We measure the deepest vacuum the pump can pull with the inlet closed. For rotary vane pumps, this is typically 0.5-1.5 Pa. For claw pumps, around 20 Pa. For roots boosters, we measure with a backing pump in the loop. We use a calibrated Pirani gauge plus a capacitance manometer for cross-verification. The pump must hit its rated ultimate within a specified time window — usually 20-30 minutes from cold start. If it does not, we look for leaks, check the oil, check the seals, and rerun. If it still fails, it gets torn down.

We measure flow rate (m³/h or L/s) at the rated inlet pressure using a mass flow meter. This is the number that tells you how fast your chamber gets pumped down. We test at multiple pressure points — not just the rated point — to generate a rough speed curve. If a pump's speed is off by more than 5% from the rated value, we investigate. Usually it is a minor assembly issue. Occasionally it is a machining tolerance problem. Either way, it gets fixed before the pump ships.

We measure the motor's current draw under load. If it is too high, something is binding — a misaligned rotor, a tight bearing, an assembly error. If it is too low, the pump might not be compressing properly — perhaps a vane is stuck or a clearance is too large. We set upper and lower limits for each model based on historical data. Readings outside those limits trigger a teardown.

Raw materials, castings, bearings, seals, motors — everything that enters the factory gets inspected before it reaches the production floor. We have a sampling plan (AQL 1.0 for critical dimensions, AQL 2.5 for non-critical) and a documented procedure for what happens when a batch fails. We have rejected entire shipments of cast iron for porosity. We have sent back motors that did not meet nameplate efficiency. The supplier does not like it. The pump does.

During machining and assembly, QC technicians perform dimensional checks at defined stages. Critical dimensions — rotor bore diameter, shaft runout, end plate flatness — are checked on every part, not sampled. Non-critical dimensions are sampled per the control plan. If a dimension is trending toward the tolerance limit, the production team gets notified before it goes out of spec.

When something fails — a test, an inspection, a customer complaint — we open a nonconformance report. Root cause analysis gets done. Corrective action gets implemented. Follow-up verification happens within 30 days. This is not a paperwork exercise. We track defect categories and look for patterns. If the same issue appears twice in a quarter, the engineering team gets involved.

Roughly 2-3% of pumps fail initial testing and require rework before they ship. Most failures are minor: a slightly high noise reading, a vacuum that takes too long to stabilize, a current draw at the edge of the limit. About 0.5% are major failures that require a teardown. We track this monthly. If the rate trends upward, we investigate.

Under 1% across all product lines over the last three years. When a pump comes back, we do a full failure analysis. The result gets shared with the customer whether or not it is a warranty claim. Common root causes we have found and fixed: inadequate rust prevention on sea freight shipments (added VCI packaging), rotor imbalance at high hours (changed balancing spec), and oil leakage during air freight at altitude (redesigned the breather plug).

We rate our suppliers quarterly on delivery performance and quality. Bearings and seals are our highest-risk purchased parts — a bad bearing can kill a pump in weeks. We have dropped two bearing suppliers in the last five years for quality issues. We now dual-source bearings from a Japanese and a German manufacturer. It costs more. The failure rate dropped from 0.4% to under 0.05%. Worth it.

Eight test benches in a row. Each bench is sound-insulated with acoustic foam panels. The benches are numbered 1 through 8. Bench 4 is the one we use for screw pumps — it has the largest vacuum chamber and the highest-capacity backing pump.

Calibrated Pirani and capacitance manometer gauges. We cross-check readings between gauge types. If the Pirani says 0.8 Pa and the capacitance manometer says 1.2 Pa, we stop testing and recalibrate. Gauge disagreement is the most common source of false failures.

The data logging rack. Every test bench feeds into a central system that records the vacuum curve, speed, current, noise, and vibration in real time. Test records are archived digitally with the pump serial number. Backup is weekly.