FIELD OF THE INVENTION

The present invention concerns method and apparatus for the analysis of scans of corrosion to reconstruct the original surface before corrosion took place.

BACKGROUND TO THE INVENTION

Corrosion affects materials, vehicles and structures in many areas of engineering, science and industry. The effect of corrosion on systems, vehicles and structures is usually negative leading to an eventual need for their repair, replacement or decommissioning. In many applications, corrosion has a high negative impact on the value of the application. Corrosion can often render a system, vehicle or structure unsafe, dangerous to human life or detrimental to the environment. For these reasons and others, the measurement and analysis of corrosion has become a significant industry.

Most types of value reducing and unsafe corrosion are non-uniform and are localised into areas of the surface of the system, vehicle or structure. These types of corrosion include general, erosion, crevice and pitting. Areas of non-uniform corrosion may be detected by visual inspection or automated systems such as pigs that travel through underground pipes. The areas of severe corrosion are usually confined to a relatively small percentage area of the surface of the system, vehicle or structure. This means that the detailed measurement and analysis of corrosion can usually be confined to the same relatively small percentage area of the surface of the system, vehicle or structure.

Surface corrosion is characterised by the degradation of material on the surface. Such degraded material is often removed by fluid action on the surface, but can also be removed by manual methods such as a wire brush or a powered abrading tool to reveal the surface below.

Corrosion measurement is most often carried out by simple contact means such as using a depth gauge to measure the depth of a pit. These contact means are relatively inaccurate. Over the last 15 years, non-contact means have been disclosed that have the advantage over contact means of increased accuracy. Typically, optical depth measurements are in surface grids with point spacings from 30 to 1000 microns and depth resolutions of 0.1-10 microns. There are both automatic and manual non-contact measurement means. Li US 5,362,962, Barborak revealed an automated motor driven machine that can scan along a structure such as

a pipeline using a laser stripe and cany out some basic corrosion analysis. This machine has drawbacks in terms of complexity, cost, size and setup time. In US 6,611,617, Crampton revealed a manually operated laser stripe scanning system that is simple, quick and comparatively light. This system has been manufactured and marketed in up to 30 countries by 3D Scanners Ltd (UK) and its distributors under the brand name ModelMaker since 1996. ModelMaker has been used for scanning corroded surfaces since before 2002. The output of these manually operated and motor driven optical systems is one or more scans containing many millions of points representing the corroded surface and usually including significant surrounding areas of relatively uncorroded surface.

Since before 2000 it has been possible to automatically mesh Corrosion scans to form a polygonal mesh in a process that involves one or more of the steps: a) meshing patches of points to form patch meshes; b) optimising the alignment of overlapping patches to each other; c) merging the optimised overlapping patches to form a single polygonal mesh; d) decimating the mesh to reduce its size without losing significant detail.

Corrosion scans may be in many forms before their analysis, including: unstructured point clouds, structured lines of points, 2D grids of depth points and polygonal meshes. The preferred form is a single polygonal mesh.

These corrosion scans are then analysed to produce information enabling an expert to make a decision as to the likely present/future usability and safety of the system, vehicle or structure. A typical analysis method involves:

(i) loading the corrosion scan into a standard surface data processing software package; (ii) converting the corrosion scan into the best form for the algorithms available in the software package;

(iii) manually constructing an approximation of the prior uncorroded surface from a geometric primitive such as a plane, cylinder or conical section, or a complex surface such as a NURBS surface; the approximation is referred to as a 'geometric surface';

(iv) aligning the geometric surface to the corrosion scan; the inventors of this invention have invented several methods prior to this invention including unwrapping a corrosion scan of a curved object such as a pipe onto a plane and automatically best fitting a floating plane geometric surface; the alignment usually requires a first step of approximate manual positioning, followed by an

automated step using a best fit algorithm that may or may not converge on a suitable minima and in any case has wide variability of resultant alignment of the geometric surface; (v) generating a difference map between the geometric surface and the corrosion scan; (vi) inspecting the difference map using standard visualisation tools such as colour depth plots and 2D sections;

(vii) (optionally) inputting the inspection results into a set of rules for determining operational parameters of the system, vehicle or structure such as safety, maximum operating speed, pressure, load.

