DRAFT PAPER MODELING AND VISUALIZATION FOR IMAGING OF SUBSURFACE DAMAGE

Transcription

1 7t Middle East NDT Conference & Exibition Gulf International Convention Center, Gulf Hotel Manama, Kingdom of Barain September 13-16, 2015 MODELING AND VISUALIZATION FOR IMAGING OF SUBSURFACE DAMAGE Neil Goldfine 1, Scott Denenberg 1, Brian Manning 1, Zacary Tomas 1, Raseed Al Rusaid 2, Frederick Haugt 2 1 JENTEK Sensors, Inc., Clematis Avenue, Waltam, MA Al Rusaid Tecnologies Co., Al Turki Business Park, Office Villa #4; 7244 King Saud Road, Ad Doa Al Janubiya, Daran 34455, Kingdom of Saudi Arabia ABSTRACT. Tis paper describes te HyperLattice database metod for visualizing and analyzing eddy current NDT data, specifically for imaging of subsurface damage in piping and vessels. HyperLattice databases are pre-computed databases of sensor responses. Tese databases are generated using a forward model of te interaction of te sensor magnetic fields wit a multiple layered material under test. Over te past tree decades, initially at te MIT Laboratory for Electromagnetics and Electronics Systems (LEES) and later at JENTEK Sensors, Inc., te autors and oters developed forward models and inverse metods tat enable rapid imaging of multiple properties simultaneously. Te HyperLattice databases provide a unique tool for visualizing multiple unknown property spaces and building intuition tat is valuable for sensor and measurement procedure adaptation as well as for calibration and data integrity verification. Tis paper describes te HyperLattice-based inverse metods, and provides a few case studies on subsurface damage imaging. Tese case studies include detection and sizing of ydrogen blisters under weld cladding overlay for refinery vessels and corrosion imaging troug insulation or fireproofing. Te focus of tis paper is on te inverse metods and te key elements needed to perform practical measurements on piping and vessels. INTRODUCTION Te innovation described in tis paper is te use of precomputed databases of sensor responses (called Grids, Lattices and Hyper-lattices) to perform Multivariate Inverse Metods (MIMs). Tis sounds teoretical, but it as uge practical implications and enables improved performance for eddy current testing (ET) of layered material constructs suc as pipelines, coating systems and aircraft structures. 1

2 First, in order to use tis metod it is necessary to meet two difficult callenges: (1) design an ET sensor winding construct tat can be accurately modeled to enable accurate prediction of te sensor response over te range of Material Under Test (MUT) properties of interest; and (2) develop an instrument tat provides extremely accurate and sufficiently precise measurements over te frequency and impedance range of interest for te ET sensor and te MUT properties. One sensor construct tat meets tis requirement is te (wit inductive sensing elements) or te MR- (wit magnetoresistive sensing elements). Tese sensors bot use rectangular drive conductors wit long linear drive segments and a row of sensing elements placed at a selected distance (λ/4) from one of te linear drive segments. Note tat wen te sensing elements are placed at te center of a simple dual rectangle drive construct, ten λ is te spatial wavelengt of te primary mode of te magnetic field created by tis drive winding construct. Tis spatial wavelengt is proportional to te magnetic field dept of penetration at lower frequencies, wile te dept of penetration is inversely proportional to te drive current frequency at iger frequencies. For more on MWM sensor constructs refer to references [1-7]. Material Under Test (MUT) Layups Tis paper considers a progression of more and more complex MUT layups, beginning wit simple infinite alf-spaces and foils. Figure 1 sows tis progression, wit simple scematics tat represent te sensor (sown above te MUT as a dased line wit a liftoff gap between te sensor and te MUT). Te caption of Figure 1 lists all te MUT layups considered in tis paper. Note tat eac of tese layups is assumed to be comprised of layers wit uniform electrical conductivity and magnetic permeability as a function of dept witin te individual layer. Note tat it is also possible to assume a linear variation in tese properties or oter more complex property variations, but it is typically not necessary for most NDT applications. One type of damage tat is of interest is general corrosion, wic reduces te tickness of a layer over a large surface area compared to te sensor s sensing element footprint (note tat te footprint is larger tan te actual size of te individual sensing elements and defines te region tat eac sensing element is sensitive to on te surface of te MUT). For damage suc as general corrosion, a simple uniform tickness layer model provides a sufficient representation of te damage. However, for local damage, suc as small corrosion pits and cracks, te damage is not well represented by a typical uniform tickness, layered media model for te underlying pysics. Tus, for small local damage, as described below, te damage will be represented as a perturbation (deviation or cange) from te simple uniform layer model. Tis perturbation approac is often sufficient for detection and even sizing of damage, if a correlation relationsip is developed between te perturbed sensor response and te damage caracteristic of interest, suc as crack dept. 2

