Common Mistakes

Why Your Robot Cell Numbers Are Wrong: Common Mistakes and Fixes

The specific errors that make workcell throughput and cycle time projections miss reality, each with a symptom, root cause, and a fix you can measure.

The most common mistake is quoting theoretical cycle time as achievable throughput. Symptom: a cell rated at 30 picks per minute delivers 22 in production. Root cause: the ideal cycle from the Robot Cycle Time tool ignores dwell, part-present waits, and vision retries. A 1.2 second nominal cycle plus 0.3 seconds of gripper settle and 0.15 seconds of conveyor index is really 1.65 seconds, or 36 per minute dropping to 27 net. Fix: multiply theoretical rate by measured availability. If the cell runs 82 percent uptime, plan 0.82 times ideal, and validate against a Robot Pick Rate reading over a full shift, not a 60 second demo.

Unit errors on speed and acceleration wreck travel estimates. Symptom: the Robot Travel Time tool says 0.4 seconds but the stopwatch says 0.9. Root cause: people plug in the datasheet maximum joint speed of, say, 3.5 meters per second while the robot never sustains it over a 400 millimeter move. Short moves are acceleration limited, not velocity limited. A move under roughly 200 millimeters may never reach top speed at all. Fix: use a trapezoidal profile. With 20 meters per second squared acceleration, reaching 2 meters per second takes 0.1 seconds and 100 millimeters, so budget the ramp explicitly rather than assuming instant top speed.

Ignoring dynamic payload derating causes missed moves and faults. Symptom: a robot rated at 12 kilograms drops cycle speed or trips overload at full reach. Root cause: rated payload assumes the load sits near the wrist with a small offset. A gripper plus part at 250 millimeters center of gravity offset can cut effective capacity by 30 to 50 percent. Fix: check Robot Payload Utilization against the wrist load chart, not the headline number. If tooling weighs 4 kilograms and the part 5, that 9 kilograms at a long offset may exceed the derated 7 kilogram limit, forcing slower moves you never budgeted.

Skimping on reach margin creates singularities and jerky paths. Symptom: the arm stutters, replans, or faults near the edge of a station. Root cause: the layout uses 98 percent of nominal reach, but wrist singularities and joint limits eat the last 5 to 10 percent of the envelope. A 900 millimeter reach robot placed to work parts at 880 millimeters has almost no margin for tool tilt. Fix: keep Robot Reach Margin at 10 to 15 percent minimum. Position the base so the worst-case point sits at 850 millimeters or less, leaving room for approach angles and clearance moves.

Averaging path efficiency across dissimilar moves hides the slow one. Symptom: the cell looks balanced on paper but one station gates the whole line. Root cause: reporting a single Robot Path Efficiency figure of 85 percent masks a station running 60 percent because of a long over-and-around detour. Fix: measure each path separately. If actual path length is 1400 millimeters but the straight-line pick to place is 900, that path is 64 percent efficient and worth 0.2 to 0.4 seconds of re-teaching. Attack the worst path first, since in a serial cell throughput is set by the slowest station, not the average.

Confusing cell capacity with cell throughput leads to overpromising. Symptom: you told the customer 1,000 units per hour and delivered 720. Root cause: Robot Cell Capacity is the ceiling at 100 percent uptime and zero blocking. Real Robot Cell Throughput subtracts starving, blocking, changeover, and micro-stops. A cell with 1,000 unit capacity, 88 percent availability, and 4 percent quality loss nets about 845, and upstream starvation can pull it to 720. Fix: quote throughput at demonstrated OEE, typically 60 to 75 percent for a new cell, and hold capacity as headroom, not as the commitment.

Forgetting dwell time as an independent variable inflates rate claims. Symptom: adding a dispense or cure step tanks a rate that math said was fine. Root cause: dwell is fixed process time the robot cannot compress. A 0.8 second glue cure dwell on a 1.5 second cycle is 35 percent of the loop, and no faster arm removes it. Fix: pull dwell out with the Robot Dwell Time tool and treat it as a floor. If dwell is 0.8 seconds, the theoretical best cycle is 0.8 plus motion, so consider parallel tooling or a second station to hide the dwell rather than chasing arm speed you cannot use.

Sampling too short is the data mistake behind all the others. Symptom: acceptance numbers look great, then week-one production disappoints. Root cause: a 5 minute buyoff never captures the once-per-300-cycle vision reject, the tote change, or the thermal drift that slows Robot Arm Speed as motors warm. Fix: base every number on at least a 2 hour continuous run, ideally a full 8 hour shift, and log faults per thousand cycles. A cell that faults 3 times per 1,000 cycles at 40 second recovery loses about 2 percent availability that a short sample will never show.

Published 2026-07-01.