Standard surface data processing software packages for steps (i) to (vi) have been available from 3D Scanners, Innovmetric (Canada), Geomagic (USA) since the early 1990s. Rules for step (vii) have been available as software algorithms such as RSTRENG from Technical Toolboxes Inc (USA) that has been available since 1989.

Unfortunately, the state of the art tools and methods for analysing corrosion scans to generate a difference map are manual and subject to significant human error, such that two experts analysing the same corrosion scans can produce widely ranging difference map results. Some drawbacks of the state of the art include: a) the approach of selecting a geometric primitive to represent the original surface before corrosion usually leads to significant errors. For example, a pipe might be manufactured to be oval in section rather than perfectly cylindrical and fitting a geometrical cylinder to a deformed pipe section is by definition imperfect; b) when the alignment process is performed by an operator, it is subject to human error; c) when the alignment process is performed automatically by an algorithm such as best- fit cylinder or unwrapping and floating plane best fit, best-fit operations usually iterate to one of several minima that is dependent upon one or more of the starting alignment and the best-fit algorithm; d) manual creation and fitting of NURBS surfaces is a complex, time consuming and skilled manual process; e) the current manual approaches are manually intensive, highly skilled, expensive, slow and usually carried out off-site, whilst the industry in many cases requires reports within minutes of the scanning process taking place; current manual approaches typically take five to eight hours; f) current automatic processes are inaccurate and unreliable;

g) for a significant amount of corrosion scans, current automatic processes are perturbed by the presence of surface features such as welds; where two or more surfaces are joined by welds in the same corrosion scan, current automatic processes break down; h) the results of both manual and automatic processes are not easy to verify by an expert; there is considerable skill and judgement involved and customer confidence in the result is low.

The current market situation is that there has not been high adoption of optical corrosion scanning systems, largely because of the unreliability of the state of the art corrosion analysis tools and methods for corrosion analysis.

The stage in existing corrosion analysis systems that is unreliable is the method of analysing corrosion scans to reconstruct the original surface before corrosion and an apparatus for carrying the method out. This stage is covered by iii) and iv) in the typical analysis method described above.

The challenge for the corrosion measurement industry has become one of providing the method and apparatus for the stage of analysing corrosion scans to reconstruct the original surface before corrosion that are:

Accurate

Repeatable

Automatic

Not subject to human error

Unskilled

Immediate reporting

Quick and easy to verify by an expert

SUMMARY OF THE INVENTION

It is a purpose of this invention to provide a method of analysing corrosion scans to reconstruct the original surface before corrosion comprising the following steps: receive a corrosion scan; analyse said corrosion scan to determine areas of corrosion; remove said areas of corrosion; reconstruct the original surface before corrosion.

This method is disclosed for the cases of a continuous surface with corrosion and a discontinuous surface with corrosion.

It is a further purpose of this invention to provide an apparatus for analysis of corrosion scans comprising:

a computer readable means connected to the computer; whereby said computer readable means receives and processes a corrosion scan to reconstruct the original surface before corrosion.

It is disclosed below that these inventions meet the requirements of customers requiring analysis of corrosion scans.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which;

Figure 1 is a flow diagram for a corrosion scan analysis process;

Figure 2 is a flow diagram for corrosion area identification and segmentation;

Figure 3 is A flow diagram for feature area identification and segmentation;

Figure 4 is a flow diagram for sub-surface identification and segmentation;

Figure 5a is a rendered Imago of a corrosion scan of a section of a pipe;

Figure 5b is a depth image of a corrosion scan of a section, of a pipe that has been unwrapped;

Figure 6a is a depth image of a corrosion scan of a section, of a pipe that shows the identified areas of corrosion;

Figure 6b is the intermediate scan with flie identified areas of corrosion removed;

Figure 7a is a depth image of a corrosion scan of a section of a pipe that shows the identiøed feature areas;

Figure 7b is a depth image of a corrosion scan of a section of a pipe that shows the segmentation of the sub-surfaces;

Figure 7c is the restructured original sub-surfaces of a section of a pipe before corrosion;

Figure 8 is an apparatus for analysing corrosion scans;

Figure 9 ia a histogram of the standard deviations of each point in a range image.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

FIRST EMBODIMENT

The inventors of the present invention have invented and tested many different methods of automatically analysing corrosion scans. Each method is tested on a large database of corrosion scans to determine its accuracy and repeatability with the main acceptance criteria being acceptable accuracy and 100% repeatability on all the corrosion scans in the database. The method of the present invention arose from the process of inventing and testing many methods that did not achieve the acceptance criteria on a significant proportion of the database of corrosion scans. For example, the method of unrolling a corrosion scan of a pipe and best fitting a floating plane was invented and tried, but found to fail on an unacceptably large proportion of the database of corrosion scans.

