How should sampling frequency be set for batch inspections?

Setting the correct sampling frequency for batch inspections is a critical quality control decision that strikes a balance between risk, cost, and efficiency. A well-designed sampling plan provides a high level of confidence in batch quality without the prohibitive expense of 100% inspection. It is a statistical tool grounded in acceptance sampling principles, primarily governed by the Acceptable Quality Limit (AQL).

Core Principles of Acceptance Sampling

The foundation of any sampling plan is understanding that it assesses batch quality based on a representative subset, not absolute certainty.

Understanding AQL and Risk

The Acceptable Quality Limit (AQL) is the worst tolerable process average quality level you are willing to accept. A sampling plan defined by an AQL (e.g., 1.0% for major defects) does not mean you accept batches with 1.0% defects; it means you have a high probability (typically 95%) of accepting batches that are at or better than the 1.0% defect level. The corresponding risk is the Lot Tolerance Percent Defective (LTPD), the quality level that you have a low probability (typically 10%) of accepting. This balances the producer's risk (rejecting a good batch) and the consumer's risk (accepting a bad batch).

Determining Inspection Levels and Sample Size

Sampling plans, such as those in ISO 2859-1, use inspection levels (General I, II, III or Special S-1 to S-4) to determine the sample size based on the batch size. General Level II is most common. The sample size code letter from the table, combined with your chosen AQL, determines the number of units to sample and the acceptance/rejection number. For critical components in industries like Medical Device or Aerospace and Aviation, a tighter AQL (e.g., 0.65% or 0.10%) and a higher inspection level (III) are typically mandated to minimize risk.

Key Factors Influencing Sampling Frequency

The sampling plan is not one-size-fits-all; it must be tailored to specific product and process factors.

Process Capability and Historical Stability

A stable and capable manufacturing process with a proven history of high yield justifies a reduced sampling frequency or a looser AQL. For instance, a well-controlled CNC Turning Service producing simple shafts might use a routine AQL of 1.5. Conversely, a new or unstable process, or one machining challenging materials like Titanium CNC Machining or Inconel 718, requires tighter inspection, perhaps even 100% verification on critical features until stability is demonstrated.

Part Criticality and Failure Severity

The potential impact of a part failure is the most important factor in setting sampling rigor. This can be broken down into three categories:

• Critical: Failure could cause injury or catastrophic system failure. Requires the tightest AQL (e.g., 0.10%) and often 100% inspection for that characteristic. This is non-negotiable for safety-critical components in Automotive or aerospace applications.

• Major: Failure would likely result in a product that is unfit for use. A standard tight AQL (e.g., 0.65% or 1.0%) is applied.

• Minor: Failure would not significantly affect usability but may impact aesthetics. A more lenient AQL (e.g., 2.5%) can be used.

Batch Homogeneity and Supplier Performance

A batch made from a single material lot in one continuous production run is more homogeneous than a batch compiled from multiple setups. For homogeneous batches, statistical sampling is more reliable. Furthermore, certified or high-performing suppliers with validated processes, such as those offering One-Stop Service, may be granted reduced sampling based on their proven quality history.

Implementing a Dynamic Sampling Strategy

A static sampling plan can become inefficient over time. A dynamic strategy responsive to data is far more effective.

Shifting Between Normal, Tightened, and Reduced Inspection

ISO 2859-1 allows for switching between inspection severities based on historical performance. If several consecutive batches are accepted under normal inspection, you may switch to reduced inspection to save cost. If two out of five consecutive batches are rejected, you must switch to a tightened inspection to protect the consumer, either by increasing the sample size or requiring a higher quality level for acceptance.

Integrating with First Article Inspection (FAI)

For new parts or after a significant process change, a full First Article Inspection is mandatory. This is essentially a 100% inspection of all designated characteristics on a small, initial production run. The data from the FAI validates the manufacturing and Prototyping Service process and provides the initial process capability data that will inform the sampling frequency for subsequent Mass Production Service batches.

In conclusion, setting the sampling frequency is a strategic decision. It begins with a standard like AQL but must be refined based on process capability, part criticality, and supplier performance. A dynamic plan that responds to quality data ensures that resources are focused on the highest risks, guaranteeing product quality and supply chain efficiency.