Statistical Tolerancing and Process Control
Every company seeks to have their products work as predicted straight from design into manufacturing. For that to happen the product’s critical design parameters need to be manufacturable in the real world, so part tolerances need to be adequately assessed. This class explores a statistical approach to tolerancing using simulation, but we don’t stop there. In order to assess how well the method works, we need to understand Statistical Process Control—specifically two aspects of SPC, Capability and Control Charting—in order to assess tolerancing results. Lastly, to ensure product capability we must be able to trust our measurement systems. As such, Measurement Systems Analysis is covered.
Statistical Process Control
Any process (whether operational or transactional) is repeatable, so how can we ensure that it is under control and predictable? Statistical Process Control (SPC) does a great job of maintaining process control before the process drifts out of specification. In other words, SPC is a proactive approach to maintaining process control. In addition, given process drift, SPC will indicate what data signature has emerged, helping in the identification of root-cause. The core content of the course is basic Control Charting for both continuous and attribute data.
In Six Sigma, statistical experimentation is the way to understand the intricate relationships between inputs (causes) and outputs (effects). But what if we don’t have the luxury of experimentation, either because the data already exists (“big data)” and/or we simply cannot turn back the wheels of time to replicate in an experimental sense? This is the world of Predictive Analytics. With the advent of data-science methods and platforms, the power of performing predictive analytics is now in the hands of the practitioner, no longer the purview of the lone statistician. We use MINITAB and methods including Principle Components Analysis, Cluster Analysis, Decision-Tree Analysis, Factor Analysis, together with the more classic Analysis of Variance and Multiple Regression on the big-data analysis challenge.