Method of Analysis of Corrosion Scans

The first embodiment of this invention is a method of analysis of corrosion scans to reconstruct the original surface before corrosion. The following cases are considered:

1. Continuous original surface: surface corrosion on an original surface that is continuous and has a relatively large radius of curvature

2. Discontinuous original surface: surface corrosion on an original surface that is discontinuous and has features (disruptions) such as welds, edges or abutting surfaces

Case 1: Continuous original surface

In Case 1 of this first embodiment, the original surface before corrosion is reconstructed from an area of surface corrosion on an original surface that is continuous and is either a plane or has a relatively large radius of curvature throughout the area of the scan data. A relatively large radius of curvature can be typically defined as greater than 20 mm, although this method can work with scan data of corroded surfaces with a radius of curvature of less than 20 mm. This method assumes that either the scan data is directly applicable or that in a preparation step any discontinuities in the surface such as from welds have been removed, usually by manually trimming the scan data using user interface tools in the software package. This method further assumes that the scan data covers a sufficient area of non-corroded surface around the corroded area. A sufficient area of non-corroded surface can be typically defined as greater than 3 mm of non-corroded surface around the corrosion surface, although this method can work with scan data of non-corroded surface around the corrosion surface of less than 3 mm. A sufficient area of non-corroded surface is typically at least 50% of the area of the corrosion scan, but could be considerably less or considerably more.

Referring to Figure 1, in a first step 101, a corrosion scan is received. In a second step 102, the corrosion scan is analysed to identify areas of corrosion. This analysis process is described in a following section. In a third step 103, the identified areas of corrosion are removed from the corrosion scan to provide an intermediate scan. In a fourth step 104, the original surface as it was before corrosion is reconstructed. This reconstruction process is described in a following section.

The analysis process 102 in which a corrosion scan is analysed to determine the areas of corrosion is now described. Referring to Figure 2, in a first step 111, the change of surface normal between each neighbouring pair of polygons is calculated, hi a second step 112, the changes in surface normal are analysed to identify rapid random changes of surface normal in neighbourhoods. In a third step 113, the areas of corrosion are segmented.

Case 2: Discontinuous original surface

In Case 2 of this first embodiment, the original surface before corrosion is reconstructed from an area of surface corrosion on an original surface that is discontinuous and has features such as welds which join at least two original surfaces (hereinafter called sub-surfaces) that are unlikely to be in perfect alignment.

The same general method of Figure 1 is followed for both Cases 1 and 2. The method of Figure 1 is now described in more detail for Case 2.

The second step 102 of the method of Figure 1 is analysis and this is expanded in the second case as follows. Referring to Figure 3, in a first step 121, the change of surface normal between each neighbouring pair of polygons is calculated as previously disclosed. In a second step 122, the changes in surface normal are analysed to identify regular changes of surface normal in areas. In a third step 123, feature areas are segmented.

In Case 2, the two or more original sub-surfaces are typically joined by a feature such as a weld. A fundamental of methods of construction of systems, vehicles and structures is that sub-surfaces have a boundary at which there is either an edge or a join to another sub-surface that exhibits CO (physical continuity) but in general does not exhibit either Cl (slope continuity) or C2 (rate of change of slope continuity). It is therefore important to look for these boundaries in order to segment the original surface into its constituent sub-surfaces. Referring now to Figure 4, in a first step 131, features segmented in the previous step 123 of the process are further analysed to construct a system of continuous segmentation features between the original sub-surfaces. In a second step 132, the continuous segmentation

features are removed from the intermediate scan. Ih a third step 133, each segmented subsurface is separately labelled. Referring again to Figure 1, the intermediate scan can now be reconstructed by repeating the process of step 104 for each sub-surface.

