⚠️ Industry Standards vs. Textbook Specs (Why LSL/UCL aren't always Auto-Calculated)
In a purely academic textbook approach, Specification Limits are often just set to ±3 Standard Deviations of whatever your historical data happens to be. World-Class Industry Standards reject this logic.
Why? Because auto-calculating limits from past data only tells you what your process is currently doing. True Specifications (LSL/UCL) are dictated by Client Requirements, Engineering Tolerances, or Safety Margins—which dictate what the process must do.
For example, if you are manufacturing a bolt that must hold a bridge, the limit is defined by the physics of the bridge (Industry Standard), not by how poorly your machine performed last week (Textbook Data). You should manually set LSL/UCL to match the engineering target, and then use the simulator to see if your current variables can successfully meet that target!
1. Monte Carlo Simulation
A mathematical technique used to estimate the possible outcomes of an uncertain event. Instead of calculating a single result, it runs thousands of trials (iterations) using random values within defined bounds to create a probability distribution of possible results.
2. LSL & UCL (Specification Limits)
- LSL (Lower Specification Limit): The absolute minimum acceptable value for your output/process. Anything below is considered a defect.
- UCL (Upper Specification Limit): The absolute maximum acceptable value. Anything above is a defect.
3. Cp & Cpk (Process Capability)
- Cp (Potential): Measures how well your process could fit within the limits if it were perfectly centered. It compares the width of your data to the width of the limits. (Target: >1.33)
- Cpk (Performance): Measures how well your process is actually performing, taking centering into account. A low Cpk but high Cp means your process is tight, but off-center.
4. Distribution Types
- Gaussian (Normal): Bell curve. Most data clusters around the mean. Needs Mean (P1) and StdDev (P2).
- Uniform: Flat line. All values between Min and Max are equally likely. Needs Min (P1) and Max (P2).
- Triangular/PERT: Needs Min (P1), Most Likely/Mode (P2), and Max (P3). Good for project timelines.
5. Tornado Analysis (Sensitivity)
A bar chart that shows which input variables have the biggest impact on the final outcome. The variable at the top (widest bar) is the most critical to control.