Variable Sampling Plan Implementation for High Capacity Prismatic Cell Shipments
Variable sampling under ANSI/ASQ Z1.9 cuts incoming prismatic cell sample sizes by 80 percent while upholding strict AQL protection via precise continuous parametric data.

Dock
Shipments of large-format lithium iron phosphate and nickel manganese cobalt prismatic cells arrive at receiving facilities in multi-pallet ocean containers carrying thousands of individual units. Incoming quality control for high-capacity cells, such as 280 Ah, 302 Ah, and 314 Ah formats, presents a physical and statistical bottleneck if managed purely through attributes sampling. Attributes inspection plans, governed by ISO 2859-1 or ANSI/ASQ Z1.4, evaluate units on a pass-fail basis.
For a lot size of 40,000 cells at General Inspection Level II, an attributes plan under an Acceptance Quality Limit of 0.65 percent demands a sample size of 500 cells. Discharging, charging, and measuring internal resistance across 500 individual 300 Ah class cells burns hundreds of channel-hours on expensive test equipment and delays cell clearance for days.
Sample size determines buyer risk.
Implementing a variable sampling plan compliant with ANSI/ASQ Z1.9 or ISO 3951 changes this economic calculation. Variables sampling utilizes the measured numerical values of continuous parameters, including alternating current internal resistance at 1 kHz, direct current internal resistance, open-circuit voltage, K-value self-discharge rates, and absolute cell capacity. Because continuous parametric data yields significantly higher statistical information per sampled unit than binary pass-fail data, a variable sampling plan achieves identical Statistical Protection and Quality Limit enforcement with sample sizes that are typically 60 to 80 percent smaller than their attributes equivalents.
For 314 Ah prismatic cells tested under ambient conditions of 25 degrees Celsius, switching from ANSI/ASQ Z1.4 attributes inspection to ANSI/ASQ Z1.9 variables inspection reduces sample sizing from 500 units to 35 units while maintaining an Acceptance Quality Limit of 0.40 percent.
The operational decision to transition receiving inspection to a variables framework requires structuring defects into distinct risk classes based on cell mechanics and electrical failure modes. Critical parameters that involve cell safety or immediate physical non-conformance remain under zero-acceptance attributes criteria, whereas continuous performance parameters move to variable statistical evaluation.
- Critical Defect Parameters Safety hazards including hermetic seal leaks detected via helium mass spectrometry, mechanical housing dents exceeding 0.5 mm depth, terminal tab thread damage, and active insulation resistance lower than 100 Megaohms at 1,000 Volts direct current.
- Major Variable Parameters Electrical performance limits encompassing 1 kHz AC internal resistance upper specification limits, 30-day K-value voltage decay rate expressed in millivolts per day, and discharge capacity below nominal rated Ampere-hours under 0.5C rate at 25 degrees Celsius.
- Minor Variable Parameters Dimensional envelope metrics, including cell body thickness under 300 kilogram-force face compression, total mass variance, and terminal height offset across the top cover plane.
Establishing the boundary between lot size and sample allocation requires aligning factory batching with inbound logistics documentation. When a supplier ships 40,000 cells across eight distinct manufacturing sub-lots, evaluating the shipment as a single master lot invalidates the statistical variance calculation if sub-lots originated from different electrode coating lines or formation aging tanks.
| Lot Size (Cell Units) | Attributes Sample Size (ANSI/ASQ Z1.4) | Variables Sample Size (ANSI/ASQ Z1.9) | Testing Time Reduction (Channel-Hours) | Statistical Basis |
|---|---|---|---|---|
| 1,201 to 3,200 | 125 | 15 | 88.0% | Normal Distribution / Standard Deviation Unknown |
| 3,201 to 10,000 | 200 | 25 | 87.5% | Normal Distribution / Standard Deviation Unknown |
| 10,001 to 35,000 | 315 | 35 | 88.9% | Normal Distribution / Standard Deviation Unknown |
| 35,001 to 150,000 | 500 | 50 | 90.0% | Normal Distribution / Standard Deviation Unknown |
Applying an attributes plan to high-capacity cell shipments either forces excessive inspection expenditure or drives buyers toward loose sampling levels that allow defective cell clusters to slip into module assembly lines.

Normality
Parametric acceptance sampling rests on the mathematical assumption that the inspected characteristic follows a Gaussian distribution within the incoming lot. Cell capacity exhibits normal dispersion. Internal resistance across automated electrolyte injection and formation lines also follows a symmetric Bell curve when manufacturing steps remain in statistical process control.
