Site Acceptance Testing: Data Signals and Condition Monitoring

The Purpose of SAT in a Live Warehouse Environment #

Site Acceptance Testing (SAT) is the bridge between installation and live operation. For a warehouse system—conveyor zones, sortation loops, lifts, palletizers, and their control networks—the test is less about proving that a device exists and more about proving that the millions of signals exchanged between mechanical parts, sensors, variable-frequency drives, and programmable logic controllers are accurate, repeatable, and correctly timed under real site conditions. SAT is also the first opportunity to capture a clean, structured condition baseline of every major moving component. That baseline becomes the reference point for all later predictive maintenance.

Three goals shape a meaningful SAT campaign:

• Prove the system meets contractual throughput and accuracy requirements under representative load and real product.
• Establish baseline health for motors, gearboxes, drives, sensors, and network devices so future anomalies can be detected as drift rather than discovered as failures.
• Expose integration issues between subsystems that were invisible during factory acceptance testing because site cabling, site power quality, ambient temperature, and physical vibration are different.

The remainder of this article explains how data signals and condition monitoring interact during SAT, what symptoms to look for, how to collect evidence, and where the decision boundaries lie.

Data Signals: What the Test Captures and Why #

Data signals are the language of the control system. During SAT, they are the primary evidence that the system is behaving as designed. A signal is not simply “on” or “off”; it has timing, edge quality, repeatability, and consistency under load. For warehouse automation, the signal families that matter most are:

• Digital inputs from photoelectric sensors, limit switches, pallet presence detectors, and mechanical position switches.
• Digital outputs to contactors, solenoid valves, indicator lamps, and audible annunciators.
• Analog signals from encoders, load cells, pressure transducers, temperature probes, and VFD analog output channels.
• Industrial network traffic such as EtherNet/IP or PROFINET tags, cyclic data, diagnostic counters, and timestamps.
• Scanner, camera, and RFID reads with payload data, retry counts, and read-quality metrics.

Each signal type requires a different acceptance discipline. A photoeye that turns on 4 ms late may still be logically correct but causes a carton to collide with a diverging gate. An encoder count that drifts only when a VFD runs near full speed is a noise or grounding problem, not an encoder problem. A scanner that reads a barcode 19 times out of 20 will fail only under the exact lighting and speed conditions of a particular packing station. For this reason, SAT must never be reduced to a blinking-light checklist. The test engineer and controls team must capture signal quality under loaded motion, not just signal presence at standstill.

It is also important to distinguish between signal presence and signal quality. Signal presence answers the question “did the input turn on?” Signal quality answers the questions “when did it turn on relative to the event?”, “did it stay stable?”, “what was its minimum pulse width?”, and “does it change with temperature or load?” A true signal with a 5 ms dropout can be invisible to a human observer but will cause a sorter to mis-divert a jumbo carton. A PLC trace sampling at 10 ms or faster is often the only tool that reveals this.

Condition Monitoring: Establishing Baselines During SAT #

Condition monitoring is often treated as a post-commissioning activity, but that is a missed opportunity. SAT is the ideal time to collect the first structured health measurements because the system is being loaded in a controlled way, with known product dimensions, known cycle counts, and limited operational pressure. The measurements captured during a properly instrumented SAT become the “as accepted” fingerprint of the machinery. Every future vibration reading, motor current trend, or drive temperature chart is compared against this fingerprint.

Useful baseline measurements to take during SAT include:

• Motor current at no-load, part-load, and full-load conditions, compared across identical motors within the same line.
• Vibration velocity and acceleration readings on motor bearings, gearbox input and output shafts, and sorter cams. The mounting point must be recorded because a reading from a housing and a reading from a stud-mounted accelerometer are not directly comparable.
• Drive and panel temperatures under sustained throughput, including VFD heatsinks and cabinet internal air temperature.
• Chain and belt tension values, or at minimum the acoustic signature of each drive, so later changes in tension can be identified.
• Encoder count repeatability over a defined number of cycles, including the variance from cycle to cycle.
• Thermal images of electrical panels, cable terminations, and busbars to highlight loose connections or undersized conductors.

Every baseline reading must be accompanied by context. Record the ambient temperature, relative humidity, product weight and size, conveyor speed setting, and the number of cycles completed prior to the reading. A motor current baseline taken at 08:00 with the warehouse at 18°C is not directly comparable to a reading taken at 15:00 with the warehouse at 33°C unless the thermal history is documented. Without context, the baseline is just a number and cannot support later analytical decisions.

Component Interactions That Influence Test Results #

Warehouse systems are not collections of isolated devices. Every component interacts with neighbours, and those interactions are precisely where SAT finds its most valuable evidence. A signal that is perfectly healthy in isolation can be the cause of a system failure because of how it interacts with a downstream component. Understanding these interactions before judging results prevents the classic error of replacing a sensor that was never at fault.

Consider a conveyor merge zone. A photoelectric sensor sees a gap between two cartons and sends a “clear” signal to the upstream belt. This is correct at slow speed. At higher speed, the previous carton may bounce off a stop and momentarily interrupt the light beam, creating a false “occupied” signal five milliseconds after the first carton has left. The interaction is between mechanical bounce and electrical timing, not between logic and the sensor. The SAT evidence must capture the complete sequence: the stop actuation, the bounce, the sensor output, and the downstream belt command.

Another common interaction involves encoders and VFDs. If the drive ramps from zero to full speed faster than the encoder filter allows, the PLC sees lost pulses and triggers a “run out” alarm even though the conveyor is moving correctly. The encoder is fine and the VFD is fine, but their response rates are incompatible. This type of fault appears only when the drive is under load with the actual mechanical inertia of the conveyor, so it rarely shows up during FAT on a test bench.

Sorter induction stations illustrate the interaction between electrical, pneumatic, and mechanical lags. The photoeye signal reaches the PLC at the correct time, but the solenoid valve is older or has a sticky spool, and the pneumatic cylinder actuates 12 ms later than it did at the start of the test. The signal is correct, the mechanical action is slow, and the sorter misses the diversion window. A condition monitoring log of the actuator response time over repeated cycles will reveal this long