3 (lift off) (permeability) (conductivity). (a) (b) Coating Cladding c, c, μ c Coating Coating layer (c) g V V Cladding (d) o i P P P wj, wj o i mes V V V Near Side Corrosion (e) Far Side Corrosion Insulation and Weater Jacket FIGURE 1. Scematics for (a) infinite alf-space; (b) conducting, non-magnetic foil; (c) single layer conducting and non-magnetic coating on a tick conducting and non-magnetic base material; (d) single layer conducting and sligtly magnetic coating wit a disbond or blister between te coating and a magnetic and conducting base material suc as for cladding overlay on an internal vessel surface; (e) corrosion, wall loss, for wit an insulating, non-conducting, coating on a conducting and magnetic single layer suc as a pipe or a vessel; (f - left) corrosion, or wall loss, for a magnetic and conducting layer, wit insulation and a weater jacket suc as for refiner piping; (f - rigt) corrosion, or wall, loss for an concrete layer wit internal wire mes - suc as for underwater pipeline weigt coat or fireproofing for vessel skirts. (f) Corrosion Under Fireproofing (wit Wire Mes) Te most important ting to understand is tat witout suc a model, beavior from varying unknown properties suc as liftoff or insulation tickness or pipe wall magnetic permeability will dominate te responses and te correlation to damage will be unreliable. Tis is te key. Using a simple model, we can remove te effects on te response caused by tings we don t care about, and focus on te response canges caused only by te damage of interest. However, under some circumstances, it is necessary to use more complex numerical models to represent te 3-dimensional defect geometry. Suc numerical models are not addressed in tis paper. Tese more complex models are used after te simple models enable detection and rapid data analysis for gross sizing. Ten a more complex model can be used to provide more reliable defect dept and extent sizing. 3

4 Grids, Lattices and Hyper-lattices for Multivariate Inverse Metods (MIMs) Precomputed databases ave been generated for eac of te configurations in Figure 1, using te simple uniform layer modeling metod. Tese databases are generated off line and used by te GridStation software to perform te MIMs. Te GridStation software uses a rapid solver (essentially a database searc metod) to find te solution for multiple unknown MUT properties of interest for eac position of te sensing elements along te inspected or monitored surface of te MUT. Figure 2 sows examples of Grids and Lattices for te two and tree unknown layups sown in Figure 1. Note tat Grids are for two unknowns and look like a curved seet of grap paper; Lattices are for tree unknowns and are simply a set of Grids. Eac Grid in te Lattice represents te same range for two selected unknown variables. For eac Grid, te tird unknown variable is eld constant. Tis tird unknown is ten varied from Grid to Grid. Hyper-lattices are difficult to represent visually, since tey are used to solve problems wit four or more unknowns. Te MIM can be performed on all unknowns simultaneously solving for tem by searcing te Grid, Lattice, or Hyper-lattice for te best solution, using a single frequency metod or a multiple frequency metod. Note tat eac frequency provides a magnitude and pase measurement (or complex real and imaginary part measurement) for te impedance (sensing element voltage divided by te primary current). Tus, for two unknown problems a single frequency is sufficient, since te pase and magnitude measurements eac provide te equivalent of one equation tat can be solved. But for tree of four unknowns, two frequencies are needed since we need eiter te same number or more equations (measurements) tan unknowns as in simple linear algebra problems. Note tat because te pysics equations are inerently nonlinear, we must use a searc routine to find te solution tus, by precomputing te solution spaces and searcing tem rapidly, we can solve complex problems very quickly. For problems wit more tan four unknowns, tree frequencies or more are needed. In te following sections, some brief descriptions are provided for a few important problems including imaging of Stress Corrosion Crack colonies troug coatings, imaging of corrosion under insulation (CUI) for piping or vessels, sizing of ydrogen blisters under overlay cladding inside vessels, and imaging of corrosion under fireproofing wit wire mes. 4

