Garment Sewing Machine Fault Prediction Algorithm Based on Machine Learning

During the garment production process, garment sewing machines frequently malfunction, often causing a series of problems such as low production efficiency, high maintenance costs and poor product quality. This paper uses time series analysis and XGBoost to predict the fault of clothing sewing machines. This paper first collects various data such as vibration signals and motor current during the operation of the sewing machine, removes data noise and outliers, and ensures data quality. Then, based on Fourier transform, key features such as the spectral characteristics of the vibration signal and the volatility of the motor current in the sewing machine operation data are extracted, the features are screened and optimized, and the feature subset that is most sensitive to fault prediction is selected. XGBoost is selected as the fault prediction model, and the model is trained and the model parameters are optimized using the extracted feature subsets and operation data. The model is preliminarily evaluated using methods such as cross-validation, and necessary adjustments and optimizations are made to the model based on the evaluation results. Compared with the SVM model, XGBoost shows advantages in both prediction accuracy and prediction time. Its prediction accuracy is always higher than SVM. The average prediction time of XGBoost is about 232.6 milliseconds, but that of SVM is 401.1 milliseconds. This shows that XGBoost has higher efficiency and accuracy in processing garment sewing machine fault prediction tasks, and can provide reliable fault warning information for garment production, helping to timely discover and deal with potential fault problems, thereby improving production efficiency and equipment reliability.