How to conduct sampling inspection is a very important issue in promoting quality control. This is because this system determines the level of value delivered to customers. Therefore, this time, we will talk about the relationship between the AQL (Acceptance Quality Limit) of Sampling Inspection and Inspection Level, which is important for quality control, and the points to be observed when deciding the quantity.

AQL and Inspection Level in Sampling Inspection

In promoting “manufacturing”, it is important to prevent defects from flowing out by building quality into the process. In order to do that, when looking at the flow of manufacturing, what kind of processing should be performed at which part, and what kind of inspection should be performed? How will the producer ensure the quality level of the designer? It is important to carefully consider these issues. In other words, it is a sampling inspection plan.

Usually, the inspection system is set up by the person in charge of quality after assessing the condition in which the product is made. At the beginning of product production, the processing state is not stable, so it is necessary to establish an inspection system that can transmit information that stabilizes.This is called initial flow management.

If you remove the unstable factor and continue production for a while, you will get used to it and stabilize little by little. When this situation occurs, we will prepare an inspection system by combining the types of sampling inspections. The quality manager sets AQL and inspection level of sampling inspections using probability so that the degree of contamination of overall defects and defects is allowed within the target range. Let’s talk about that how to select AQL and Inspection Level for your sampling inspection plan!

“AQL” stands for Acceptance Quality Limit, which means “Acceptable Quality Limit”. AQL is the upper limit that a product inspected by sampling inspection can pass, and is the inspection level used for general sampling inspection. In an AQL sampling inspection, if the product is sampled from a process where the quality of the product is not below AQL, it is likely to pass. In addition, this standard is based on ISO 2859-1, so it can be used as a common inspection standard even when expanding overseas.

1 – Determine the principles of AQL

AQL is the dividing line between acceptable and unacceptable process means, is a quality parameter related to sampling plans, and is also an index to retrieve sampling plans from a sampling system such as ISO 9001. However, so far there is no one method that can be applied to all different situations.

Here are some principles to follow when determining AQL:

  • 1: The AQL value is a satisfactory average quality of the process, and the average batch quality of the supplier is expected to not exceed this value.
  • 2: Consider the purpose of the product and the consequences caused by the failure of the product. For example, the AQL value of electronic components of the same specification can be larger when used in general civilian equipment than when used in military equipment; The AQL value can be larger for ordinary instruments than for precision instruments.
  • 3: Although AQL is not a requirement for the quality of individual batches, if it is known that the number of unqualified products per 100 units of products (or the number of unqualified products per 100 units of products) of a certain batch does not exceed the specified AQL value, the batch should be received; Otherwise, the batch should be rejected.
  • 4: From the consideration of the user, the AQL value cannot be required to be too small, otherwise, or the inspection fee and product cost will increase, or the product batch is often rejected, so that the supplier refuses to sign the contract.
  • 5: If the quality of the producer is difficult to improve the quality of the product urgently needed by the user, in order to obtain the product, the AQL value must be selected appropriately larger, and the AQL value will be adjusted after the quality of the producer is improved.
  • 6: When there are many spare parts, and the unqualified parts are easy to find and replace with qualified parts when assembled as a whole piece, the AQL can be selected larger; If a part fails and cannot be removed from the whole piece and replaced, so that the whole piece fails, AQL should be selected as smaller.
  • 7: Considering the assembly sequence, if the unqualified products in the previous process will cause waste of time and material resources to the next process, the AQL value of the previous process should be smaller than the AQL value of the latter process.
  • 8: You can specify AQL value for a single inspection item, or you can jointly specify an AQL value for a group of inspection items.
  • 9: Considering the severity of the impact of unqualified (product) on product performance, in the same acceptance sampling problem, it is generally required that the AQL value of Class A unqualified (product) is less than the AQL value of Class B unqualified (product); The AQL value of Class B unqualified (product) is less than the AQL value of Class C unqualified (product). For example, the AQL values of Class A, B and C nonconformities (products) are 0.15, 0.25 and 0.40, respectively.
  • 10: AQL can not tell the user, a single batch of products can get quality protection how, in order to clarify the quality protection of the user, need to look at the design value of GB2828 Figure 1 ~ Figure 11 given the design value of a sampling scheme OC curve and its corresponding value table.
  • 11: The determination of AQL value depends on the compromise between the quality that the producer may provide and the quality that the user considers ideal, that is, it is a compromise between the quality that the user wants and the quality he can afford, the stricter the quality requirements, the more difficult it is for the producer to meet, the higher the inspection fee, and these costs are ultimately calculated on the product.
  • 12: The AQL value is not reassigned for each batch, but is specified by the responsible department (or responsible person) in the contract. Once specified, it cannot be changed arbitrarily.

