Measurement Stability and Reference Monitoring

Why it matters

When a measurement system is examined on a single working day, it may give a good result. But contact tip wear, instrument warm-up, an environment change or deterioration of the reference can change the reading over time. A stability study asks whether the behavior of the measurement setup is maintained over time. For this reason the same reference is measured at regular intervals and under comparable conditions. The time order of the records is part of the analysis; calculating only the overall mean is not enough. Low readings on the first day and high readings on the last day can offset each other in a single mean. This monitoring is needed to tell whether a small increase seen in production results is a real process change or a drift in the measurement setup.

Daily repeat scatter and the change of level between days are different pieces of information. While one day's readings are very close to each other, the same closeness the next day may form around another mean. Repeatability may look good, but the center of the system is moving. Conversely, even if the daily mean stays in place, the scatter can grow. In that case the uncertainty of decisions increases. For this reason both the center and the spread are monitored in reference tracking. Looking at a single line on the same chart does not show the whole behavior. If the reading conditions have changed, the meaning of the difference on the series changes too. How the daily practice is carried out is as important as the numerical result; interventions such as cleaning or zeroing must be marked on the time axis.

Reference monitoring and the acceptance limits of the product are separate. The product tolerance defines the conformity of the product; the investigation threshold in the reference series flags a change in the measurement setup that needs to be investigated. This threshold must come from a justified control plan. The thresholds given in this lesson are predefined for training; they are not presented as if statistical control limits had been calculated. A trend over time can call for attention even while single points stay within the threshold. On the other hand, a single threshold exceedance does not directly prove that the instrument is faulty. The reference, the environment, recording errors and the application are investigated together. The meaning of the signal is not that a cause has been found, but that the normal decision flow is stopped and the evidence is examined.

Diagram: [Same reference] → [time-ordered daily center and spread] → [predefined signal] → [investigation of cause and impact] → [new verification]. Center and spread are tracked separately; a constant range does not cancel a signal in a rising center. The value of the monitoring plan appears when it is clear what will be done after a signal. Producing a chart alone does not manage measurement risk. Which measurements are affected, when the last reliable check was made and which product decisions will be reviewed must be defined. The check frequency is also chosen by risk: a long interval means a possible drift is noticed later. However, reading more often does not increase the reliability of the same faulty reference. The reference must be protected and, where necessary, independently verified. A stability result is valid for a specific time period, measurement conditions and reference. It is not a document that no change will ever occur in the future; sustained monitoring keeps the evidence current. All numerical scenarios in this lesson's short quiz are also covered by the following notice: This dataset was generated synthetically for training; it does not represent a real business, employee or customer study.

Formula