The reconstruction process in which the original surface is reconstructed as it was before corrosion is described. Two ways of surface reconstruction are provided in following disclosure: interpolation and morphing. Both ways work well in practice. The scope of this invention is not limited to these two ways of surface reconstruction, but includes any way of surface reconstruction in which the original surface is accurately reconstructed as it was before corrosion. The two cases will be described by images of the results of the methods disclosed.

Case 1 Images

Referring now to Figure 5a, a typical Case 1 corrosion scan 201 is in the form of a polygonal mesh and is shown rendered. With a high density non-contact scan, the corrosion scan is accurate and repeatable enough to represent the surface of the pipe for the purpose of determining the amount of corrosion. This can be shown by carrying out repeated scans of the same area of corroded pipe and comparing them; the differences between the corrosion scans are small enough to be disregarded. Referring now to Figure 5b, the surface of the corrosion scan 201 has been unwrapped onto a plane and can be seen as a depth image. This image of Figure 5b is the example Corrosion scan for Case 1 as in method disclosed in Figure 1. Referring now to Figure 6a, the Corrosion scan 201 has been analysed and the corrosion areas 204 identified as dark areas. The corrosion areas 204 identified as dark areas in Corrosion scan 201 are removed to provide voids in the intermediate scan. Referring now to Figure 6b, the voids in the intermediate scan have been replaced by reconstructed areas 206 in a reconstructed original surface before corrosion 203.

Case 2 Images

Referring now to Figure 7a, a typical Case 2 corrosion scan 201 has been analysed and the feature areas 210 identified as dark areas. Referring now to Figure 7b, the feature areas 210 have been analysed to segment the surface into sub-surfaces 211. Referring now to Figure 7c, reconstructed areas 206 have been generated in a reconstructed original set of sub-surfaces before corrosion 203.

Verification, Editing and Repetition

The further steps of comparing the reconstructed surface(s) and the Corrosion Scan and automatically generating a report are well known. The types of graphical illustration and

algorithmic results in the report are also well known. The problem with the state of the art is for an expert provided with an automatic report, to assess whether the reconstructed surface using, for example a best-fit cylinder, is sufficiently accurate.

It is a purpose of this invention that the disclosed method coupled with standard inspection processes and reports will produce results that are both accurate and repeatable. However, a method such as the one disclosed herein, cannot be assured accurate for all possible surfaces with all possible perturbations, particularly if there is an error in the corrosion scan. It is therefore essential that the report enables an expert to decide whether the results are accurate. It is an object of this invention that the combination of repeatable results and standard reporting is sufficient for an expert to easily see where a perturbation of the Corrosion Scan has led to an inaccurate result. The method is as follows:

1. The expert verifies the inspection report and identifies an inaccuracy, usually due to a perturbation;

2. The expert manually intervenes to eliminate this perturbation using tools in a Standard surface data processing software package to output an edited Corrosion Scan;

3. The edited Corrosion Scan is automatically analysed and a report produced by the disclosed method;

4. The expert verifies the inspection report and if an inaccuracy is found, steps 2-4 are repeated, otherwise the inspection report is approved.

Detailed description of Case 2 for an approximately cylindrical pipe

There are many ways of providing a method of Analysis of Corrosion Scans for Cases 1 and 2 (Continuous and Discontinuous original surfaces). There now follows a detailed disclosure of a method for the case of an approximately cylindrical pipe with a discontinuous original surface.

Preparation

The Corrosion Scan is a typical output of the ModelMaker laser stripe scanning system supplied by 3D Scanners Ltd. It consists of an optimised and merged polygonal mesh representing the surface of the discontinuous cylindrical pipe. As a pre-process to the Corrosion Scan analysis, a geometric cylinder primitive is manually or automatically fitted to the Corrosion Scan. The accuracy of fit is not important, providing that the largest error is typically within 2 or 3 mm but it could be more. A range image consisting of a dense grid of depths between the Corrosion Scan and the geometric cylinder primitive is automatically generated in cylindrical coordinates. The cylindrical range image is then unwrapped into a

planar range image (Rl). This method of pre-processing based on an unwrapped range image is one of many methods that could be used and has been chosen because the resulting data structure is fast to process. This method of pre-processing can be easily implemented by a person skilled in the art.