5 Pase (deg) (a) -10 Ta (%IACS) Increasing lift-off Increasing tickness Magnitude (H) (b) (c) (d) (e) (f) FIGURE 2. Grids and Lattices corresponding to scematics from Figure 1, (a) troug (f). Grids a(left) and b(left) were generated in 1989 [4,7]. Te b(left) grid is for conductivity and tickness. Before tis grid is used a iger frequency measurement provides te lift-off in a ierarcical 3-unknown inverse metod. 5

6 Stress Corrosion Cracking (SCC) Problem Description Figure 2a (middle and rigt) provides two different representations of a pipeline SCC colony mapping metod. In Figure 2 (middle) te MUT as been represented by a tick layer wit a constant magnetic permeability and te effective electrical conductivity is allowed to vary. Tis is a two unknown problem wit te unknowns being liftoff and electrical conductivity. Figure 2 (rigt) instead, assumes te electrical conductivity is constant and allows te magnetic permeability to vary as one of te two unknowns, along wit liftoff as te second unknown. Goldfine [4] sowed tat te magnetic permeability and te electrical conductivity cannot be measured independently at relatively ig frequencies (witout use of a variable bias field); tus, eiter of tese representations are sufficient for use of te perturbation metod for crack detection and sizing, described earlier. In a perturbation metod, te crack sows up as a cange in one of te unknowns. Tis works best if te cange is primarily in one unknown of te uniform layer representation. However, for cracks, te estimated liftoff value is also affected for relatively open cracks. Tigt closed cracks ave a smaller effect on liftoff. As sown in Figure 3, te magnetic permeability response can also be correlated very well wit crack dept and tis can be displayed as a C-Scan image of dept for eac crack or as a B-Scan for te individual cannels. Figure 4 sows examples of crack maps tat matc very well wit te actual crack morpology sown also in te Figure. FIGURE 3. (left) Data sowing magnetic permeability response can also be correlated very well wit crack dept; (rigt) displayed as a C-Scan image of dept for eac crack or as a B-Scan for te individual cannels. 6

7 FIGURE 4: (left, top and bottom) crack response C-Scan images. (rigt, top and bottom) corresponding enanced potograps of te crack morpology. CUI (Wit and Witout a Weater Jacket) Figure 5 and Figure 6 provide results for corrosion under insulation imaging witout and wit a weater jacket, respectively. For applications witout a weater jacket, tis is a simple tree unknown problem, if we assume te electrical conductivity of te pipe wall layer is constant. If we add te weater jacket, ten tis becomes a five unknown problem. Wen we solve five unknowns we typically use a ierarcical metod, were we solve for a subset of te unknowns first. FIGURE 5. Images of lift-off (wic is te same as te coating tickness in tis case) and magnetic permeability for a riser wit a relatively insulating coating. Bot te lift-off and te magnetic permeability images sow te defect, even wen te insulating coating tickness is varied witout informing te metod. 7

