Novel Approaches To Data Handling Of ATS Calibration

With the rapid development of computer and electronic technologies,a large number of equipments and products become more and more digital,automatic,and intelligent,which makes ATS(automatic test)more and more popular.Measurability is the premise of calibration,which draws a lot of people's attention in recent years.Measurability starts from the design step of a product,so the concept of calibration exits in the whole life cycle of the products.Because there are little references about measurability at present,this paper proposes the concepts of measurability and its connotation and indexes.From the relationships between measurability and testability,the information flow model of measurability is established based on the idea of testability model.The evaluation indexes are calculated according to the model,which is an effective approach to the measurability evaluation.Furthermore,the calibration interval of on-station calibration is also optimized by applying the metrology contribution rate.Another model of measurability which is called multi-signal cloud model is also applied to analyze the analog part of ATS.The calibration tree is obtained to direct the calibration process.The calibration data of ATS should be reasonably treated and analyzed,which includes the determination of gross error,the evaluation of measurement uncertainty,and the decision making of calibration conclusion.Because the rapid development of uncertainty theories(grey theory,cloud model,fuzzy theory,rough set theory,and so on),this paper focus on the applications of these new methods to handle the calibration data of ATS:(1)Since the tradition approaches to detect gross errors are absolute,and also sometimes the decisions are different for different rules,the concept of outlierness is proposed in this paper.The absolute treating of gross error changes into the uncertainty handling in this way.The inference rules of gross error are analyzed,and the traditional Mamdani and cloud model approaches are applied.The comparisons are made and the result approves the validity of this approach.(2)Measurement uncertainty is an important factor for the quality of ATS calibration,and the traditional GUM approach starts form the theory of randomness,which can only handle the uncertainty caused by random effects.The RFV(random-fuzzy variable)method combines the probability and fuzzy theories into the evidence theory,which can handle the random and nonrandom effects at the same time.The basis of RFV approach is analyzed in this paper,and a new method for the transformation between the probability distribution function and membership function is proposed.This novel approach is applied for the calibration of DMM voltage in ATS and the results shows that the random part is in accordance with the GUM method,though the systematic part is different.(3)Calibration always ends up with decision making,according to the result of a comparison between the measurement result and the standard value.Because the calibration is not under an ideal condition,there are decision making risks about the final result,which include false accept rate and false reject rate.The risk of a single calibration conclusion is analyzed in this paper,and the relations between the TUR(test uncertainty rate)and risk are also discussed.Furthermore,Monte Carlo method is used to evaluate the risk of decision making for the DMM calibration,which shows how to decrease the false-accept and false-reject rate.(4)When using RFV to represent the calibration data,it is often necessary to compare one fuzzy variable with another,which is the ranking problem of fuzzy variables.According to the characters of measurement data,a new ranking method is proposed in this paper to decide whether the unit under test is in tolerance or not.This new approach can obtain the credibility coefficients.The proposed method is simpler and direct comparing with the random method.