy to obtain and were an expensive option due to the high number of variables involved and large amount of time Unifi quality technicians needed to invest.
Unifi’s false-twist texturing process is very complex and contains a number of manipulated factors that are highly correlated. The presence of highly correlated factors, known as "multicollinearity," often results in redundant measurements and reduced statistical efficiency. Unifi set out to find a more efficient, cost-effective approach for the data analysis needed to optimize the false-twist texturing process. They found the solution in Minitab Statistical Software.
Unifi quality technicians used Minitab’s powerful data analysis to narrow their statistical approach to the analysis best suited for practical daily use. Taking advantage of the large quantity of data the company already had available, they used Minitab to evaluate and compare models from three different methods—multiple linear regression, principal components analysis, and partial least squares (PLS)—to find which would work best for optimizing the false-twist texturing process.
Minitab analysis of data collected at the Unifi spinning plant in Yadkinville, North Carolina, revealed that PLS analysis was the fastest and simplest method for generating accurate models, and was also the best of the three at remedying multicollinearity. The other approaches required additional experimental work, which was much more expensive and time-intensive—and thus not feasible for everyday use. With PLS, Unifi technicians also were able to identify outlier data sooner than they could when using classical linear regression. This saved time and allowed technicians to catch and fix potential problems with the process much earlier than they had been able to do before using PLS.
Minitab also made it possible for technicians to quickly update new production data based upon process adjustments, and then view the changes visually with graphs. Seeing how variable settings changed yarn properties made it easy to achieve the quality characteristics their customers need.
Unifi’s successful PLS approach shows how process optimization techniques can bring true benefits to false-twist texturing. Because the yarn industry’s widespread use of process improvement techniques and statistical software is still relatively new, Unifi’s success with Minitab demonstrates that these techniques can be applied more frequently and more extensively in the future.
The improvements to the false-twist texturing process have not only made it easier for Unifi to provide customers with highly-customized yarns of the highest quality, but improvements have also helped keep costs low. The time Unifi quality technicians spend completing experimental work to optimize the process has been dramatically reduced. "We’ve been able to reduce or completely eliminate experimental trials from our process," says Edmir Silva, technical manager at Unifi’s Yadkinville, North Carolina plant. "Minitab helped us bring actual value to our process—without incurring any major costs" he says.
Unifi Manufacturing, Inc.
Optimize and improve the false-twist texturing process