ORIGINAL COMPANY-CONTEXT PRACTICE
Separate sensor drift from process change
Original practice scenario inspired by Siemens Aktiengesellschaft's public business context; this is not a reported interview question or a description of its internal systems. A synthetic industrial dashboard shows a gradual rise in one sensor's readings. Production conditions and maintenance schedules also changed. How would you investigate whether the sensor is drifting?
Written for independent practice using company context; this is not a verified question from a company interview.
A REASONED APPROACH
Model answer.
A trend in readings does not by itself distinguish an instrument problem from a real process change. Use reference measurements, operating context, and maintenance events to compare competing explanations.
Work through it
- Check units, calibration history, sampling frequency, and missingness. Plot the sensor alongside related measurements under comparable operating regimes.
- Compare residuals against a trusted reference or a model fitted on stable periods, while keeping evaluation periods separate. Inspect changes around maintenance and equipment replacement.
- Quantify uncertainty and investigate flagged periods with domain operators. Set an alert rule with persistence and review criteria rather than automatically recalibrating from the trend alone.
Tradeoffs to discuss
- Reference instruments add cost, and model residuals can also drift when the process changes. Combining independent evidence is stronger than one statistical test.
Common pitfalls
- Normalizing every period independently can erase genuine drift.
- Correlated sensors sharing the same failure mode are not independent confirmation.