Modeling of Nde Reliability: Development of a ‘pod-generator’

NDE in an in-service situation is used as a basis for integrity management as well as planning of future and maintenance actions. In such an environment it is of paramount importance that the performance level of the NDT method(s) used in terms of Probability Of Detection (POD) is known. It is widely acknowledged that there is no such thing as ‘THE POD’ of an NDE method or technique. This is because the POD does not only depend on the physics of the NDE method and the procedure used, but also on factors such as the degradation mechanism, the geometry of the component, accessibility, presence of coating or insulation, scope of sampling and personnel qualification and experience. A number of industries in The Netherlands, among which chemical plant and pipeline owners as well as NDE service companies have initiated a Joint Industry Project to develop a model and corresponding software tool (‘POD-generator’) to assess the inspection effectiveness in a specific situation. TNO (institute of technology) is leading the project that has been launched in October 2003 and will take 3 years. The focus in the project is on the inspection of pipelines and piping, newly developed and conventional NDE techniques as well as a number of specific degradation mechanisms. The paper will deal with the results of the project gained so far. Introduction: With the introduction of Risk Based Inspection methodologies and similar prioritizing strategies for in-service asset management, the need to quantify the effectiveness of inspection procedures is greater than ever. The shift from intrusive inspection to non-intrusive inspection has compounded this need because it entails the loss of visual inspection as a tool. The performance of the non-destructive inspection is of critical importance to the management of the integrity because non-destructive inspection is increasingly the inspector’s only ‘sense’. To devise the correct inspection specification, knowledge is required of the performance of nondestructive inspection and the factors that influence it. An inspection specification must be tailored to the specific application. For in-service inspection this means tailoring to the specific degradation mechanism and the way of managing the integrity. Inspection performance is a key factor in controlling the structural integrity of piping and pipelines. At the end of 2003 a joined industry project has been initiated in The Netherlands with the objective to develop a comprehensive model for the assessment and optimization of a certain inspection approach. This model can be subdivided in three models, the inspection model, the degradation model and the integrity model. The main output of the degradation model is a surface containing modeled defects. The output of the degradation model serves as input of the inspection model. The output of the model is a probability of detection curve (POD-curve) for a specific situation. The main advantage of modeling POD-curves instead of measuring them is the reduction of time and cost. However this requires accurate and well-validated models. The POD-curve(s) from the inspection model and the output of the degradation model serve as input for the integrity model. This model calculates the probability of failure. Currently TNO is developing these three models. In this paper we will explain the contents and the mutual relation of these models in more detail and we show the progress that was made over the last nine months. The ‘POD-generator’: The so-called POD-generator is a model, which allows the assessment and optimization of an inspection program for in-service assets. It is recognized that for an optimal operation of in-service assets, issues like degradation processes, inspection performance and structural integrity are all related and therefore should not be treated as separate issues. Within the ‘POD-generator, there are three models, the degradation model, the inspection model and the integrity model. The degradation model predicts the initiation and growth of defects. With this project, the focus will be on uniform corrosion, localized corrosion and corrosion under insulation. The inspection model simulates the performance of an inspection method. It receives information from the degradation model in the form of defects. Information about the inspection performance is passed to the integrity model. This model predicts the probability of failure. These models are all interrelated, which we will illustrate on a simple example. Obviously, the inspection method depends on degradation mechanisms. The frequency and extent of an inspection depends on the consequences of a possible failure. During an inspection defects will be detected and action will be taken, if required, to repair these defects. This will improve the structural integrity. However for proper assessment of the integrity of an asset, knowledge about the performance of an inspection method is required for that given asset under inspection. Properties like the detection probability of (critical) defects quantify the performance of an inspection method. This property is commonly expressed in the form of a POD-curve (Probability Of Detection curve). After each inspection, information is obtained about the growth and extent of different types of corrosion. This allows an update of the degradation model. The degradation model predicts the evolution in time of the defects. This information can be used to evaluate the probability of failure as function of time. It may for example turn out that the evolution of defects progresses faster than expected. Therefore it may affect the inspection interval, to maintain an acceptable level. The objective is to develop a software package, which incorporates these three models. The user can use the software to evaluate different scenarios to evaluate or to optimize an inspection approach. The degradation model: The degradation model describes the shape of a corroded surface in the form of a surface profile as function of lateral (xy) coordinates. Currently the main focus is on uniform and localized corrosion on the inside of pipelines and piping and on corrosion under insulation (CUI) and other degradation on the outside of pipelines and piping (e.g. supports). The method could in principle also be applied on other forms of corrosion as well, but this is beyond the scope of the current project. In order to describe initiation of degradation and subsequent movement of a degradation front into a material, a descriptive modeling can be used to describe the evolution of the corrosion process as function of time. A Monte Carlo type calculation will be needed to come to a defect distribution, which can act as input for integrity and/or inspection models. The initiation stage is defined as follows: The probability that a surface cell A [m2] is initiated for the degradation mechanism in time step dt [s] equals Pi*dt [-]. The probability Pi [s] depends on many parameters, like: temperature, medium, metal, surface condition, etc. But it depends also on the fact if a neighbouring surface cell is already initiated or not. Now, if the surface cell is initiated, the propagation starts. The propagation can be modeled like, (i) a growth rate da/dt [m/s] that occurs in every time-step after the initiation. The growth rate depends on parameters like: temperature, medium, metal, etc. But also the depth of the defect (a [m]), and possible the shape of the defect (da/dx [-]) can play a role. However it is also possible to model the propagation in a statistical way too (ii): the growth rate da/dt can be considered to consist of a probability Pp times a propagation step da, for each time-step. Upon fine-tuning such a statistical model, a typical surface geometry of a corroded surface can be simulated, and the time-evolution can be calculated. This model is in principle able to describe all possible degradation mechanisms origination from the surface. With some modifications the model can be adjusted to describe internal defects as well. The main development during this project will be aimed on “calibrating” the input-parameters for good descriptive and predictive value of the model. Inspection results might be incorporated in a feedback loop, such that the corrosion model is becoming more and more realistic with time. The model is capable of producing different types of “corroded” surfaces, depending on the input parameters. Therefore, the challenge is to define the appropriate input parameters that are applicable to the corrosion under consideration. Corrosion may range from pitting to uniform corrosion. Examples of five typical corroded surfaces are given in figure 1. The figure illustrates that the general features of these types of corrosion can be reproduced by properly adjusting the model parameters. Statistical model 3D-scan of corroded surface Photo of corroded surface Type I: Isolated pitting corrosion Type II: Overlapping pitting corrosion Type III: Severe overlapping pitting corrosion Figure 1: Examples of simulated corrosion types compared to real examples from the field. The inspection model: The inspection model simulates an inspection technique. The focus in this project is on modeling ultrasonic and radiographic inspections. For modeling ultrasonic inspection techniques several algorithms are available based on finite difference scheme’s [6,8,10] and the Kirchhoff integral [11]. Modeling of radiographic inspection techniques is done using a Ray-tracing algorithm [12]. The output of the inspection model is either a inspection result for specific situation or a PODcurve based on a sufficiently large number of simulations. Figure 2 summarizes the modeling approach; it basically consists of two steps: • Numerical simulation of inspection; • Generation of the inspection result. In the first step the physical experiment is being modeled. The output of this simulation could be a measured signal at a certain location by a specific sensor. This result itself is not yet an inspection result it requires a certain processing/interpretation. To obtain an inspection result, the inspection procedure should be applied and combined with certain detection criteria. After applying this procedure, an inspection result is obtained. This can either be the detection of a defect and/or characterization of a defect in terms of size, location and type of defect.