关键词: 计算机辅助配棉, 纱线质量预测, 机器学习, 优化算法, 产业互联

Abstract:

Significance Computer-aided cotton blending integrates advanced intelligent technology and traditional manufacturing, which is an important foothold of intelligent transformation of textile industry. The "Fourteenth Five-Year Plan" for developing the textile industry and demand for intelligent textile manufacturing call for comprehensively accelerating the industry's digital transformation, optimizing the production process, improving production efficiency, and achieving lean manufacturing. The textile industry's intelligent infrastructure has received a lot of attention during the "Thirteenth Five-Year Plan" period. Many technically advanced raw cotton information platforms and production information systems have emerged, gathering a sizable amount of raw cotton sales and public inspection data on the supply side of the raw cotton, and forming an internal enterprise including procurement, inventory, process, scheduling, products, sales and other dimensions of Standardized production data, constituting a sizable set of "supply and production" big-data system. For the intellectual development of China's cotton spinning firms, it is now imperative to find a way to maximize the value of supply and production data and to investigate intelligent management technologies that can significantly boost production efficiency and process level.

Progress The system framework of computer-aided cotton blending is introduced, and its development and application are summarized and analyzed around technical modules and technical connotations, in order to explore the future development of computer-aided cotton blending technology and promote the improvement of advanced management level and production efficiency of cotton spinning enterprises. The current research on intelligent raw cotton management aims to solve the optimization of raw cotton usage, which mainly includes the yarn quality prediction model and the cotton blending optimization model. (1) The yarn quality prediction model methods use supervised machine learning models such as multiple linear regression, support vector machines, artificial neural networks, and other improved models. In terms of model training approaches, evolutionary optimization algorithms have gained considerable attention in addition to the conventional analytical solution method and gradient descent method. (2) The cotton blending optimization model prioritizes cotton cost and yarn quality and creates a multi-objective optimization model based on inventory, total cotton consumption, cotton type, and cotton similarity. (3) To meet the needs of cotton spinning businesses for yarn quality management, a set of cotton blending technology management decision support system will be created in practical applications, integrating four functional modules of raw cotton inventory database maintenance, yarn quality prediction and management, cotton blending program formulation, cotton blending and yarn quality files, and a human-computer interactive interface.

Conclusion and Prospect After analyzing the two key components of the computer-aided cotton blending process, namely yarn quality prediction and the core technology utilized in the design of the cotton blending scheme, a number of difficulties with the current research are proposed: (1) The current study yarn quality prediction model lacks useful characteristics to characterize the performance distribution data of raw cotton in the cotton blending scheme and cannot adapt to the cotton blending scheme with length variation. (2) The current cotton blending optimization model optimizes the formulation of each cotton blending scheme as a single task and only for the production of a single variety (or a single production line), ignoring the fact that cotton blending is a time series task for multiple varieties on the multi-production line. Methods to improve the efficiency, precision, and generalizability of models must be investigated from the viewpoints of feature expression, model structure, and optimization technique. Also, the processing efficiency of large data and the universality of production mode must be enhanced. On the one hand, the computer-aided cotton blending system continues to progress the standardization and expansion of raw cotton quality inspection, while focusing on the in-depth application of big data for raw cotton. On the other hand, it will likely accelerate the intelligent transformation and upgrading of the textile sector. As big data and artificial intelligence technologies continue to improve, it is expected that the computer-aided cotton blending system study will make major strides in integrating cloud market data and adapting to the personalized production mode of future businesses.

Key words: computer aided cotton blending, yarn quality prediction, machine learning, optimization algorithm, industrial interconnection