Giriş kalite kontrolunda kullanılan kabul örneklemeleri ve bir işletmede uygulanması
Giriş kalite kontrolunda kullanılan kabul örneklemeleri ve bir işletmede uygulanması
Dosyalar
Tarih
1996
Yazarlar
Topgül, Onur
Süreli Yayın başlığı
Süreli Yayın ISSN
Cilt Başlığı
Yayınevi
Fen Bilimleri Enstitüsü
Özet
Bir malı, malzemeyi veya hammadeyi kullanacak kişiler ile bunu üretenler karşı karşıya geldiğinde uygulanan kalite kontrolü Giriş Kalite Kontrolü veya muayenesi olarak adlandırılır. Seri üretime dayanan büyük miktarlarda malların, belirli zaman aralıklarına göre, partiler halinde teslimi söz konusu olduğu zaman, malzemelerin istenilen standatlara uygun olup olmadıklarına, saptanan spesifikasyonu karşılayıp karşılamadığına bakılarak parti hakkında kabul veya red karan verilir. Bu karar, belirli standartlara göre üretilen malların oluşturduğu partinin muayenesine dayanır ki, burada kullanılan örnekleme muayenesidir. Kabul örneklemelerinin, kabul veya red karan verilmesi için dayanılan örnek sayısı bakımından; - Birli Kabul Örneklemeleri, - Çift Kabul Örneklemeleri, - Ardışık - İkili Kabul Örneklemeleri, olmak üzere üç türü vardır. Kalite kusurlu oranına dayanılarak belirtilirse ve her parti, bir tek örneğin muayenesine göre kabul veya red edilirse, buna kusurlu oranına dayanan birli kabul örneklemesi denir. Birli kabul örneklemesi, parti büyüklüğü N, örnek büyüklüğü n ve kabul için örnekte çıkmasına izin verilecek en yüksek kusurlu sayısı c olmak üzere üç faktöre göre belirlenir ve karar her zaman tek bir örnekle verilir. Çift örnekleme planlarında, partiyi red veya kabul karan partiden çekilen tek bir örnek grubuna göre değil, İM örnek grubunun muayenesi sonucunda verilir. Alınan örnek grubu birli örnekleme planındakinden daha küçüktür. Bu durumda partinin çapı büyüdükçe çift örneklemeden sağlanan kontrol giderlerinden tasarruf daha fazla olur. Ardışık örnekleme planlarında, örnek büyüklüğü bir sabit değer değil, bir tesadüfi değişkendir. Bu yöntemin en belirgin özelliği aşamalardan oluşmasıdır. Muayenenin her aşamasında, yeni örnekler alınmasına gereksinim olup olmadığına, yani muayenenin devam edip etmeyeceğine belirli kurallar çerçevesinde karar verilir. Bu örnekleme planlarının hazırlanması ve kullanıma geçirilmesi için kolay anlaşılır ve uygulanabilir tablolar geliştirilmiştir ki bunlar standart örnekleme tablolar olarak adlandırılırlar. Bu yöntemlerden yola çıkılarak bir örnekleme planı dizayn edilmiş ve bir işletmede uygulanmıştır.
