Preventive and Predictive Maintenance and Troubleshooting
Two different jobs
Preventive maintenance is calendar or running-hour based - lubrication, filter changes, inspections, belt checks. Predictive (condition-based) maintenance acts on measured condition: vibration, thermography, oil analysis, ultrasound and performance trending. A good program uses both and lets condition data stretch or shorten the calendar interval.
Reading the signals
Vibration is measured as overall velocity in mm/s RMS against a zone standard, with direction (horizontal, vertical, axial) recorded each time so trends are comparable. Frequency content identifies the fault: 1× rpm suggests unbalance, 2× suggests misalignment or looseness, high-frequency energy with harmonics points to rolling element defects, and a once-per-revolution spike on a belt drive suggests belt or sheave trouble. Thermography finds hot connections and failing bearings; oil analysis reports wear metals, water and particle counts; ultrasound detects early bearing distress, leaks and electrical arcing.
Troubleshooting method
Confirm the symptom, gather data, form a hypothesis, test it cheaply, then correct the root cause. Tools such as 5 Whys and fishbone diagrams keep the analysis honest. Replacing a failed part without finding why it failed guarantees a repeat.
Worked example
A pump shows vibration rising over weeks at 1× rpm, highest in the horizontal plane, and the trend accelerates after a coupling change. Unbalance or misalignment is the leading hypothesis; check alignment and runout before touching the impeller.
Traps
Judging a machine by one reading instead of a trend; comparing measurements from different points or directions; assuming a hot bearing is always a lubrication problem.