·Home ·Table of Contents ·Workshop - Reliability

Systematics of NDE Reliability: A Practical Point of View Gregory Ya. Dymkin, Vera N. Konshina (NDT Center of RIBNT, St.Petersburg RUSSIA) Contact

Introduction

NDE Reliability is the degree to which an NDE systems achieves its purpose of detection and describing discontinuities (defect states). In general NDE reliability depends on a technique's Ideal (Intrinsic) Capability (IC), and is generally degraded by the effects of application parameters (AP) and Human Factors (HF). However the factors forming AP and HF are intimately related and should not be considered as independent variables. This paper addresses and experimental approach to assessing the reliability of NDE systems.

Systematics of NDE reliability

From a practical point of view, quantitative evaluation of NDE reliability is of special interest because it can be used to validate NDE techniques. Modern methods for evaluating of NDE system and technique performance should reflect the ability to detect defects and, in some cases, their technical-economic impact, e.g., cost of false calls. Statistical indexes currently used for NDE system (technique) evaluation include:
1. reliability indexes - these define the ability of NDE system to perform its functions; i.e. ability to test for and evaluate defects,
2. in-service reliability indexes - these define the ability of NDE system to maintain installed values of reliability indexes and
3. efficiently indexes - these define the technical and/or economic effects resulting from the use of an NDE system.
From the technical point of view NDE system could mostly be evaluated with reliability indexes.

Practical approach

While doing the validation of NDE techniques one have to: select the reliability indexes and the collection of test pieces; carry out NDE with the technique have to be validated; study defect state; evaluate (calculate and compare) the reliability indexes. The steps of an "ideal" and proposed validation procedures are shown in the table 1.

An "ideal" way ensure complete and precise evaluation but it is not reproducible and could not be realized because of difficulties to select the representative collection of a test pieces with nature flaws and their destructive examination.

The proposed "practical" way is based on using the test pieces with an artificial reflectors. The only requirement is to make an adequate artificial reflector in the object under NDE and to proof that their are covering the same range of reflection properties that an nature flaws are.
If a manual NDE techniques are validating we must to take into account the effects of AP and HF. That is why practical assessment have to be made

1. as a blind trial,
2. by personnel,
which qualification level and experience are likely the same as in industrial application.

Stage	"Ideal" procedure	"Practical" procedure
Selection of the Reliability Indexes	Justification and choosing of the Reliability Indexes from the list of recommended ones
1. Selection of the test pieces	The representative collection of a test pieces with the real flaws	The number of a test pieces with an artificial reflectors of adequate reflection properties
2. Practical assessment	Blind Trial using technique under validation and appropriate qualified personnel
3. Description of the defective conditions (defect state)	Open trial with an expert technique and Destructive examination	-
4. Statistical assessment	Signal interpretation and calculation of the Reliability Indexes
5. Evaluation of the result	Comparing of the Reliability Indexes to requirements or other NDE technique, that have been already validated
Table 1 : Essential Steps and Features of NDE Technique Validation Procedures

No	Terms of the Reliability Indexes	Name	Method
1.	Probability of truth detection	P11	POD
2.	Probability of false call	P01
3.	Indicative size of the defect under given Probability of truth detection	X(P11*)
4.	Probability of truth detection under given Indicative size of the defect	P11(X*)
5.	Probability of truth detection under given Probability of false call	P11(P01*)	ROC
6.	Probability of false call under given Probability of truth detection	P01(P11*)
7.	"Reliability function"	D =1 - (P01 + P10)
8.	Square under ROC-curve
9.	"Truth rate"
10.	Average "Truth rate"	(P11/ P01)av
Table 2 : Statistical Reliability Indexes

Note: X- is the measured equivalent square (s_e) or measured length (dL) of the defect

The statistical indexes (Table 2) are based on the probability of true detection and false calls resulting from use of an NDE system. They are derived and constructed using POD and ROC methods.

Experimental investigation

A special experiment based on comparing two sets of experimental data obtained by two separate techniques makes it possible to experimentally check the performance of various reliability indexes. Testing results was obtained using one set of the test pieces and two ultrasonic testing techniques of a priori different reliability (fig.1).

Fig 1: Investigation of the stastical reliability indexes

The test pieces contains the segment reflectors of different size. Reflection properties of segment reflectors extremely depends on the direction of insonification.. The experiment made it possible to verify the following:

• several indexes including
  1. S;
  2. P₀₁(P₁₁^*); P₁₁(P₀₁^*).
  These indexes were statistically indistinguishable for the two compared techniques, and
• other indexes such as
  1. P₁₁;
  2. P₀₁;
  3. P₁₁(X^*), as well as the function D which allowed us to distinguish the reliability for two inspection techniques.
Depends on the goal of the techniques and the costs of testing mistakes the indexes could be used in such a ways: P₁₁; P₁₁(X^*) are preferable in the case of the cost of false negative indications is much more than the cost of false calls; D - if mentioned costs are comparable; P₁₁ and P₀₁ - if a customer would like to use the new technique instead of old one which detectability is quite good.

In order to check the above mentioned "practical" validation procedure we have compared the techniques of manual (M) and automated (A) ultrasonic testing of welds. The test pieces contain an artificial reflectors such as vertical cylinders and segments. To proof the adequate reflection properties of artificial reflectors the experiment (fig. 2a, b) have been done. Practical assessment carried out by personnel of UT level II (EN473) qualification. Statistical assessment has been carried out in agree with procedure described in [1]. The results are:
P₁₁^A > P₁₁^M ; P₁₁^A(dL^*) > P₁₁^M(dL^*); P₀₁^A > P₀₁^M; D^A ~ D^M.
The evaluation of the results of validation is based on the values of reliability indexes and depends on practical requirements for the techniques.

Fig 2:

Conclusion

The main characteristics of described "practical" procedure of validation are:
• test pieces with artificial reflectors;
• blind trial;
• appropriate qualified personnel;
• statistical reliability indexes.

References

1. 1. Nockemann C, Tillack G.-R, Wessel Y., Hobbs C., Konchina V. Performance demonstration in NDT by statistical methods: ROC and POD for ultrasonic and radiographic testing. - 6-th European Conference on NDT. - Nice, 24-28 Oct., 1994, 1, p. 37-44.

© AIPnD , created by NDT.net

|Home|    |Top|