Corrosion Scan analysis summary

The Corrosion Scan analysis may be summarised as the combined process of segmenting the surface into sub-surfaces and locating potential corrosion areas in each sub-surface.

The current process for analysis of an approximately cylindrical pipe with a discontinuous original surface consists of the following steps with a RI as input: calculate changes in surface normal in the neighbourhood for each point in RI create a histogram of standard deviations of surface normal perform rough segmentation of RI into good and disrupted/corroded surface which are labelled Areas of Interest clean Areas of Interest remove Areas of Interest reconstruct surface compare reconstructed and original surface separate corrosion and disruption Areas of Interest flood fill Areas of Interest segment disruptions into edges/welds and bumps convert bumps back to original surface boundary grow edges/welds to segment surface into sub-surfaces output segmented sub-surfaces with actual corrosion areas and reconstructed estimated sub-surfaces before corrosion

The segmenting into sub-surfaces is carried out by looking for areas that have the characteristics of disruptions such as an edge or a weld feature and finding their directions. The disruption directions are grown to become sub-surface boundary lines. Disruptions are found by comparing the normals over a neighbouring area and checking if their deviation is within or beyond what is considered a typical normal deviation for a continuous surface. Areas having non-continuous deviation are considered either disruptions or corrosion areas. They are further analysed to conclude which type exactly they are. A characteristic of the corrosion area is that it is below the surface formed by its boundary points (current method of recognition). Edges are identified by their normals being at a very steep angle to the pipe surface (current method of recognition). Welds are identified by being above the surface

formed by its boundary points (current method of recognition). Another characteristic of the corrosion is that it has rapidly and randomly changing normals within a small neighbourhood, which is a further criteria for distinguishing between corrosion and disruptions in the general case. The current methods of recognition disclosed above have been found to work but the invention is not limited to these methods and other methods of recognition could be used within the scope of the invention. A detailed description of the algorithms currently used to perform the above steps is now made.

Calculating Changes in Surface Normal

The input to the process of finding Areas of Interest is the planar range image (RI) of the preparation stage. The StdDev (StandardDeviation) value for each point in the RI is calculated. The StdDev for each point is calculated as a deviation of the angle between the Normal vector of the current point and the Normal vectors of the points in the surrounding local area with size (AnalysisAreaSize). The AnalysisAreaSize is an input parameter to the process. The preferred value of AnalysisAreaSize is for a square area of 5mm x 5mm which works out at 50 points x 50 points in a RI with 0.1 mm spacing. The analysis is centred on each point in turn. The output is a StdDev value for each point in the RI. Please check SC It is OK

Segmenting by histogram

Referring now to Figure 9, a histogram 401 of the StdDev values for the RI is created. The peak value 402 of the histogram is calculated (PHV). The assumption is that the good pipeline surface that is non-corroded and non-disrupted forms a significant percentage of the area and that the deviation of normals over that area is normally distributed. In other words this assumes that a significant amount of good pipeline surface is scanned around the areas of corrosion and disruption. As a rough guideline, 50% of the scan should be good areas surrounding the disruptions and corroded areas. We take the peak 402 of the histogram of deviations (PHV) and multiply it by two to give us the maximum deviation 405 of typical surface that is non-corroded and non-disrupted. The output is a rough segmentation of the RI into good surface 403 and Areas of Interest (disrupted or corroded surface) 404.

Fine-finding Areas of Interest

The next step is to carry out a cleaner segmentation of the RI into good surface and Areas of Interest. The rough segmentation by histogram produces Areas of Interest that are extended into the good surface by distance up to the AnalysisAreaSize parameter.. The fine-finding of Areas of Interest finds these good surface points along the boundaries of the Areas of Interest and marks them as good surface points. This is done in the following way:

1. for each boundary point calculating the Expected Surface Normal (ESN) by averaging the normals of good surface points in the neighbourhood, typically in the AnalysisAreaSize square (current method); the ESN can be calculated by using other methods, such as extrapolation of the good surface to that point.

2. comparing the ESN to the normal of the boundary point; if the difference is above a certain threshold, the boundary point remains the Areas of Merest else it is marked as good surface. The threshold can be either an input parameter or automatically calculated. The current approach is to use as threshold 3 times the average of the Stdandard Deviations of all good surface points (the StdDev is calculated for each point in Calculating Changes in Surface Normal above).

