تجزیه و تحلیل سیستم اندازه گیری با استفاده از آزمایش های طراحی شده با حداقل خطرات α-β و n

Quality and productivity improvement are most effective when they are an integral part of the product and process development cycle. The main object of this study is re-identifying the variability sources that lead to errors in the measurements made for the correct evaluation of whether the targeted quality standards are reached in quality assurance systems, as well as at re-establishing a model with designed experiments, by virtue of including laboratory factor as a measurement variability factor into the Measurement System Analysis (MSA) studies, whereby it is currently ignored. The measurement systems that need to be examined and kept under control, in order to set the extent to which the products meet the customers requirements and expectations, have been analyzed statistically. Besides, new producer (α)–consumer (β) risks and the required minimum sample size (n) for its design will also be identified. As for the business organization chosen for the application of the model, a new determined sample size with α, β error probabilities has been identified as a result of the applications with the new model.

As today's favorite approach, “Zero-error” in quality, can be achieved only through efforts towards identifying the reasons of errors and preventing their occurrence by applying quality control and quality assurance systems. QS-9000 quality standard, whose primary aims are continual improvement, error prevention, minimization of extravagance and variabilities, and improvement of quality systems, is, in addition to ISO-9000 quality assurance conditions, a new quality assurance system that aims at perfection and zero-error. MSA, one of the reference fields of QS-9000, is a statistical approach used in the accuracy and sensitivity estimations made for a measurement device [1]. In establishing a quality system, MSA vested with the necessary assumptions should be applied perfectly since it is essential to be accurate in each measurement and evaluation [2].

In this study, MSA (Measurement Systems Analysis), one of the reference fields of QS-9000 Quality Assurance System, has been statistically investigated. Some generalized results have been achieved in consequence of the operations in a business management, and some suggestions have been put forward for managements that search for the accuracy and sensitivity of measurement systems. This study has tried to point out that “laboratory” factor, lower though considered as insignificant in “Measurement Systems Analysis” applied in today's organizations and accordingly ignored as a variability source, can lead to a significant measurement variability; as environment and atmosphere conditions of the measurement area are summarized in all descriptions. Although this hypothesis could not be verified in the specific organizations where the application was made, it is quite clear that laboratory factor can be an important factor too in affecting measurement errors. With the inclusion of laboratory factor into these applications, not only it will be possible to prevent different measurement results that could be obtained in producer and customer laboratories and therefore avoid conflicts between firms; but also will it be possible to prove that devices thrown away due to failure in adequate measurement are probably sensitive enough, and that simply the conditions of the laboratory are inappropriate; thus easily find out for which variability factor should the real improvement activities be planned. Furthermore, by means of indicating a quantitative variable for each feature of the laboratory conditions (heat, pressure, vibration, etc.), it would also be possible to estimate under which conditions and to what extends the laboratory could meet measurement errors. According to the results obtained after the analysis, some factors could be excluded from, or some new factors that are considered to be effective can be included in the experiment. This will help minimize the costs of the application and waste of time. The sample size n, as suggested by MSA studies, should be questioned. With an assessment in the framework of the acceptance plan, sample size has been re-estimated and a new sample size measurement (n=166) has been suggested. But also, in this study it is proven that getting sample size n, 180 is necessary and sufficient condition for the experiment in a (3 × 2 × 10) mode. So, in a MSA study, when taken into consideration all factors and levels that are deemed to be able to affect the measurement variance, it will be best fit to set the experiment according to a randomized complete block design and it will be necessary to repeat the experiments at least three times instead of twice in order to estimate experimental error under the given conditions. The determination of producer (α=0.0015) and consumer (β=0.0006) risks, as have been estimated in this study, will both enhance the point of view of the producer as well as customer firms; and supply information about at what intervals the measurement systems should be controlled and make guidance in the estimation of the adequacy of the sample measurements made during experiments. It goes without saying that the table that has been provided for the estimation of these risks through the shortest way possible will bring considerable advantages to the management concerning speed and costs. Furthermore, the number of the sample measurements should not be less than that provided in this study in MSA where the producer and consumer risks are very small, and that the sample measurements that has been used supports the sampling assumptions set worth in the statistical analyses are also among significant findings achieved through this study.