Acceptable quality level

from Wikipedia, the free encyclopedia

Acceptable Quality Level ( AQL ) is a procedure or sample system for determining sample sizes in the company or for determining the acceptable quality limit in the area of quality management . Here, from a precisely defined procedure, a quantity of goods, for example from a delivery lot , a certain partial quantity is determined as a sample . This random sample must be checked according to certain regulations, such as DIN ISO 2859-1, which have been agreed by the supplier and customer .

description

This AQL method was developed by the US Army during the Second World War , but in some industries this method is no longer relevant because it is rather imprecise. Because with this sampling system, a certain subset of a delivery lot is defined. The test plan is precisely defined in the ISO standard, depending on the AQL value agreed between the customer and the supplier. AQL, also known as the acceptable quality limit, specifies how high the maximum proportion of a defective subset that is considered to be of acceptable quality in a sample. A partial quantity is selected from a total quantity of a certain product. A random sample is carried out on this in accordance with the statutory or regulated specifications and standards. Based on the final result of this sample, conclusions are drawn regarding the quality of the total quantity of a commodity. The severity of the guidelines set may vary. The standards for a certain product are higher and therefore the specified guidelines are also stricter. For this reason, AQL is considered a static method for determining quality.

requirements

As a rule, the AQL is expressed in percent , but without adding the percentage sign itself. However, this is only the case if the AQL is to be presented as the “proportion of defective units”. In addition, the representation in “Number of defects per hundred units” is known. For the limit value of the AQL, p = 1% or p = 0.01 was specified. In order to be able to guarantee an appropriately strict quality test and test procedure, it is important to determine an appropriate AQL using certain numerical values. This means that the supplier or sub-supplier knows what minimum quality this must have in production . This avoids the risk of an uneconomical rejection rate. The AQL system makes it possible in many cases to draw meaningful and cost-effective conclusions about the success or failure of the quality requirements. Of course, this is only possible at our discretion if a reasonable AQL value is established. As soon as this is chosen to be too high, the customer is exposed to the risk of receiving too high a proportion of defective units in a delivery lot (quantity of units that can be delivered at the same time) for the intended purpose without objection. Conversely, an AQL that is too low can also lead to an impairment of the economic efficiency of the process, since the testing is extremely costly. Both the buyer and the supplier will of course primarily insist on their own points of view, which can lead to problems. For example, the supplier can choose an AQL value that is too high, while the buyer could choose one that is too low. A successful business relationship is of course more beneficial for both parties and is feasible with a corresponding willingness on both sides. So there has to be a clear agreement on both sides. A solution can be found with less effort if both sides understand the effect of the AQL definition. In this way, we can also come closer to the economic optimum together. It is therefore assumed in the AQL system that both partners are familiar with the testing procedures of the sampling system for qualitative characteristics . It must be known that ...

  • ... an AQL application can only be implemented for delivery series.
  • ... a test only works properly if both parties are familiar with the technical specifications and test methods.
  • ... error classification and error weighting should be set up or how they should be set up and how they are sorted in error lists.

Exceeding the limit value or the AQL (this is the limit value) can have fatal consequences for a supplier in the long term, as it is confronted with a rejection. For this reason, the AQL with a sampling system according to ABC-STD-105 (Sampling Procedures and Tables for Inspection by Attributes, Superintendent of Documents, US Government Printing Office, Washington DC, 20402) has also established itself internationally. The AQL value determined by the buyer is primarily intended to point out to the supplier that the goods submitted for inspection by the supplier are subjected to a sampling plan and that this is only accepted if the proportion of defects does not exceed the defined AQL value. This is the description of the AQL as a so-called nominal value according to DIN 40 080.

construction

The test plan of DIN ISO 2859-1 is dependent on:

- AQL value → this is determined between the supplier and the customer

- Lot size N → is the number of the delivered quantity (= lot).

- Test level / levels → ISO 2859 differentiates between 7 test levels

  • 3 general, like I (good), II (sure), III (very sure)
  • 4 special S -1 to S - 4

The choice of test level depends on the cost optimization, but the most common level II is used.

If a series of inspection lots is tested, the results of the previous sample test must be taken into account when selecting a lot. As a schematic example, testing is carried out using the assessment levels (from DIN 40080).

1. Beginning of the exam

Unless otherwise agreed, the normal test will begin.

2. Change from normal to aggravated

If the normal test has started, a change must take place too intensely if 2 out of 5 consecutive test lots have failed the first test.

3. Then again change from aggravated to normal

If the test was tightened with, it must be skipped to normal after 5 lots have been accepted in the initial test.