3. the process is repeated for any newly formed boundary point until there are none such

Separating Areas of Interest into Corrosion and Disruption Areas

The next step is to carry out a separation of the Areas of Interest into one of the two types: Corrosion and Disruption. This is currently implemented by the following method:

1. Marking all the points belonging to the Areas of Interest as Points For Reconstruction;

2. Creating a Reference surface, using the Surface Reconstruction method described below;

3. Comparing the Reference surface to the original RI and finding for each point if it is below the Reference surface (the point could be Corrosion or Disruption) or above it (the point must be in a Disruption);

4. For each point below the Reference surface, the normal is checked as previously disclosed and depending on its steepness the point is marked as either Disruption (very steep angle) or Corrosion;

At the end of this step, all the points in all the Areas of Interest are marked as either Corrosion or Disruption.

Using Disruption Areas to segment the surface into sub-surfaces

Disruptions include edges, welds and bumps (a term used to describe features on the surface of no particular importance). Both edges and welds form sub-surface boundaries, whereas bumps have a neutral role in both surface subdivision and the subsequent surface reconstruction. The process of surface subdivision includes the following:

1. Segmenting the Disruption areas into continuous Disruption entities

2. Identifying Disruption entities of type Bump entity

3. Finding sub-surface boundary regions in the non-Bump Disruption entities

4. Extracting boundaries from boundary regions and constructing a network of boundaries segmenting sub-surfaces

Segmenting Disruption Areas into continuous Disruption Entities

The segmentation is done by applying a standard flood-fill algorithm for finding continuous areas as will be understood by a person skilled in the art. The result is a set of Disruption Entities, each one isolated from all the others.

Identifying Bump Entities

Bump entities are Disruption entities with little direction. There is a minimum length:width ratio (5:1 is the preferred maximum) and if it is not reached it is considered that the entity is neither an edge nor a weld and the Disruption Area is marked as a Bump entity.

Finding sub-surface boundary regions in the non-Bump Disruption entities

In the general case the, non-Bump Disruption Entities could contain both Bump regions and Surface Boundary regions (a region is a part of an Entity). In order to segment Surface Boundary regions it is assumed that they form Thick Line Sections; it is also assumed that a number of Thick Line Sections can be joined together in a continuous region so that they form complex shapes such as two welds forming a T-shape. In order to find all Thick Line Sections the following method is used for each non-Bump Disruption entity:

1. All the boundary points of the non-Bump Disruption Entity are found and a search is made for the two boundary points with the greatest distance between them; the points must be such that most of the points between them belong to the same non-Bump Disruption Entity. This will result in finding the longest line section that could be formed within the non-Bump Disruption Entity.

2. Moving perpendicularly along the line section with a step of one point, Entity points on both sides of the line sections are counted. A histogram of the counts is created and the same approach with Histogram Peak Value as disclosed above is used to define a threshold for marking the Entity points on the sides of the line section as belonging to the thick line - if the count of the points is less than the threshold, the points are part of the Thick Line Section.

3. After finding the Thick Line Section it is checked to see if it has the characteristics of a Bump and if so all its points are converted to Bump; else all its points are marked as LINEi, where i is a counter of the think line sections extracted so far.

4. The process is repeated for the remaining points of the Entity from step 1 until there are no more points left

After all the non-Bump Disruption entities have been analysed the result is a set of N Lines

LINE ₀ ... LINE _N .

Constructing a network of Boundary Line Sections segmenting sub-surfaces

In this step, the input is the set of N Lines LINE ₀ ... LINE _N . For each LINE; the maximum length:width-ratio orientation is calculated and the line section going through the middle of the extents for this orientation is marked as a Boundary Line Section. Some of the Boundary Line Sections can be merged if they have almost the same direction (preferably within 5 degrees) and if the distance between them in direction perpendicular to them is small enough (preferably within 5 mm). All resulting Boundary Line Sections are extended to the edges of the scan or to an intersection with another Boundary Line Section, starting with the longest. The resulting Boundary Line Sections are assumed to be the network of boundaries segmenting the sub-surfaces.