If a supplier blends cells from two separate production lines or combines different aging lots to fulfill a single purchase order, the resulting distribution becomes bimodal. Applying standard variable calculations to a bimodal or heavily skewed distribution severely distorts the calculated quality index, leading to incorrect acceptance of sub-standard lots or false rejection of compliant shipments.
Skewed distributions invalidate Z1.9 plans.
Verification of population normality forms the mandatory prerequisite step before executing the variable calculation algorithms of ANSI/ASQ Z1.9. The receiving inspector measures the target parameter across all units in the sample and evaluates the data set using the Shapiro-Wilk test or the Anderson-Darling test at a significance level of alpha equals 0.05. When the sample data rejects the null hypothesis of normality, the inspection team cannot use standard variable formulas.
In such cases, the lot must either be evaluated under a non-parametric variable procedure, transformed via Box-Cox power transformations, or reverted to an attributes sampling plan for that specific inspection lot.
- Draw a random sample of n units from the designated shipment lot according to the sample size code letter determined by lot volume.
- Measure the continuous variable for each sample unit under controlled ambient temperature and standardized fixture mechanical loading.
- Calculate the sample mean by summing all measured values and dividing by the sample count n.
- Calculate the sample standard deviation s using the unbiased sample variance formula with n minus one degrees of freedom.
- Compute the Upper Quality Index QU by subtracting the sample mean from the Upper Specification Limit USL and dividing by the sample standard deviation s.
- Compute the Lower Quality Index QL by subtracting the Lower Specification Limit LSL from the sample mean and dividing by the sample standard deviation s.
- Estimate the lot percent non-conforming by cross-referencing QU and QL into the standard Z1.9 distribution tables.
- Compare the total estimated percent non-conforming against the Maximum Allowable Percent Non-conforming M corresponding to the contractual AQL.
Distribution symmetry across cell internal resistance readings indicates stable factory formation processes and confirms the statistical validity of reduced sample lot acceptance.
When measuring parameters with two-sided specification limits, such as cell thickness under pre-charge compression or 1 kHz AC impedance, both upper and lower boundaries must be evaluated concurrently. A cell format with nominal AC internal resistance of 0.18 mOhm may carry an Upper Specification Limit of 0.23 mOhm and a Lower Specification Limit of 0.13 mOhm. An abnormally low internal resistance reading often signals internal micro-shorting or separator degradation rather than superior electrochemical performance.
The double specification variable procedure accounts for non-conformance risks at both extremes of the measurement spectrum.
Parametric data collected across multiple consecutive lots provides the historical standard deviation needed to transition from the Standard Deviation Unknown method to the Standard Deviation Known method, which further reduces required sample counts by approximately forty percent.

Crate
Uncrating large prismatic cells for bench top testing introduces physical measurement errors if environmental and mechanical boundary conditions are not strictly governed. High-capacity aluminum-sheathed cells experience dimension changes based on state of charge and internal gas pressure. Electrical contact resistance at the copper and aluminum terminal busbar surfaces varies by orders of magnitude depending on terminal clamping force and surface oxidation layers.

What Distinguishes Variables Sampling from Attributes Plans?
Attributes plans record only whether a cell passes or fails a set threshold, requiring large sample volumes to establish statistical confidence. Variables plans record precise continuous numerical values, extracting higher statistical confidence per unit and enabling smaller sample sizes. This distinction shifts the burden from sheer testing volume to high-precision metrology and fixture repeatability.
Voltage drop reveals micro-shorts.
Temperature shifts bias voltage readings. A 314 Ah cell stabilized at 15 degrees Celsius exhibits a higher internal resistance and lower open-circuit voltage than the same cell measured at 25 degrees Celsius. To maintain data integrity across variable sampling runs, cells pulled from shipping containers must sit in a climate-controlled room at 25 degrees Celsius plus or minus 2 degrees for a minimum of twelve hours before electrical measurement.
Fixtures applying face compression to prismatic cells during thickness measurement must enforce a uniform clamp force of 300 kilograms-force plus or minus 10 kilograms-force across the broad faces of the cell casing to replicate pack enclosure pressures.
- Thermal Stabilization Protocol Soak sampled cells in a temperature-controlled environment until internal thermal equilibrium reaches 25 degrees Celsius within a tolerance of 1 degree Celsius before taking voltage or resistance readings.
- Terminal Contact Load Apply a calibrated 4.0 Newton-meter torque to spring-loaded four-wire Kelvin sensing probes during impedance measurements to eliminate contact resistance variability across bare aluminum and nickel-plated copper terminals.