4. Change from normal to reduced

Now, after the normal test, the reduced test can be passed over, whereby the following conditions must be met

  • At least 10 test items were successfully tested beforehand
  • When production runs smoothly without interruption
  • It was previously agreed that a change to the reduced examination may take place

5. From reduced to normal

A change back to the normal test takes place here if the following occurs during the initial test:

  • 1 lot was rejected
  • The sampling procedure was ended without the rejection criterion being met; the first lot is assessed as accepted and the next lot is checked according to the normal check
  • Production is interrupted or runs erratically
  • It was agreed that a switch to the normal exam could take place for other reasons

6. Abort the exam

If 10 (or another agreed number) consecutive inspection lots are to remain in the more stringent inspection, this should be canceled on the grounds that the quality of the products must be improved.

Tools for AQL determination

Using tools for the AQL determination, influencing factors can be assessed and classified well. However, a major disadvantage of such tools is that the influencing factors cannot be effectively assessed for industrial or all sectors. Influencing factors, such as business policy aspects, i.e. those that are usually assessed on the basis of feelings, cannot be recorded with tools. For this reason, it is believed that experience can be used to determine a useful AQL value, which is true in most cases. As a result, experience is also an important factor in determining and assessing an AQL value. It is advisable to only work with experience if the exact consequences of errors and clear and simple cases of investigation are known.

Meanings of the AQL value

Importance from the sampling system itself

Normally the interests of both parties, the supplier and the buyer should be taken into account. This means: A small number of delivery lots is not subjected to an overly extensive test. The reverse is also the case with large quantities of delivery lots: Overly imprecise tests are avoided. It is economically of great importance that the instruction of the random samples or the separability changes with the size of the delivery lot. Because the more extensive an examination lot, the more serious an incorrect examination result, which is due to the statistical probability of error, has. It is all the more important to keep this low, given a growing number of test lots with an equally growing separation capacity.

Meaning by applying the sampling system

Both the buyer and the supplier benefit from joint coordination. These can be both inside and outside the company. The necessary prerequisites to achieve optimization in the economic sense are only met if there is joint coordination. The relevant knowledge is of course also of great importance. It should not be forgotten that using the AQL is only an aid. If you want to get a more concrete and more economical result about the quality of the product, it is advisable to put the sampling system in connection with other information systems. As a result, the test planning is influenced indirectly or directly by the design, work preparation, market analysis and development. It is possible that experiences from another or similar delivery are also brought in. The quality capability of the supplier and possibly the sub-supplier also plays a major role. The proportion of errors that usually occur and the associated effort to keep them low, for the different quality features, must be known to the supplier himself in order to be able to carry out reliable quality planning. This enables the supplier to better negotiate the price with the customer. Of course, he needs this knowledge to agree on an exam. Otherwise the supplier cannot use his right to propose an appropriate AQL. The customer, on the other hand, should prepare an error analysis for his part. The buyer should inform the supplier about the consequences of certain errors. For this reason, error analysis is also recommended. All criteria of both parties should be discussed and debated together. Because the deliveries can vary depending on the value and size of the goods. The function of the product as well as the manufacturing process should also be known to both sides, because the agreement thus serves a careful examination.5 Because a random sample can only be carried out with the help of the lot size and the agreed AQL value. As a rule, this is done according to the standardized random sampling instructions according to DIN ISO 2859 Part 1.

procedure

On the basis of the random samples, a company learns whether the products in a lot meet the quality requirements. AQL can also allow errors and is therefore not suitable for all test cases such as B. suitable for pacemakers. It gives the manufacturer indications for his quality assurance. If a customer, the buyer, checks his delivery from the supplier, this must be agreed in advance. The lot (delivered quantity) must be unique, and it must be ensured that only accepted lots can be used. The acceptance or return of a delivery quantity is established precisely with the help of the random sample check. Rejected inspection lots are initially held back, the decision is to be made depending on the case (defect classification). In some cases, a stratified random sample is advantageous; the stratifications of a lot are understood to mean a so-called “subdivision” of the lot, which in itself can be more homogeneous than the total lot. The result is shown in the present lot from handling N, from which a sample of size n is taken at random. And the acceptance number contained in the sampling instruction is used to determine the significance of the lot.

Before the test, the actual value of the feature to be tested must be clearly defined and approved, which is taken from the test specification. If a lot does not meet the requirements, it is an error.

  1. Metal plates are checked for their dimensions at the reception. Now it is compared whether the measured values ​​of the thickness agree with the limit values ​​and thus one sees whether the test objects meet these requirements or not. If a test item does not meet this requirement, it is faulty and is named as a faulty unit.

Choice of conditions related to errors

Before you can start with the calculation of the AQL value, the so-called "boundary conditions" should be clear. In addition, both sides should of course define exactly what is defined as an error. In specialist circles, boundary conditions are understood to mean general conditions and requirements. This includes, among other things, whether it is possible to perceive an error check for every hundred units or a proportion of the defective units and what consequences this can have. For this reason it is important to be able to distinguish the terms:

  • Sampling based on the proportion of defective units:

→ With this type of random sample check, the defective portion is expressed as a percentage and this cannot exceed 100%. Only everyone can

Units will be faulty and no more.