Reconstructing a non-Corroded (Reference) Sub-surface or Surface

This reconstruction step can be applied to an unsegmented continuous surface or it can be applied to individual sub-surfaces. In both cases there are no discontinuous areas in the surface being reconstructed. The resulting Reconstructed surface is accurate. Unlike best fit approaches, the result is repeatable. This reconstruction step can also be applied to a Discontinuous surface to create an approximate Reconstructed surface.

AU points that are not good surface points are removed from the RI and the resulting holes are reconstructed in a way that represents the best guess of what the surface originally was. The current method of surface reconstruction is by interpolating all the removed points using the boundary points of the area they form and weights based on the distance to each point to the power of three, i.e. Interpolated Point = Sum (Boundary Point n, weight n = (1/ Distance To Interpolated Point) ^Λ 3), where the sum of weights is 1. Other methods for interpolating can be used, e.g. by extrapolating the non-removed surface inwards the area being reconstructed and then weight averaging again. Interpolation is the preferred approach of reconstructing a sub-surface but this invention is not limited to the approach of interpolation. This interpolation approach of surface reconstruction is sufficient for the current invention, but this invention is not limited to this approach.

Case 2 for any discontinuous surface with large radius of curvature

A method for the case of an approximately cylindrical pipe with a discontinuous original surface has been disclosed. This invention is not limited to this case or the detailed method disclosed based on a RI.

The general case is a discontinuous surface with large radius of curvature in each subsurface. This surface can be automatically provided as a dense mesh. This mesh can be automatically tessellated into a large number of small polygons such that the vertices of the 3D tessellated mesh are at approximately uniform distance from each other if the distance from the large radius of curvature surface is ignored. The main difference between the 3D tessellated mesh and the 2.5D RI of the previously disclosed method for cylindrical pipes is that the tessellated mesh vertices are not on a 2.5D square grid. It will be understood by a person skilled in the art that the method followed for the 2.5D RI of the cylindrical pipe can be adapted to a 3D tessellated mesh and that all such methods fall within the scope of this invention. For example, in the step of surface reconstruction, a tessellated Reconstructed surface could be morphed across the 3D Corrosion hole instead of using the 2.5D interpolation approach disclosed above. The morphing of the tessellated Reconstructed surface could in the case of large radius of curvature, maintain CO, C1 and C2 continuity with the good surface surrounding the Corrosion hole. In the event of more complex surfaces such as manifolds, the problem is simplified by analysing smaller areas of corrosion by automatically splitting the surface into sub-surfaces on a topological basis whilst avoiding needing to split an area of corrosion.

SECOND EMBODIMENT

Apparatus for Analysis of Corrosion Scans

The second embodiment of this invention is an apparatus for analysis of corrosion scans to reconstruct the original surface before corrosion.

Referring to Figure 8, the apparatus for analysis of corrosion scans comprises a computer 301 and a computer readable memory 302 whereby the computer receives and processes a corrosion scan to reconstruct the original surface before corrosion.

The apparatus for analysis of corrosion scans can be used by at least salesmen, application engineers, R&D personnel, service companies, end-user customers, teaching staff and students. Data and Databases of analysed Corrosion scans can be produced from the disclosed apparatus by means of the disclosed method.

INDUSTRIAL APPLICATION

The method and apparatus according to the invention can be used in a variety of applications wherein corrosion scans need to be analysed to reconstruct the original surface prior to corrosion.

The corrosion scans may be of surfaces of systems, vehicles and structures. Systems include but are not limited to: engines, turbine blades, process plant, reactors, vessels, pipelines. Vehicles include but are not limited to: aircraft, spacecraft, cars, lorries, ships, boats, motorbikes and bicycles. Structures include but are not limited to: bridges, buildings, platforms (onshore and offshore), cranes. The largest current markets for corrosion scan analysis are nuclear reactors and pipelines (primarily gas and oil). Corrosion scans may also be indirect such as corrosion scans of casts or other forms of solid copy taken of a corroded surface.

This invention may also be applied to other types of surface deformation than corrosion. Examples of other types of surface deformation include but are not limited to abrasion, erosion, dents arising from impacts with other objects, buckles, bends or deformations arising from applied bending moments or pressures, surface cracks from fatigue, thermal stresses and localised heating such as from welding.

It is evident that there are numerous other applications of this invention and the scope of applicability of this invention is not limited to analysis of corrosion and surface deformation.