The trend in industry has been to shift the burden of ensuring incoming quality to the vendor. Various quality assurance programs are being developed in which the vendor provides proof of product quality to the purchaser. Vendor documentation of ongoing quality improvement programs is usually required, along with access to control charts at critical process points. The receiving companies also have a vendor-rating system based on the quality history of the vendor. Sampling does a good job of accepting very good lots and rejecting very bad lots. Unfortunately, a large area on inspection lies in the middle. The sampling rules in all the formal sample plans are based on probability, but the application of probability predicts the acceptance of lots with substandard quality. When production has already taken place, we often wish to know the quality level of the lot. When a supplier ships a batch of parts, for example, should they be accepted as good or not? Acceptance sampling is the statistical quality control technique for making these kinds of decisions. To specify a particular sampling plan, we indicate the sample size, n, and the number of defectives in the sample permitted, c (acceptance number), before the entire lot from which the sample was drawn is to be rejected. The operating characteristics (OC) curve for a particular combination of n and c shows how well the plan discriminates between good and bad lots. Therefore, if the actual quality is good, the plan provides for a high probability of acceptance, but if the actual quality is poor, the probability of acceptance is low. Thus, the OC curve shows how well a given plan discriminates between good and poor quality. A sampling plan that discriminates perfectly between good and bad lots would have vertical OC curve. For all lots having percent defectives to the right of the line, the probability of acceptance is zero. Unfortunately, the only plan that could achieve this discrimination is one requiring 100 percent inspection. Therefore, the justification of acceptance sampling turns on a balance between inspection costs and the probable costs of passing bad parts. By making sampling plans more cuscriminating (increasing sample size) or tighter (decreasing acceptance numbers), we can approach any desired level of outgoing quality that we please, but at increasing inspection costs. This increased inspection effort would result in lower probable costs of passing defective parts. At some point the combination of these incremental costs is a minimum. This minimum point defines the most economical XI sampling plan for a given situation. Obviously, if the cost of passing defective products is high, a great deal of inspection is economically justified. To justify 100 percent inspection of a sample, the probable losses due to the passing of bad products would have to be large in relation to inspection costs, perhaps resulting in the loss of contracts and customers. It is on this basis that the Japanese objective of zero defects can be justified. On the other hand, to justify no inspection at all, inspection costs would have to be very large in relation to the probable losses due to passing bad parts. The most usual situation is between these extremes, where there is a risk of not accepting lots that are actually good and a risk of accepting lots that are bad. OC curves can be constructed from data obtained from normal or Poisson distributions. If lots are large, perhaps greater than 10 times the sample size, probabilities for the OC curve can be obtained from the binomial distribution. However, if samples are large, the normal or Poisson approximations are also very good, and they are much more convenient to use. Usually, the lot percent defectives is small and the lots are relatively large, so the Poisson distribution is used to calculate values for the percentage probability of acceptance, for OC curves. We can explain some terms of sampling; AQL (Acceptable Quality Level): Lots of this level of quality are regarded as good, and we wish to have a high probability for their acceptance. a (Producer' s Risk): The probability that lots of the quality level AQL will not be accepted. Usually a=5 percent in practice. LTPD (Lot Tolerance Percent Defective): The dividing line selected between good and bad lots. Lots of this level of quality are regarded as poor, and we wish to have a low probability for their acceptance. P (Consumer' s Risk): The probability that lots of the quality level LTPD will be accepted. Usually (3=10 percent in practice. To specify a plan that meets the requirements for AQL, a, LTPD, and p, we must find a combination of n an c with an OC curve that passes through these points. The mechanics of actually finding specific plans that fit can be accomplished by using standard tables, charts, or formulas that result in the specification of a combination of n and c that closely approximates the requirements set of AQL, a, LTPD, and p. When we set the levels for each of these four values, we are determining two critical points on the OC curve that we desired, points (AQL, a) and (LTPD, P). xu To specify a plan that meets the requirements for AQL, a, LTPD, and P, we must find a combination of n and c with an OC curve that passes through points (AQL, a) and (LTPD, P). The mechanics of actually finding specific plans that fit can be accomplished by using standard tables, charts, or formulas that result in the specification of a combination of n and c that closely approximates the requirements set for (AQL, a) and (LTPD, P). We can also calculating AOQ ( Average Outgoing Quality). The random sample of size n is inspected, and any defectives found in the sample are replaced with good parts. Based on the number of defectives, c', found in the sample, the entire lot is accepted if c'
