NDT 2010 Conference Topics Session 3B (5) Inspection Qualification Chairman M Mienczakowski 12.05 PICASSO - improved reliability inspection of Aeronautic structure through Simulation Supported POD Author - Nancy Maléo PICASSO is a research collaborative project registered in the Seventh Framework Programme of the European Commission and coordinated by Snecma (Safran Group). The aim of PICASSO is to build a new and original concept of simulation supported Probability of Detection (POD) curves based on Non Destructive Testing (NDT) simulation in addition to existing experimental data base. The POD method of analysis is generally used for quantitative NDT process capabilities assessment. NDT performances must be evaluated in order to ensure the desired level of safety and the POD curve provides a very useful method of quantifying the performance capability of an NDT technique. POD curves are thus used as a basis for: establishing design acceptance requirements, setting inspection maintenance intervals. Nowadays POD curves are obtained by expensive fully experimental campaigns and are even technically often impossible to realize. The objectives of PICASSO are : to have more accurate and reliable POD curves in order to get appropriate damage tolerance assessments for critical parts to address the potential risk of failure from induced anomalies (from material, manufacturing or service) within the Approved Life of the part, to overcome cost issue of the Probability of Detection campaign with Non Destructive Testing simulation techniques. This challenge is particularly important for metallic parts within the engine and aircraft industry. This paper presents PICASSO project : partners, objectives, overall strategy and progress.
PICASSO improved reliability inspection of Aeronautic structure through Simulation Supported POD Nancy Maléo Snecma (Safran Group), Villaroche site, Rond-point René Ravaud, Réau, 77550 Moissy-Cramayel, France Telephone : Tel : +33 1 60 59 91 12 / Fax : +33 1 60 59 75 84 e-mail : nancy.maleo@snecma.fr Abstract PICASSO is a research collaborative project registered in the Seventh Framework Programme of the European Commission and coordinated by Snecma (Safran Group). The aim of PICASSO is to build a new and original concept of simulation supported Probability of Detection (POD) curves based on Non Destructive Testing (NDT) simulation in addition to existing experimental data base. The POD method of analysis is generally used for quantitative NDT process capabilities assessment. NDT performances must be evaluated in order to ensure the desired level of safety and the POD curve provides a very useful method of quantifying the performance capability of an NDT technique. POD curves are thus used as a basis for: establishing design acceptance requirements, setting inspection maintenance intervals. Nowadays POD curves are obtained by expensive fully experimental campaigns and are even technically often impossible to realize. The objectives of PICASSO are : to have more accurate and reliable POD curves in order to get appropriate damage tolerance assessments for critical parts to address the potential risk of failure from induced anomalies (from material, manufacturing or service) within the Approved Life of the part, to overcome cost issue of the Probability of Detection campaign with Non Destructive Testing simulation techniques. This challenge is particularly important for metallic parts within the engine and aircraft industry. This paper presents PICASSO project : partners, objectives, overall strategy and progress. 1
1 Introduction PICASSO is a research collaborative project registered in the Seventh Framework Programme of the European Commission. The aim of PICASSO is to build a new and original concept of simulation supported Probability of Detection (POD) curves based on Non Destructive Testing (NDT) simulation in addition to existing experimental data base. 2 Technical context Most aeronautical components have to be designed on damage tolerance concept. The damage tolerance design assumes that aeronautical structures contain flaws. This means that the potential existence of component imperfections as the result of inherent material structure, material processing, component design, manufacturing or usage is recognized and this situation is addressed through the incorporation of fracture resistant design, process control and Non-Destructive Testing (NDT) inspections. Therefore, periodic maintenance intervals are defined by the design office on the base of maximum size of defect potentially present in the component and material properties (crack growth rate). This process requires to perfectly know the minimum size of indication that can be reliably detected during manufacturing and in-service inspections. This information is obtained through Probability of Detection (POD) campaigns (see MIL-HDBK-1823A (1) for recommendations concerning POD determination). A POD campaign consists in applying the NDT procedure associated to the part by several NDT operators on a large panel of samples with and without defects representative of possible damage on the part. Statistical studies are performed on NDT results. The objective is to determine the curve which calculates the probability to detect a defect type in the proposed configuration (probability vs. defect size). An example of POD curve is given hereafter (see figure 1). Probability of Detection 100% 90% POD(a) 50% POD(a) 95% 0% a 90-50 a 90-95 Size of defect :a Figure 1 : Example of POD curve Design offices generally use the minimum detectable defect size with a probability of 90% with a confidence interval of 95%. : a 90-95. 2
