ISO 28597:2017 – Acceptance Sampling Procedures by Attributes — Specified Quality Levels in Nonconforming Items per Million

ppm-Based Acceptance Sampling for Modern High-Quality Manufacturing

ISO 28597:2017 addresses the limitations of traditional percentage-based AQL values in modern manufacturing environments where defect rates are measured in parts per million (ppm) rather than percentages. As manufacturing capabilities have advanced to Six Sigma levels (3.4 ppm), the conventional AQL series (0.01% to 10%) has become too coarse for meaningful sampling. This standard introduces quality levels specified directly in nonconforming items per million, enabling more discriminatory sampling for high-quality processes.

Traditional AQL = 0.01% represents 100 ppm, which is far above the defect levels achievable by modern automated manufacturing. ISO 28597 extends the quality level range to as low as 1 ppm, providing appropriate sampling plans for electronics, automotive, and medical device manufacturing.

Quality Level Series and Sample Size Tables

The standard defines quality levels (QL) in a preferred series: 1, 2, 5, 10, 20, 50, 100, 200, 500, 1000, 2000, 5000, and 10000 ppm. For each QL, the standard provides master tables for normal, tightened, and reduced inspection. The sample sizes are calculated to provide a consumer’s risk (β) of 10% at the limiting quality level (LQL), which is typically set at 3 to 5 times the QL. For QL = 100 ppm (0.01%), the normal inspection sample size for a moderate lot is approximately 1250 units.

Quality Level (ppm)	Normal n	Tightened n	Reduced n	Ac/Re for Normal
1	500,000	750,000	200,000	0/1
5	100,000	150,000	40,000	0/1
10	50,000	75,000	20,000	0/1
50	10,000	15,000	4,000	0/1
100	5,000	7,500	2,000	0/1
500	1,000	1,500	400	0/1
1,000	500	750	200	0/1
10,000	50	75	20	0/1

For ultra-high quality levels (≤ 10 ppm), the sample sizes become impractically large for lot-by-lot inspection. The standard recommends transitioning to process control-based acceptance (per ISO 28594) or using cumulative multi-lot sampling for these quality levels.

Switching Rules and Data Exclusion Protocols

ISO 28597 includes normative annex A on data exclusion rules. When a nonconforming item is found during inspection, the standard provides specific protocols for determining whether it represents an assignable cause that should trigger tightened inspection or a random fluctuation that can be managed through normal switching rules. The credit-based switching mechanism (similar to ISO 28593) is adopted, with credit scores calibrated for ppm-level quality. At the 100 ppm level, 10 consecutive conforming lots add one credit, while a single nonconforming lot deducts 5 credits.

The ppm-based plans are extremely sensitive to sample integrity. A single contaminated sample (e.g., due to cross-contamination from a previous high-defect lot) can trigger tightened inspection for 20+ subsequent lots. Strict segregation of samples by lot is essential for correct operation of the standard.

Engineering Application for High-Volume Production

For high-volume production lines (e.g., automotive electronics with daily output of 50,000+ units), ISO 28597 provides a rational basis for sampling that balances inspection cost against quality risk. The standard is particularly well-suited to automated optical inspection (AOI) systems where 100% inspection is technically feasible but economically undesirable. By establishing a ppm-based QL, manufacturers can calibrate their AOI systems to sample at rates that provide statistically valid quality estimates without slowing production throughput.

ISO 28597 should never be used for go/no-go decision on safety-critical characteristics where the acceptable defect rate is truly zero. The statistical nature of sampling means there is always a non-zero probability of accepting a nonconforming lot, regardless of the QL selected.

Statistical Basis and Sampling Plan Derivation

The ppm-based sampling plans in ISO 28597 are derived from the hypergeometric distribution for finite lot sampling, with the Poisson approximation applicable when the sample size is less than 10% of the lot size and the nonconforming rate is below 1%. For ultra-high quality levels (≤ 100 ppm), the Poisson approximation becomes inaccurate at the tails, and the standard requires exact calculation using the binomial or hypergeometric distribution for sample size determination. The plans are designed to provide a consumer’s risk (β) of 10% at the limiting quality level (LQL), which is typically set at 3 to 5 times the quality level (QL). For example, at QL = 1000 ppm (0.1%), the LQL is approximately 5000 ppm (0.5%), meaning there is only a 10% chance of accepting a lot with 0.5% nonconforming items. The producer’s risk (α) at the QL is held at approximately 5% for normal inspection, ensuring that suppliers producing at the agreed quality level experience rejection rates no higher than 5% due to sampling variation. The standard provides master tables for code letters A through R (following the ISO 2859-1 code letter system), with sample sizes ranging from 5 to 5000 units depending on the QL and lot size. For QL = 100 ppm with lot size 50,000, the normal inspection code letter is N with sample size 1250 and acceptance number Ac = 1 (since the accept-zero criterion for single sampling would require impractically large samples at this quality level).

The ppm-based approach has been widely adopted in the automotive and electronics industries, where defect rates below 100 ppm are routine for mature production processes. By providing sampling plans calibrated for ppm-level quality rather than percentage-level, ISO 28597 enables meaningful acceptance sampling in these industries without requiring impractically large sample sizes that would be specified by percentage-based standards like ISO 2859-1.

Application to Automated Inspection Systems

ISO 28597 is particularly well-suited for integration with automated optical inspection (AOI) and other high-speed inspection systems. The standard provides guidance for converting between 100% automated inspection and statistical sampling: when an AOI system achieves a demonstrated probability of detection (POD) greater than 99.5% for all defect types of interest (validated through formal POD studies per MIL-HDBK-1823A), the sampling intensity can be reduced by applying an equivalence factor. For example, an AOI system with 99.5% POD at QL = 500 ppm provides quality assurance equivalent to traditional visual inspection at QL = 1000 ppm, allowing a 50% reduction in sample size. The standard requires that the AOI system’s POD be revalidated at least quarterly and whenever the product design or defect criteria change. For mixed manual and automated inspection lines, the standard provides a framework for combining results from both inspection modalities into a unified acceptance decision, accounting for the different POD characteristics of human inspectors versus automated systems. This flexibility makes ISO 28597 adaptable to modern smart factory environments where inspection resources are dynamically allocated based on real-time quality data.

When using automated inspection systems for ppm-level sampling, an important consideration is the false rejection rate. AOI systems, particularly those using machine vision with deep learning classification, can have false rejection rates of 1-5% even when correctly calibrated. These false rejections must be distinguished from true defect detections through a verification stage, as they would otherwise artificially inflate the apparent nonconforming rate and trigger unnecessary tightened inspection. The standard recommends a two-stage inspection approach: an initial automated screen with sensitive (low-threshold) defect detection, followed by human verification of all flagged items to separate true defects from false calls.

FAQ

Q: How does ISO 28597 compare with the C=0 plans in ISO 28593?
A: ISO 28597 uses ppm-based quality levels while ISO 28593 uses percentage-based AQL with a credit system. ISO 28597 is better suited for high-quality processes with ppm defect levels.

Q: What is the minimum lot size for applying ISO 28597?
A: The standard requires a minimum lot size of at least 10 times the normal sample size. For QL = 1000 ppm, where normal n = 500, the minimum lot size is 5,000 units.

Q: Can ISO 28597 be used for destructive testing?
A: Yes, but the economics of destructive testing at ppm-level quality must be carefully evaluated. Cumulative sampling across multiple lots may be more practical.