Abstract:

Objective The objective of this research is to develop an efficient and accurate positioning method tailored for the complex environment of winding machine assembly workshops. These workshops present unique challenges due to their large size, intricate layouts, and the high precision required for the accurate placement and movement of materials and equipment. Effective positioning is crucial for optimizing production processes, improving resource allocation, and reducing operational costs. Traditional positioning methods have certain limitations: for instance, global positioning system has low positioning accuracy in indoor environments, while other methods (such as infrared or ultra-wideband (UWB)) can provide higher precision but are costly to deploy and complex to maintain. Therefore, this study selects wireless networks(WiFi) as the foundational positioning technology, as its signals are widely distributed in indoor environments, possess strong penetration capabilities, and offer a more cost-effective and deployable solution compared to other approaches. This research aims to develop a hierarchical positioning method based on WiFi to meet the specific needs of winding machine assembly workshops.

Method The proposed method integrates three key algorithms, i.e., the XGBoost classification model, the K-means clustering algorithm, and the weighted K-nearest neighbors (WKNN) algorithm. The methodology is divided into several stages. ① Analysis of wireless network environment. A thorough analysis of the wireless network environment in the assembly workshop is conducted to understand its characteristics and specific positioning requirements. ② Initial fingerprint database construction. The workshop is divided into functional areas, such as assembly, testing, and processing zones. Reference points are strategically placed throughout these areas, and the received signal strength indicator (RSSI) values are collected to create an initial fingerprint database. ③ Clustering and database refinement: K-means clustering is used to segment the fingerprint database into smaller, more manageable clusters. This step ensures that the RSSI data within each cluster are more homogeneous, improving the accuracy of subsequent localization steps. ④ Model training. The XGBoost classification model is trained using the refined fingerprint database. This model is responsible for coarse localization, identifying the sub-region where the target is located. ⑤ Fine localization. Within the identified sub-region, the WKNN algorithm is applied to achieve precise positioning. This two-tiered approach, combining coarse and fine localization, enhances both the accuracy and efficiency of the positioning process.

Results The proposed hierarchical WiFi positioning method was evaluated in a simulated workshop environment. The experimental results demonstrate significant improvements in both positioning accuracy and efficiency compared to traditional methods. Specifically, the method achieves a 143.82% improvement in positioning accuracy over conventional WKNN algorithms, with the average positioning error reduced to 0.89 m. Additionally, the average positioning time is shortened to 2.76 ms. Detailed performance metrics, including average positioning errors and cumulative distribution functions (CDF) for various algorithms, highlight the advantages of the proposed method. For instance, when positioning error thresholds are set at 3 m, the proposed method achieves a 100% success rate, significantly outperforming other methods such as RF and unclustered hierarchical algorithms. These results underscore the method's effectiveness in complex and dynamic workshop environments.

Conclusion This study introduces a novel WiFi-based hierarchical positioning algorithm specifically designed for the winding machine assembly workshop environment. By integrating XGBoost, K-means, and WKNN algorithms, the proposed method significantly enhances positioning accuracy and system performance. The hierarchical approach not only improves precision but also reduces the computational load, making real-time positioning feasible in large-scale industrial settings. The improvements in positioning precision and efficiency are particularly beneficial for industrial applications, where accurate and timely location data are critical for optimizing operations and ensuring safety. Future research should focus on testing the algorithm in more varied and complex real-world environments to further validate its robustness. Additionally, strategies to mitigate the impact of environmental changes on positioning accuracy should be explored. A comprehensive cost-benefit analysis of implementing this system in actual industrial settings would also provide valuable insights for practical deployment. Overall, this research contributes to the advancement of wireless network positioning technologies, offering a practical and efficient solution for the challenges faced in winding machine assembly workshops. The proposed method aligns with the goals of Industry 4.0, supporting the transition towards smarter, more automated manufacturing processes.

Key words: winding machine assembly workshop, wireless network, hierarchical positioning method, XGBoost classification model, K-means clustering algorithm, weighted K-nearest neighbors algorithm