The trend in industry has been to shift the burden of ensuring incoming quality to the vendor. Various quality assurance programs are being developed in which the vendor provides proof of product quality to the purchaser. Vendor documentation of ongoing quality improvement programs is usually required, along with access to control charts at critical process points. The receiving companies also have a vendor-rating system based on the quality history of the vendor. Sampling does a good job of accepting very good lots and rejecting very bad lots. Unfortunately, a large area on inspection lies in the middle. The sampling rules in all the formal sample plans are based on probability, but the application of probability predicts the acceptance of lots with substandard quality. When production has already taken place, we often wish to know the quality level of the lot. When a supplier ships a batch of parts, for example, should they be accepted as good or not? Acceptance sampling is the statistical quality control technique for making these kinds of decisions. To specify a particular sampling plan, we indicate the sample size, n, and the number of defectives in the sample permitted, c (acceptance number), before the entire lot from which the sample was drawn is to be rejected. The operating characteristics (OC) curve for a particular combination of n and c shows how well the plan discriminates between good and bad lots. Therefore, if the actual quality is good, the plan provides for a high probability of acceptance, but if the actual quality is poor, the probability of acceptance is low. Thus, the OC curve shows how well a given plan discriminates between good and poor quality. A sampling plan that discriminates perfectly between good and bad lots would have vertical OC curve. For all lots having percent defectives to the right of the line, the probability of acceptance is zero. Unfortunately, the only plan that could achieve this discrimination is one requiring 100 percent inspection. Therefore, the justification of acceptance sampling turns on a balance between inspection costs and the probable costs of passing bad parts. By making sampling plans more cuscriminating (increasing sample size) or tighter (decreasing acceptance numbers), we can approach any desired level of outgoing quality that we please, but at increasing inspection costs. This increased inspection effort would result in lower probable costs of passing defective parts. At some point the combination of these incremental costs is a minimum. This minimum point defines the most economical XI sampling plan for a given situation. Obviously, if the cost of passing defective products is high, a great deal of inspection is economically justified. To justify 100 percent inspection of a sample, the probable losses due to the passing of bad products would have to be large in relation to inspection costs, perhaps resulting in the loss of contracts and customers. It is on this basis that the Japanese objective of zero defects can be justified. On the other hand, to justify no inspection at all, inspection costs would have to be very large in relation to the probable losses due to passing bad parts. The most usual situation is between these extremes, where there is a risk of not accepting lots that are actually good and a risk of accepting lots that are bad. OC curves can be constructed from data obtained from normal or Poisson distributions. If lots are large, perhaps greater than 10 times the sample size, probabilities for the OC curve can be obtained from the binomial distribution. However, if samples are large, the normal or Poisson approximations are also very good, and they are much more convenient to use. Usually, the lot percent defectives is small and the lots are relatively large, so the Poisson distribution is used to calculate values for the percentage probability of acceptance, for OC curves. We can explain some terms of sampling; AQL (Acceptable Quality Level): Lots of this level of quality are regarded as good, and we wish to have a high probability for their acceptance. a (Producer' s Risk): The probability that lots of the quality level AQL will not be accepted. Usually a=5 percent in practice. LTPD (Lot Tolerance Percent Defective): The dividing line selected between good and bad lots. Lots of this level of quality are regarded as poor, and we wish to have a low probability for their acceptance. P (Consumer' s Risk): The probability that lots of the quality level LTPD will be accepted. Usually (3=10 percent in practice. To specify a plan that meets the requirements for AQL, a, LTPD, and p, we must find a combination of n an c with an OC curve that passes through these points. The mechanics of actually finding specific plans that fit can be accomplished by using standard tables, charts, or formulas that result in the specification of a combination of n and c that closely approximates the requirements set of AQL, a, LTPD, and p. When we set the levels for each of these four values, we are determining two critical points on the OC curve that we desired, points (AQL, a) and (LTPD, P). xu To specify a plan that meets the requirements for AQL, a, LTPD, and P, we must find a combination of n and c with an OC curve that passes through points (AQL, a) and (LTPD, P). The mechanics of actually finding specific plans that fit can be accomplished by using standard tables, charts, or formulas that result in the specification of a combination of n and c that closely approximates the requirements set for (AQL, a) and (LTPD, P). We can also calculating AOQ ( Average Outgoing Quality). The random sample of size n is inspected, and any defectives found in the sample are replaced with good parts. Based on the number of defectives, c', found in the sample, the entire lot is accepted if c'
Açıklama
Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Sosyal Bilimler Enstitüsü, 1996
Anahtar kelimeler
Kalite kontrol,
Quality control