A machine that is running is not necessarily a machine that is performing. A press cycling at 72% of its rated speed has a performance loss that does not appear as a stoppage in any manually maintained downtime log. A machine monitoring system that captures only binary running/stopped states misses the performance component of OEE entirely, which is why comparing systems on the basis of method matters before making a purchasing decision.
This guide covers the three main methods for machine monitoring, the metrics each method can and cannot capture, and realistic deployment cost ranges for different scales.
What is a machine monitoring system?
A machine monitoring system is hardware and software that tracks the operational state of production machinery continuously, aggregates that state data into performance metrics, and surfaces it to production management in useful time frames. The critical phrase is “in useful time frames”: a system that delivers yesterday’s machine data as this morning’s report is not a monitoring system, it is a reporting system.
Genuine monitoring delivers state data within minutes of events occurring. Useful thresholds:
- Machine state change: visible in dashboard within 2 minutes
- Downtime alert: delivered to responsible supervisor within 5 minutes
- Shift-level OEE: available without manual calculation at any point during the shift
Method 1: PLC and SCADA integration
PLC integration connects the monitoring system directly to a machine’s programmable logic controller, extracting state data, cycle counts, alarm codes, and in some cases sensor readings directly from the machine’s control system.
What it captures: Machine state, cycle count, cycle time, alarm codes, throughput, and sometimes temperature, pressure, and torque data depending on what the PLC exposes.
What it misses: Process compliance (operator behavior), material flow between machines, and any machine without a PLC (older equipment, sub-contract assemblies, manual stations).
Deployment timeline: 2-6 weeks per machine type for PLC mapping and integration. A 40-machine floor with three machine generations and two control system vendors typically takes 4-8 months to fully deploy.
Typical cost range: Rs 15-40 lakh for a 20-machine deployment including hardware, integration, and first-year support. Higher for older equipment requiring protocol translation.
Method 2: Sensor-based monitoring
Sensor-based systems attach current transducers, vibration sensors, temperature probes, or optical sensors to machines to infer state from physical signals. A current transducer on a motor circuit detects whether the motor is drawing power; vibration sensors detect whether a machine is in motion.
What it captures: Machine running state (reliably), cycle count (for discrete processes with clear cycle signatures), vibration trends for predictive maintenance, and temperature patterns.
What it misses: Operator activity, process compliance, quality outcomes, and cycle time variation for processes without strong physical signatures.
Deployment timeline: 1-3 days per machine for sensor installation. A 40-machine floor can be instrumented in 4-8 weeks including calibration.
Typical cost range: Rs 8-25 lakh for a 20-machine deployment. Lower than PLC integration but sensors require periodic calibration and replacement.
Method 3: Camera-based AI monitoring
Camera-based systems use computer vision models to classify machine state, count cycles, detect process deviations, and observe operator activity from video feeds. When deployed on existing CCTV infrastructure, the camera hardware cost is eliminated.
What it captures: Machine state, cycle count, process compliance, operator activity, changeover sequence adherence, and material flow visibility between stations.
What it misses: Internal machine signals (temperatures, pressures, torque) that are not externally observable. For predictive maintenance based on internal signals, sensor augmentation is necessary.
Deployment timeline: 4-8 weeks for a 40-machine floor using existing cameras. Faster than PLC integration; comparable to sensor deployment but without per-machine hardware installation.
Typical cost range: Rs 6-20 lakh for a 20-machine deployment on existing camera infrastructure. The largest cost variable is whether cameras need repositioning or supplementation.
What deployment costs depend on
Three factors drive deployment cost more than the method itself:
Equipment age and standardisation. Newer, standardised equipment with modern PLC communication protocols costs less to integrate than a mixed-vintage floor where every machine requires individual assessment.
Camera coverage gaps. Camera-based deployment costs rise when significant coverage gaps require new camera installation rather than reuse of existing infrastructure.
Integration with existing systems. Connecting monitoring output to MES, ERP, or quality management systems adds cost regardless of monitoring method.