check
check
check
check
check
check
An inclination sensor "drifts" on a vibrating machine when the vibration is coupled into the measurement instead of being rejected: a mounting bracket that resonates or flexes converts machine vibration into real tilt at the sensor; vibration components alias into the slow tilt reading when sampling and filtering do not separate them; the sensitive axis is misaligned so machine vibration projects onto it; temperature shifts add genuine drift; and sometimes the reading is honest — the structure actually tilts under load. The first safe check is the static-versus-dynamic comparison: read the sensor machine-stopped versus machine-running at the same nominal attitude, and on a rigid reference surface if possible. Never mount, remount or wiring-work a sensor on a running machine — stopped-and-locked-out work only.
Vibration must be rejected, not averaged away. Averaging a vibration-corrupted reading reduces noise but biases the tilt value; the fix is coupling (mounting) and filtering designed for the vibration spectrum — with sampling that avoids aliasing.
The mounting is part of the measurement. A bracket that flexes or resonates makes the sensor measure its own motion; stiff, short, direct-to-structure mounting is the baseline, and the sensor's own filtering (e.g., the fifth-order filtering documented in KJT Sensors' inclination products) is the second line.
Axis alignment decides what the sensor sees. A single-axis sensor mounted a few degrees off its intended plane projects machine vibration onto the sensitive axis — apparent drift that vanishes with correct alignment.
Static-vs-dynamic is the split. Same attitude, machine stopped vs. running: identical readings → drift is thermal or electronic; different readings → vibration coupling or real structural tilt under load.
Don't recalibrate to hide a coupling fault. Zeroing a vibration-biased reading corrupts the static accuracy; fix the coupling, then zero if the model provides it and the zero state is real.