- Volumetric Compression Fixtures Measure cell thickness using a rigid dual-platen pneumatic fixture equipped with digital linear variable differential transformers accurate to 0.005 millimeters under a standardized face pressure of 300 kilograms-force.
- Decay Timing Controls Record open-circuit voltage measurements using a 6.5-digit digital multimeter exactly 120 seconds after removing the cell from charging or discharging pulses to standardize electrochemical relaxation dynamics.
| Parameter | Measurement Unit | Lower Spec Limit (LSL) | Upper Spec Limit (USL) | Assigned AQL (%) | Test Instrument Requirement |
|---|---|---|---|---|---|
| AC Impedance (1 kHz) | mOhm | 0.130 | 0.230 | 0.40 | 4-Wire AC Milliohmmeter (±0.001 mOhm) |
| DC Internal Resistance (10s 1C) | mOhm | 0.300 | 0.450 | 0.65 | High-Current DC Electronic Load (±0.5%) |
| Open-Circuit Voltage (50% SOC) | Volts | 3.2850 | 3.3050 | 0.25 | 6.5-Digit Digital Multimeter (±0.005%) |
| K-Value (Self-Discharge) | mV/day | 0.000 | 0.800 | 0.40 | Dual-Point OCV Delta over 14 Days |
| Cell Thickness (Compressed) | mm | 71.30 | 72.20 | 1.00 | LVDT Automated Gauge (±0.01 mm) |
| Discharge Capacity (0.5C) | Ah | 314.0 | 330.0 | 0.40 | Regenerative Cycler Channel (±0.1%) |
Standard commercial purchase agreements specify that incoming acceptance testing conducted outside specified thermal and mechanical tolerances voids buyer rejection claims against cell manufacturers.
Cell suppliers frequently defend incoming lot failures by asserting that receiving warehouse thermal gradients skewed open-circuit voltage measurements or that uncalibrated Kelvin probes induced artificial impedance spikes.

Arithmetic
Execution of a variable sampling inspection for a incoming shipment of 40,000 units of 314 Ah LFP prismatic cells requires working through the full standard deviation method under ANSI/ASQ Z1.9, Inspection Level II, Normal Single Sampling with an AQL of 0.40 percent for AC internal resistance. The specification limits for AC internal resistance are set at a Lower Specification Limit of 0.130 mOhm and an Upper Specification Limit of 0.230 mOhm. From the ANSI/ASQ Z1.9 sample size code tables, a lot size of 40,000 cells corresponds to Sample Size Code Letter N, which mandates a sample size n of 35 cells.
Sample size determines buyer risk.
The inspector selects 35 cells at random from across the pallets using a stratified random sampling pattern. Each cell sits in the thermal equilibration chamber at 25 degrees Celsius for 12 hours. The AC internal resistance of each cell is measured at 1 kHz using a calibrated 4-wire Kelvin fixture.
The resulting 35 discrete data points, expressed in milliohms, are compiled to execute the statistical decision algorithm.
| Sample Unit | Measured Impedance (mOhm) | Sample Unit | Measured Impedance (mOhm) | Sample Unit | Measured Impedance (mOhm) |
|---|---|---|---|---|---|
| 1 | 0.172 | 13 | 0.178 | 25 | 0.181 |
| 2 | 0.168 | 14 | 0.185 | 26 | 0.174 |
| 3 | 0.181 | 15 | 0.169 | 27 | 0.188 |
| 4 | 0.175 | 16 | 0.191 | 28 | 0.179 |
| 5 | 0.189 | 17 | 0.173 | 29 | 0.171 |
| 6 | 0.166 | 18 | 0.182 | 30 | 0.195 |
| 7 | 0.194 | 19 | 0.176 | 31 | 0.167 |
| 8 | 0.177 | 20 | 0.184 | 32 | 0.183 |
| 9 | 0.183 | 21 | 0.170 | 33 | 0.186 |
| 10 | 0.171 | 22 | 0.187 | 34 | 0.172 |
| 11 | 0.190 | 23 | 0.179 | 35 | 0.180 |
| 12 | 0.184 | 24 | 0.192 | – | – |
Calculating the sample mean yields a value of X-bar equals 0.1800 mOhm. Summing the squared deviations from the mean across all 35 observations and dividing by 34 gives a sample variance of 0.0000624 mOhm squared, taking the square root yields a sample standard deviation s of 0.00790 mOhm.