3 Innovation context Today s conventional methodology for POD campaigns is too much cost and time consuming and configurations equivalence, which could avoid new POD campaigns, are not available. In the framework of PICASSO project, the benefits of an innovative concept will be evaluated. This concept is the simulation supported POD based on using NDT simulation tools for future POD determinations. The simulation supported POD approach constitutes very promising solution for the evaluation of aeronautic NDT inspection. By incorporating simulated data to the experimental statistical campaigns it becomes possible to reduce the cost and time while increasing the reliability of POD evaluations. The use of quantitative and validated numerical models will allow investigating more representative samples population than the usual purely experimental POD campaigns, to cross more completely the effects of the different influential parameters and to reduce the a priori hypothesis done today on the statistical models. The table 1 hereafter illustrates the main advantages of the innovative simulation supported POD concept proposed by PICASSO over the traditionally used experimental approach: Table 1. Description of the main advantages of the innovative simulation supported POD concept over the traditionally used experimental approach Sample manufacturing NDT inspection campaign POD data exploitation Experimental POD Simulation supported POD - Difficult to manufacture real defects - simulation of complex defect geometries, easy positioning by modelling - Limited number of samples - Large number of samples - Limited number of defects - Large number of defects - Several operators needed - NDT procedure is respected - Respect of the NDT procedure (sometimes operators change parameters for a better detection) - Data collection - Find equivalence between artificial and real defects by modelling (in order to manufacture only easy defects) - Human factor will not be treated in the framework of this project - Automatic data exploitation can be implemented with simulation tools New POD configuration - Need to do a new POD campaign - Use simulation to find transfer function to find equivalence between POD configurations - Reduced POD campaign sufficient for a new configuration 4 Objectives The objectives of PICASSO are : to have more accurate and reliable POD curves in order to get appropriate damage tolerance assessments for critical parts to address the potential risk of failure from induced anomalies (from material, manufacturing or service) within the Approved Life the part, 3
to overcome cost issue of the Probability of Detection campaign with Non Destructive Testing simulation techniques. The main result of PICASSO will be the validation of the concept of simulation supported POD by initial realistic results, implementation and methodologies. 5 Partners involved in PICASSO project The technology breakthrough developments proposed in PICASSO rely on a cooperation of 14 organisations, comprising European leading engine manufacturers, the engine-industry supply chain, key European research institutes and small and medium enterprises (SMEs) with specific expertise. The consortium gathers 6 large partners from industry, 3 SMEs, and 5 from university and research organisations. In total, 5 countries (France, Germany, Poland, Sweden, United Kingdom) are represented, all of them being members of the European Union. Snecma is the coordinator of this project. The members of the PICASSO Consortium are listed here-under : Universities and Research Organisations : Bundesanstalt für Materialforschung und -prüfung / Nondestructive Testing/Radiology (BAM) Commissariat à l Energie Atomique (CEA) Chalmers tekniska hoegskola AB Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.v. Fraunhofer Institute for Nondestructive Testing, Dresden Branch (IZFP-D) TWI Ltd SMEs : Hoch Technologie Systeme GmbH (HTS) Phimeca Engineering S.A. Technic Control Industries : Snecma (Safran Group) EADS France (Innovation Works Department) (EADS) MTU Aero Engines GmbH Rolls-Royce plc Turbomeca (Safran Group) VOLVO AERO AKTIEBOLAG 6 Overall strategy of the work plan The PICASSO project has started on July 1str, 2009 and will last three years. It is organised in six work packages : four technical work packages (WP1 to WP4), one dedicated to management and coordination (WP5), and one to Exploitation and Dissemination (WP6). 4
The following diagram (see figure 2) presents the overall strategy of the work plan : Figure 2 : Overall strategy Initial input (WP1) Reliable NDT simulations models are based on initial input depending on material properties (noise, grain size, anisotropy,..), defect description (cracks, inclusions, ), and data from equipment (sensors, detectors).the initial real conditions of the component must be precisely known to determine representative modeling inputs. The results of the NDT inspections may be strongly affected in function of the material characteristics (grain size, noise, anisotropy ), the real shape and knowledge of the (un)expected defects, and data from equipment. The aim of this WP1 is to identify the case studies that simulation supported POD method will be develop against and generate all the necessary material property, defect shape and Non Destructive Evaluation (NDE) method data that will be needed as input into the simulation software. Objectives of WP1 : Selection of validation cases by the industrial partners Description of material properties, defects geometries (crack and volumetric defects) and equipment Description of the influential parameters Modelling - NDT simulation software (WP2) Simulation tools already exist in Europe for NDT techniques, as Ultrasonic Testing (UT), Radiographic Testing (RT) and Eddy Current Testing (ET). The figure 3 presents the simulations softwares involved in PICASSO : 5
