Abstract:

Objective Each machine used in the weaving workshop production scheduling scheme is closely related to the textile enterprise production efficiency. In order to solve the problem of existing loom machine caused by time depending on human experience and hard to predict production scheduling, this paper, through the study of dynamic loom weaving process of actual production data prediction algorithm, puts forward a kind of loom machine to help production personnel according to the loom machine time, determines the best production scheduling strategy of the whole production line, improves the production efficiency of the production line and increases the profit of the enterprise. This method does not depend on hardware cost, greatly improves enterprise profit, and is easy to be implemented.

Method This paper proposes a machine prediction algorithm based on K-means theory. Taking each loom in the weaving workshop as the object, the whole production process of air-jet loom is reasonably divided into five production states by establishing the theoretical model of loom machine prediction. In addition, the dynamic data (weaving yield, sampling time point, current shift, remaining yarn length, variety number) and static data (warp axis number, set axis length) of loom machine production process in the workshop were recorded in time sequence, and the 7-dimensional vector matrix X was established as the original data sample set. Finally, the problem was solved by Python built-in mathematical processing module.

Results The time series data of characteristic samples in a certain period of time in the actual production process of the equipment were input into the dichotomous K-means algorithm with a clustering index of 5. The algorithm divided the production state data into five clusters, and the orange point was the centroid of the corresponding cluster region. According to the production characteristics, the predicted value of the remaining machine time of the loom machine at this time can be calculated. Sample data of 10 loom machines (the time period of complete production of a warp shaft) were randomly selected from 480 loom equipment in the workshop. The sample data were input into the algorithm to calculate the predicted values of 144 h, 96 h, 84 h, 72 h and 24 h from the actual loom machine and compare the calculation errors with the actual machine values. The results showed that with the passage of time and the accumulation of data. The error between the predicted loom machine value and the actual loom machine value calculated by the dichotomous algorithm model gradually decreased, and the average error of the predicted loom machine value was less than 0.9 h within 72 h before the machine.

Conclusion The data results show that the error between the predicted loom machine time and the actual loom machine time is small, which proves the correctness of the theoretical model of loom machine prediction, and the maximum absolute error is not more than 0.9 h (far less than the 2 h error required for production scheduling), which meets the timeliness and accuracy required for production scheduling in the weaving workshop. In addition, the weaving machine prediction algorithm based on bisection K-means theory established in this paper needs to be built on the accuracy and integrity of the data collected by the equipment. Therefore, in the future, we should start to ensure the full coverage of the weaving workshop network and stable and efficient data collection and transmission mode. Although theoretical model of the machine does not support the prediction of warping and sizing equipment, it is also applicable to the spinning machine equipment with similar processes, such as water spraying and rapier type, and has important engineering application value.

Key words: loom, production scheduling, machine prediction, binary K-means, weaving process