With X-bar and s established, the inspector calculates the Upper Quality Index QU and Lower Quality Index QL:
QU = (USL – X-bar) / s = (0.230 – 0.1800) / 0.00790 = 0.0500 / 0.00790 = 6.329
QL = (X-bar – LSL) / s = (0.1800 – 0.130) / 0.00790 = 0.0500 / 0.00790 = 6.329
Impedance spikes alter thermal margins.
To evaluate lot acceptability under ANSI/ASQ Z1.9 Table B-5, the calculated Quality Indices QU and QL are converted to estimated lot percent non-conforming values PU and PL. A Quality Index value of 6.329 far exceeds the minimum required Quality Index value for Code Letter N at AQL 0.40, which specifies a Maximum Allowable Percent Non-conforming M equal to 1.33 percent. The estimated percent non-conforming for both upper and lower limits is effectively 0.00 percent.
Because the total estimated percent non-conforming (P = PU + PL = 0.00%) is less than or equal to M (1.33%), the shipment of 40,000 cells passes variable inspection for AC internal resistance.
Assume now a second scenario where the sample standard deviation s broadens to 0.0220 mOhm due to inconsistent laser welding of internal current collector tabs on the cell manufacturing line. The sample mean remains 0.1800 mOhm.
QU = (0.230 – 0.1800) / 0.0220 = 0.0500 / 0.0220 = 2.272
QL = (0.1800 – 0.130) / 0.0220 = 0.0500 / 0.0220 = 2.272
Cell capacity exhibits normal dispersion.
Cross-referencing a Quality Index of 2.27 with ANSI/ASQ Z1.9 Table B-5 for sample size n equals 35 yields an estimated percent non-conforming PU of 0.98 percent and PL of 0.98 percent. Summing these values gives a total estimated lot percent non-conforming P of 1.96 percent. Because P (1.96%) exceeds the Maximum Allowable Percent Non-conforming M (1.33%), the lot fails acceptance criteria, triggering immediate formal quarantine.
Section 8.2 of standard supply master agreements establishes that when an incoming lot fails variable acceptance limits, the buyer retains the contractual right to reject the entire shipment container without incurring demurrable storage costs or return freight liabilities.

Rejection
A failed variable sampling plan initiates a strict commercial disposition sequence. The buyer issues a formal Notice of Non-Conformance within five business days of container receipt, appending the parametric raw dataset, environmental log files, and calculated Quality Index derivations. The cell manufacturer receives duplicate retain samples pulled during the initial quarantine sampling to execute parallel verification testing at an accredited third-party laboratory.
Quarantine blocks non-conforming lots.
Re-testing failed lots masks defect spikes. Attempting to resample a rejected shipment using the same or larger sample size without performing engineering root-cause analysis violates basic statistical governance. If a lot fails variable inspection due to excessive variance, simply resampling until a passing outcome occurs conceals active process instability from pack assembly teams.
- Notice of Material Non-Conformance A formal document detailing the shipment lot ID, container bill of lading, specific variable parameter failure metrics, sample calculations, and immediate quarantine location status.
- Third-Party Calibration Certificate Documentation proving that all metrology equipment, digital multimeters, thermal chambers, and Kelvin fixtures used during the failed inspection held active ISO/IEC 17025 calibration stamps at the moment of test execution.
- Root-Cause Corrective Action Request A mandatory engineering inquiry demanding the manufacturer provide an 8D failure analysis report identifying the process breakdown on the coating, assembly, or grading line.
- Commercial Disposition Invoice A financial ledger entry detailing accrued port demurrage, warehouse handling charges, administrative inspection costs, and return ocean freight billed directly to the cell vendor.
When statistical failure occurs under ANSI/ASQ Z1.9, switching rules mandate that subsequent incoming shipments from the same manufacturer move automatically from Normal Inspection to Tightened Inspection. Under Tightened Inspection, the required Quality Index threshold increases, effectively shrinking the Maximum Allowable Percent Non-conforming M and requiring tighter parametric clustering from the cell factory.
Tightened inspection doubles sample counts.
Sorting costs fall to the manufacturer. If project schedules prevent returning the rejected lot to the overseas factory, the buyer may execute a 100 percent sorting protocol on site, charging all screening equipment rental fees and labor hours back to the supplier through direct invoice credit memos. Variable data collected during 100 percent sorting isolates the specific non-conforming sub-lots, allowing compliant cells to enter production while defective units are scrapped under joint supervision.
Data integrity dictates shipment release.
Capacity margins determine battery pack life. Resolving disposition disputes requires pre-negotiated master supply contract terms that define variable statistical boundaries, standard deviation computation methods, and laboratory arbitration protocols prior to issuing purchase orders for high-capacity prismatic cells.