Eddy Current Testing VIC-3D Owner : HTS NDT technic concerned : EC Simulation of EC technique VAC Owner : Chalmers / Volvo Simulation software of EC technique to build simulated data base. CIVA EC Owner : CEA List NDT simulation software which offers the user physics based models for EC technic. X-Ray Testing CIVA RT Owner : CEA List NDT simulation software which offers the user physics based models for the X-Ray technic. Analytical RT inspection simulation tool (artist ( artist) Owner : BAM Simulation of radiographic imaging. Modeling of the real inspection scenario by defining a virtual setup. Figure 3 : Simulation softwares involved in PICASSO Ultrasonic Testing Elastodynamic Finite Integration Technique (EFIT) Owner : IZFP-D Numerical 1-D, 2-D, and 3-D simulation modules for elastic wave propagation problems CIVA UT Owner : CEA List NDT simulation software which offers the user physics based models for UT technic. The simulations tools are adapted to aeronautical needs and their maturity is now enough to start working on simulation supported POD. Nevertheless, their performance must be improved to ensure the maximal reliability on simulation for POD calculations The simulation supported POD approach is based on the application of these quantitative NDT models which have to take into account the complexity of real situations, the influential parameters and their possible fluctuations. In particular, models have to be able to consider the phenomena induced by the complexity of parts and materials (texture, microstructure, structural noise ), the complexity of geometries, of defects, and the fluctuations coming from a large number of influential parameters. Objectives of the WP2 : Models will be highly improved by developers to achieve accurate, reliable and numerically efficient simulation tools. Simulation supported POD - POD methodology (WP3) WP3 is the core of the PICASSO project since it focuses on Probability of Detection (POD) issues and deals with the introduction of simulated data in POD studies. This WP will develop a simulation supported POD methodology which includes : the modelling of uncertainties in order to develop a methodology to generate appropriate simulated data including uncertainties (Design of numerical Experiment) ; the combination of experimental measurements and modelling data and the feasibility of delta-pod (transfer function approach - study of extrapolation procedure on defects response or NDT parameters in different contexts) ; the improvement of POD estimation models. Objectives of the WP3 : 6
The goal of WP3 is to provide common methodologies and tools for low cost POD determination using a simulation supported strategy. The major expected result is the development of a prototype POD software platform. Assessment of new POD methodology (WP4) In the framework of the WP4, the models will be tested to verify if the specificity of aeronautical metallic parts providing sensitivity of inspection is taken into account and if models are reliable enough for an intensive use for POD calculations. It will be performed using the validations cases defined in WP1 by the comparison of the experimental and the simulated data. WP4 will assess and validate the new simulation supported POD methodology developed in the framework of the WP3 by the comparison between experimental and simulation POD data. Objectives of the WP4 : WP4 has 4 objectives: Production of POD data ; Calculation and comparison of POD curves, conventional and simulation supported ; Capability criteria for simulation supported POD curves ; Recommendations and training curses. 7 Some illustrations of works performed in the framework of the technical WPs during the 1rst year 7.1 WP1 - Selection of validation cases by the industrial partners The validation cases have to be representative of materials and processes commonly used in aeronautical industry. The type of defects present in the validation cases should also correspond to real defect. - material : Titanium, Nickel, - processes : forging, casting, welding, machining,... - Types of defects : inclusions, fatigue cracks, welding defects, These validation cases should also covered the 3 NDT techniques addressed in PICASSO : Eddy Current, Ultrasonic and Radiographic inspections. The table 2 here-under presents the 10 validations cases selected by the industrial partners. Table 2. Selected validation cases 7
Validation case Industrial sponsor Ultrasonic inspection of Ti Billet MTU Titanium alloy Material Type of defects Technique Cylinders and spherical inclusions UT Radiographic inspection of forged Ti MTU Titanium alloy Pores RT Eddy current of bolt holes Turboméca Nickel base alloy Fatigue cracks ET Radiographic inspection of electron beam welds Turboméca Iron based alloy Tungsten inclusions Lack of fusion RT Ultrasonic inspection of forged Ti Snecma Titanium alloy Spherical inclusions UT Radiographic inspection of single crystal Ni turbine blades Eddy current of flat surface Snecma Rolls-Royce Nickel based super alloy Titanium Nickel alloy Steel Ultrasonic inspection of Ni alloy Rolls-Royce Nickel super alloy Subsurface defects Cracks Needles, equiax and spherical inclusions RT ET UT Ultrasonic PA inspection of electron beam welds EADS Iron based alloy Voids Lacks of penetration UT Rotating probe eddy currents of bore holes EADS Titanium alloy Steel Fatigue cracks ET 7.2 WP2 - Consideration of structural ultrasonic noise Due to the interaction of the ultrasonic waves with the microstructure a part of the energy is redirected. These phenomena (see figure 4) can cause significant loss in ultrasonic non destructive evaluation of highly scattering materials like titanium for example. Figure 4 : Illustration of structural noise on an ultrasonic C-scan image The simulation supported POD for ultrasonic inspection of engine discs needs the improvement of the modelling of ultrasonic noise induced by the microstructure of the materials. To simulate the structural noise of the materials in UT, two possible approaches can be addressed : External noise approach; Consideration of the microstructure of the material. External noise approach : 8