  • Sampling based on the number of defects per hundred units:

→ The number of defects per hundred units is referred to here. In contrast to the other test, it is possible to exceed the value above 100 here. It is believed,

that on average one error per unit is acceptable, since per hundred units, one hundred errors are equal to 100%.

Failure units

An equally suitable random sample test must be prescribed or determined for a suitable AQL value. This is then associated with the term “defective unit”. Quality features that one would like to subject to a desired test are therefore closely linked to the concept of a defective unit. The reasons for classifying a unit as defective are basically divided into seven cases:

  1. A faulty value can be identified at most once if a quality feature occurs only once in a unit.
  2. Several incorrect values ​​can be marked if a quality feature occurs only once in a unit.
  3. An incorrect value can be identified k times if a quality feature occurs only k times in a unit.
  4. In the event of only one fault, the entire unit can be considered faulty as soon as some quality features form a cause for a unit to be classified as faulty. (Based on the first case)
  5. In the event of only one fault, the entire unit can be considered faulty as soon as some quality features form a cause for a unit to be classified as faulty. (Relating to the second case)
  6. In the event of only one fault, the entire unit can be considered faulty as soon as some quality features form a cause for a unit to be classified as faulty. (Relating to the third case)
  7. In the event of only one fault, the entire unit can be considered faulty as soon as some quality features form a cause for a unit to be classified as faulty. (In all cases from one to three)

Fault classifications

If an error occurs in the quality implementation of the random sample, due to the fact that a previously discussed feature does not correspond, this error is broken down into certain subdivisions. It is necessary that there is an assignment to a certain defect class for each quality feature. The following options for classification are:

Critical error → possible danger to human life

Main fault → may not be useful

Secondary errors → possibly limited usability

If necessary, this classification can be further subdivided into A-errors and B-errors.

Application with possible failure consequences

If a customer, the buyer, checks his delivery from the supplier, this must be agreed in advance. The lot (delivered quantity) must be unique, and it must be ensured that only accepted lots can be used. The acceptance or return of a delivery quantity is established precisely with the help of the random sample check.

If a checked lot meets the acceptance conditions, this is accepted and can usually be used further. However, if a sample test is negative, the lot is stopped. The decision about the treatment is made by considering the error classes. There are several options that can be taken if the lot fails the exam, either:

  • Returns and exchanges
  • Rework
  • sort by
  • Scrapping (at the expense of the buyer, costs are incurred by the supplier)
  • Use through discount

Individual evidence

  1. ↑ BTSSB Bartmann Total Solutions in Steel Business: AQL - Acceptable Quality Level. Retrieved June 22, 2018 .
  2. ↑ Katrin Melzer, Hochschule-Esslingen: Statistical Quality Control Lecture Chapter 6. from p. 8, accessed on June 22, 2018 (PDF).
  3. ↑ Sempermed, AQL - a guarantee for quality. Inform number 2, Vienna, 05/2015, accessed on June 22, 2018 (PDF)
  4. ↑ German Society for Quality eV and Swiss Working Group for Quality Promotion (ed.): Methods for determining suitable AQL values . Beuth Verlag GmbH, Berlin 1990, ISBN 3-410-32823-8 .
  5. ↑ a b c d e f Dipl.-Ing. Hans-J. Cloodt: Course materials for quality management: Materials on the subject of QM in production . tape 302 . Cloodt Verlag, September 2016.
  6. ↑ Friedhelm Denkeler, Denkeler Quality Management, Berlin: Denkeler QM: The AQL sampling system according to DIN ISO 2859-1. Retrieved June 30, 2018 .
  7. ↑ German Society for Quality eV Ffm (ed.): Sampling based on qualitative characteristics, procedures and tables according to DIN 40080 . 9th edition. Beuth Verlag GmbH, Berlin 1986, ISBN 3-410-32796-7 .
  8. ↑ German Society for Quality e. V. Ffm. (Ed.): Sampling plans for quantitative characteristics (variable sampling plans) according to ISO 3951 . 2nd Edition. Beuth Verlag GmbH, Berlin 1988, ISBN 3-410-32814-9 , pp. 16-17 .
  9. ↑ Dipl.-Ing. Hans-J. Cloodt: AQL samples: Cloodt quality management . Cloodt Verlag ( cloodt.de [PDF]).
  10. ↑ a b c d e f g h Gerhard Linß: Quality management for engineers . 4th edition. Carl Hanser Verlag (ebook), Munich 2018, ISBN 978-3-446-43936-8 , pp. 487 ( google.de ).
  11. ^ F. Denkeler: Quality management, AQL system. accessed on June 18, 2018 ( [1] PDF).