8 Note tat te use of liftoff for te response in tis metod requires careful maintenance of relatively constant sensor proximity to te coating, since any variation will sow up in te lift-off response. FIGURE 6. Pipe wall tickness images are sown above for imaging of corrosion troug weater jacket and insulation. (top) Insulation tickness is 2 inces, te weater jacket is 0.5 mm tick aluminum, and te pipe walls are 0.25 inces and te corrosion is external. (Bottom) Insulation tickness is 2 inces, te weater jacket is 0.5 mm tick aluminum, and te pipe walls are 0.5 inces and te corrosion is internal. Hydrogen Blister Imaging for Overlay Cladding and Base Material Crack Detection One callenging application is te imaging of disbonds between overlay cladding and te base material for large vessels. To solve tis problem te disbond is modeled as a simple gap and tis gap is estimated as one of te unknowns. Details for tis application are provided in a complementary paper at tis conference [5]. Tis problem can be solved as a tree unknown problem for liftoff, cladding tickness and gap; if we assume tat te cladding conductivity and magnetic permeability are constant and known, and tat te base material conductivity and permeability are also constant and known. For te more complex problem were te goal is to detect cracks in te base material, te base material magnetic permeability can be added as a fort unknown. Corrosion under Fireproofing (CUF) or Weigt Coat CUF imaging or corrosion imaging troug weigt coat for underwater pipelines is also of interest. Tis problem adds te complexity of aving a magnetic layer (te wire mes) inside an insulating (non-conducting) layer of concrete. Tis problem can be solved by simply estimating te magnetic permeability of te wire mes and ten treating it as a constant, if we can assume we know te approximate position of te mes witin te concrete layer. One way to accomplis tis is to assume tat te nominal vessel or pipe wall tickness is known at a few locations and ten use tese locations to estimate te wire mes properties and position, first. Tis as been demonstrated successfully bot in te laboratory and in te field, and is also described in a complimentary paper at tis conference [6]. 8

9 SUMMARY Tis paper provides a brief overview of te visualization value provided by Grids and Lattices for complex multiple unknown problems. Scematic representations ave been described for problems tat range from a simple foil, to a pipe wit insulation and weater jacket or a vessel skirt covered by concrete wit wire mes. Te key is to identify te unknowns of interest and to solve tem rapidly wit a MIM. Te MIM relies on aving a sensor tat as a response tat can be accurately predicted over te range of MUT properties of interest and an instrument tat can provide accurate measurement of impedance at multiple frequencies over impedance range of interest. Tis metod as numerous practical applications, suc as SCC crack dept measurement, internal and external corrosion imaging, and ydrogen blister volume estimation. Tis metod is in use wit te and MR- sensors, providing reliable services for a range of applications. REFERENCES [1] Denenberg, S.A., T.M. Dunford, N.J. Goldfine, and Y.K. Seiretov, Metod and Apparatus for Inspection of Corrosion and Oter Defects Troug Insulation, US A1, 16 May [2] Sclicker, Darrell E.; Goldfine, Neil J.; Wasabaug, Andrew P.; Walrat, Karen E.; Say, Ian C.; Grundy, David C.; Windoloski, Mark.; Test Circuit Having Parallel Drive Segments and a Plurality of Sense Elements, US Patent # 7,049,811 B2, May 23, [3] Goldfine, Neil J.; Zilberstein, Vladimir A.; Sclicker, Darrell E.; Grundy, David C.; Say, Ian C.; Wasabaug, Andrew P.; Hig Resolution Inductive Sensor Arrays for Material and Defect Caracterization of Welds, US Patent # 6,995,557 B2, Feb. 7, [4] Goldfine, N. J., Uncalibrated, Absolute Property Estimation and Measurement Optimization for Conducting and Magnetic Media Using Imposed -k Magnetometry, Sc.D. tesis, Department of Mecanical Engineering, MIT, September, [5] Goldfine, N., Seiretov, Y., Manning, B., Tomas, Z., Dunford, T., Denenberg, S., Al Rusaid, R., Haugt, F., Bayangos, J., Minaci, A., Inspection of Steel Vessels wit Cladding Overlay, using Tecnology 7t Middle East NDT Conference & Exibition, Manama, Kingdom of Barain, September 13-16, [6] Goldfine, N., Manning, B., Tomas, Z., Seiretov, Y., Denenberg, S., Dunford, T., Haque, S., Al Rusaid, R., Haugt, F., Imaging Corrosion under Insulation and under Fireproofing, using MR-s, 7t Middle East NDT Conference & Exibition, Manama, Kingdom of Barain, September 13-16, 2015 [7] Goldfine, N., Magnetometers for Improved Materials Caracterization in Aerospace Applications, Materials Evaluation, Vol. 51, No. 3, p ; Marc