This kind of noise generation can be seen as a post-processing step following the real calculation. It allows a very fast calculation of a large number of noisy signals based on only one numerical simulation. The simulations presented here-after have been performed using EFIT software. The figure 5 here-under present the demonstrator set-up used : 45 Angle beam shear wave probe Metal sample Surface breaking notch / crack Angle mirror detection Figure 5 : Demonstrator set-up Different simulation runs have been performed with each time a different height (h) for the notch. For each height of the notch, the ultrasonic response of the notch have been computed (see on figure 6 the signal without noise ) and the same level of noise have been added on these signal (see figure 6 the signal with noise ). Without noise With noise h h/ = 5.4 h/ = 0.81 h/ = 0.27 Notch no more detected Figure 6 : Signal response for different height of the notch The figure 6 shows that : - the angle of reflection of the ultrasonic wave in the notch varies with the h/ ratio, - the amplitude of the notch echo decrease with the h/ ratio, - for a known level of the noise, under a certain value of the ratio h/, the notch can t be detected because its echo becomes lower than the noise. Advantages of the external noise approach : - noise generation separated from simulation - very fast computation - possibility to generate hundreds of noisy signals from one single simulation run 9
Disadvantages of the external approach : - needs calibration measurements - cannot be generalized to other frequencies and other materials Consideration of the microstructure of the material : In that case, the noise becomes an inherent property of the model itself and an analytical model computes the noise directly from material properties. Some US simulation tools like CIVA US already proposed a generator of noise based on an empirical method where the structural noise is modelled by a set of randomly distributed scatterers and is described by two parameters: the scatterers density ( ) and the variance of reflectivity distribution ( ). This nowadays generator of noise absolutely need a reference experimental measurement to determine and adjust the two parameters and and then to compute noise. To address this problematic, an analytical model is being developed in order to model the interaction of ultrasonic wave and the microstructure and then compute noise directly from material properties and avoid the reference experimental measurement actually needed to determine and adjust the two parameters and The figure 7 presents the work plan of this development: Initial Input : Micrographic analysis Backscattering coefficient : analytical model (, ) Computation of interaction of the incident ultrasonic field with a set of point-like scatterers (, ) Noise level Figure 7 : Structural noise development work plan The initial inputs are given by the WP1. They consist in experimental informations concerning the microstructure : - elastic properties : elastic constant, proportion of each phase, crystallographic orientation, relationship between the phases; - morphologic properties : size and shape of macrograins and colonies. The material model theory is available and the development of the model is in progress. The experimental validation will be performed using samples supplied in the framework of the WP1. The simulated measurements will be compared to the experimental ones in order to check the validity of the model. These samples are cubes extracted from a pre-machined forged Ti17 part. These cubes have different sizes and are extracted from different zones of the part in order to represent the different flow lines orientations. 10
The figure 8 presents an example of the ultrasonic noise and Signal to Noise Ratio (SNR) measured in one 20mm x 20mm cube. Three entry sides (A, B and C) have been tested. The horizontal blue lines on one surface of the cube illustrate the flaw lines. Figure 8 : SNR measurement on one 20mm x 20mm cube 7.3 Simulation supported POD (WP3) The use of probabilities expresses that the NDT result is made non-deterministic because of uncertainties. The possible sources of uncertainties are : the defect itself (ex : fatigue cracks), the NDT procedure, the material, the mechanical repeatability, the human appreciation, The simulation supported POD concept supposed the use of data issued from NDT simulation softwares as data for the determination of the POD curves. The outputs of the simulation tools are purely deterministic. The first challenge of this WP3 is thus : How to generate data with uncertainties with fully deterministic simulation tool? The figure 9 here-under illustrates this problematic. 11
Classical simulation output: deterministic S Targeted simulation output: accounting for uncertainty S S(a) = (a) Input parameters Deterministic X S(a) = (a) + a (M) a a Output Values Deterministic NDT simulation software Y Deterministic With uncertainty With uncertainty Figure 9 : Output of the NDT simulation sofwares 7.4 Assessment of the new POD methodology (WP4) The kick-off of this WP4 is planed in December 2010. 8 Conclusions The simulation supported POD concept improves the reliability of POD calculations by increasing the population of POD samples with defects and by enabling the simulation of sub-components configurations. The main result of the project will be the validation of the concept of simulation supported POD by first realistic results, implementation and methodologies. Its intensive use and the building of a future European standard on simulation supported POD will be the next step towards the entry into service of such tools. Acknowledgements The author would like to thanks the members of the consortium of PICASSO project who have participated to this paper. Reference 1. MIL-HDBK-1823A, Non-Destructive Evaluation system reliability assessment, USA Department of Defense Hadbook, April 7th, 2009. 12