2 – Methods for determining AQL

Here are some methods to determine AQL values for your reference.

2.1 – Engineering Law

The engineering method is to take the quality that must be technically guaranteed as AQL according to the performance, life, interchangeability, assembly, safety and other quality requirements of the product.

2.2 – Similarity Law

New products do not have historical data on quality, in order to determine the AQL value, you can refer to the AQL value of similar products (raw materials, structure, process, etc.).

2.3 – Rule of thumb

AQL is agreed upon by the personnel in charge of the department, engineering and technical personnel, and quality management personnel based on the experience of process capabilities.

2.4 – Experimental method

The experimental method is to temporarily specify an AQL value without any clear information to refer to, and then adjust the AQL according to the relevant quality information obtained during use.

2.5 – Calculation method

For multi-component complex systems, AQL can be calculated by considering the relationship between parts, components and the whole machine. If a system or component consists of several identical parts or different parts with the same AQL. Then the AQL value of the part (denoted as AQLx) should be appropriately less than the AQL value of the whole piece (denoted as AQLX). Assuming that the whole machine is composed of n identical parts and is in series, that is, the failure of one part will cause the failure of the whole machine, in this case, AQLx and AQLX have the following relationship:(AQLX/100)=1(1-(AQLx/100))n

2.6 – Estimation method

Within a factory or enterprise, in order to promote the improvement of product quality, the estimated process average or a value slightly less than it can be used as the AQL value.

2.7 – Reverse Thrust Method

The backward method does not directly specify the AQL value, but finds a normal inspection sampling scheme, so that its OC curve passes through an important point with a relatively low probability of reception, such as an indiscriminate point (the control point of the OC curve) and a user-side risk point, and inverts the required AQL value.

2.8 – Model method

Build a cost model and select the AQL value so that the total cost of the corresponding sampling plan is minimized.

3 – Allocation of AQL

3.1 – General

There are various ways to assign AQL values to different categories of nonconformances. Perhaps the simplest thing is to divide all nonconformities into two categories, A and B, and assign an AQL value to each class, for example:

After specifying the type of sampling scheme, two different sampling schemes can be retrieved from the normal inspection sampling scheme, and when both schemes make a qualified judgment, the batch of products can be judged to be qualified for normal inspection; As long as a sampling scheme makes a judgment of non-conformity, the batch cannot be judged to be qualified for normal inspection.

3.2 – There are more than two types of non-conformities

For more than two scenarios, such as:

To determine that the inspection batch is a qualified batch, the three corresponding sampling schemes of A, B and C must make a qualified judgment.

3.3 – Combined with synthesis, respectively

Each performance of the product is assigned an independent AQL, and an AQL is specified for all performances combined. For example, a certain type of nonconformance includes three types of performance, as shown in the following table:

This practice has some reference value for complex products with many independent performance tests to be tested.

3.4 – Single vs. Combination

Provision one AQL for Class A nonconformances, and one AQL for Category A nonconformities and Category B nonconformities combined.

So, what did you think? It may seem a little complicated, but in short, if you decide on various conditions and use the AQL tables stipulated by ISO, you can build a statistically managed quality assurance system.