What data do you need before deploying predictive maintenance?
Before deploying predictive maintenance you need three things. First, failure history: at least 30 to 50 documented failure events per failure mode, with timestamps and sensor readings at the time of failure. Second, a live sensor feed from the asset at a minimum of 10 Hz for vibration or 1 Hz for temperature and current signature. Third, 4 to 6 weeks of healthy baseline operation so the model can learn what normal looks like. Most Indian plants lack the failure history, which is why every engagement starts with a 4-week data audit.
Talk to an engineer about thisWhy failure history is the hardest requirement
A predictive maintenance model learns to distinguish pre-failure signatures from healthy signatures. For it to learn, it needs examples of both. Healthy data is easy: run the machine normally for a few weeks with sensors attached. Failure data is hard: most plants do not record sensor readings at the moment of failure, or their CMMS records the failure date but not the sensor readings leading up to it.
If your plant does not have 30 to 50 documented failure events with sensor context per failure mode, you have two options. First, run the machine with sensors attached and wait for failures over 6 to 12 months: this is accurate but slow. Second, use physics-based models to simulate pre-failure signatures and augment the real data: this is faster but less accurate. We recommend the first approach for safety-critical assets and the second for high-frequency, low-consequence failures.
What sensors you need for common failure modes
Bearing failure (the most common rotating-machine failure): triaxial accelerometer at 10 kHz minimum, mounted on the bearing housing. A vibration envelope spectrum detects bearing defect frequencies at 2 to 8 weeks before failure.
Motor winding failure: current transformer on all three phases sampling at 1 kHz minimum. Motor Current Signature Analysis (MCSA) detects rotor bar defects, eccentricity and winding faults.
Heat exchanger fouling and pump cavitation: pressure differential across the exchanger sampled at 1 Hz; flow rate and outlet temperature. The rate of fouling is predictable from first-principles thermodynamics once baseline is established.
What to do if you do not have the data yet
Start with a 4-week pilot on 5 to 10 of your highest-criticality assets. Install sensors, establish connectivity to a time-series database (InfluxDB, TimescaleDB, or AWS Timestream), and begin logging. During this period, run a data quality audit: check for sensor dropouts, timestamp irregularities and calibration drift.
At the end of 4 weeks, assess whether you have enough signal to train even a simple anomaly detection model (Z-score or Isolation Forest). If yes, deploy the model and collect labelled failure events as they occur. If no, extend the pilot and revisit the sensor placement.
Related questions
Ajinkya Technologies deploys predictive maintenance on rotating equipment across steel, aluminium and automotive plants in India. The data audit is the starting point for every engagement. Contact hello@ajinkyatechnologies.in to discuss your asset list.
References and scope
External references explain the underlying technology or regulatory context. Ajinkya pricing, schedules and project results are company provided estimates or examples, not independently audited industry benchmarks. Confirm scope and current requirements before making a purchasing decision.
What would this look like at your plant?
Share your production process, current systems and the problem you want to solve. An engineer, not a salesperson, replies.
Step 1
An engineer reads it
Not a sales bot. Our team replies within 24 hours with questions or times for a call.
Step 2
30 minute call
Your process, machines, systems and the number you want to move. We tell you honestly if we are a fit.
Step 3
A written scope for one line
What we would connect, what you would see, how long it takes. You decide; nothing starts without your approval.
Free scoping call. We start with one line, never change your machine logic, and you own the code and the data.
Reviewed by Amey Kadle, Founder, Ajinkya Technologies. Last reviewed: .