Abstract
Statistical Process Control (SPC) is a cornerstone of quality management in production environments. However, traditional Shewhart control charts often lack the sensitivity required to detect small process shifts, particularly in customized manufacturing contexts like Electronic Power Steering (EPS) production. This paper presents a weighted optimization approach to SPC that prioritizes recent data to improve responsiveness. By implementing this method within a 3-layer Client/Server software architecture, the system enables real-time monitoring and earlier detection of process deviations compared to standard methods.
The Challenge of Traditional SPC
Traditional SPC control charts, such as the $\bar{X}-R$ chart, have been industry standards for nearly a century. Despite their longevity, they possess a significant weakness: they treat all historical data points with equal importance. In modern manufacturing—where production is shifting toward smaller batches and highly customized parts—these traditional charts are often too "blunt." They struggle to detect small deviations (referred to as "small warps") and fail to leverage historical trends effectively. This lack of sensitivity can lead to missed opportunities to correct a process before it produces defective parts.
Methodology: Weighted Optimization
The core innovation of this research is the Weighted Optimization SPC Control Chart. This approach introduces a weight assignment ($W_i$) to every variable, where recent data points are given larger weights and older data points are given smaller weights. This is achieved through a decay factor $\beta$ (between 0 and 1), creating a sliding scale of importance that maintains statistical validity while emphasizing current performance.
By calculating weighted statistics for the mean ($\bar{X}$), variance ($S^2$), and standard deviation ($S$), the system establishes dynamic Upper and Lower Control Limits (UCL and LCL). These limits become more responsive to recent trends, effectively narrowing or shifting the "envelope" of acceptable performance based on the direction of the data.
Software Realization
To bring this theoretical model into a production environment, the authors developed an SPC control system using a 3-layer Client/Server (C/S) architecture. Built on the .NET framework with a SQL Server 2008 backend, the system is organized into distinct modules for resource management, test definition, and data processing. This architecture ensures the system is capable of real-time data processing and responsive visualization.
Comparative Analysis and Results
The effectiveness of the weighted optimization was tested using 20 sets of output voltage data from torque sensors on an actual EPS production line. A direct comparison was made between the traditional $\bar{X}-R$ chart and the new weighted chart.
In the traditional analysis, the process appeared to be in a state of statistical control, with no alarms triggered as the points remained within the static control limits.
However, the Weighted Optimization chart revealed a different story. By mathematically encoding a "recency bias," the control chart became a predictive tool rather than just a reactive one. It successfully identified an "additional trend" in the final data points that the traditional method ignored.
As shown in Figure 2, the 20th data point was flagged as "out of control" by the weighted limits. This demonstrates that the weighted approach is significantly more sensitive to process shifts, allowing for proactive intervention before the process drifts into a failure state.
Conclusion
The Weighted Optimization SPC Control Chart offers a robust solution for modern manufacturing challenges. By mathematically prioritizing recent data and implementing the logic within a responsive software framework, enterprises can achieve a higher level of quality assurance. Future work will focus on refining the selection of the decay factor $\beta$ to minimize false alarms while maintaining high sensitivity across various manufacturing cycles.
Summary updated