GB2568536A

GB2568536A - Defect detection and correction

Info

Publication number: GB2568536A
Application number: GB1719239.4A
Authority: GB
Inventors: Cui Yuxing
Original assignee: GKN Aerospace Services Ltd
Current assignee: GKN Aerospace Services Ltd
Priority date: 2017-11-20
Filing date: 2017-11-20
Publication date: 2019-05-22
Anticipated expiration: 2037-11-20
Also published as: GB201719239D0; GB2568536B; WO2019097265A1

Abstract

Boundary detection for an object manufactured by 3D printing comprises receiving 21 an, IR or near-infrared, first image of a first layer of the object. A computer aided design (CAD) model representing the object is sliced 23 into slices, a slice corresponding to the first layer is selected 24 and output 25 as a second image having a border representing an edge of the object. The second image is correlated 26 with the first image and pixels in the first image corresponding to the border in the second image are identified 27 as representing the periphery. Pixels representing the boundary and outside the boundary may be removed 28 from the first image. A defect detection method in which the brightness of a pixel is compared to the average brightness of an image region may be applied to the first image. The AM method may be a powder bed manufacturing method such as electron beam melting or laser metal deposition. Also disclosed is a method of correcting a defect.

Description

DEFECT DETECTION AND CORRECTION

FIELD

This invention relates to a method and system for boundary detection for an object manufactured by additive manufacturing. It also relates to a method and system for defect detection for an object manufactured by additive manufacturing, and to a method and system for defect correction in such an object.

BACKGROUND

In additive manufacturing (AM), parts are produced from computer models divided into slices which are translated into layers that are built on top of one another. AM has considerable advantages over conventional manufacturing techniques. For example, the weight ratio of raw material to a finished part is much smaller for parts formed by AM techniques than for parts formed by techniques in which material is removed to form the object. By comparison to conventional manufacturing techniques, AM techniques also allow parts with relatively complex geometries to be built to a high level of precision. These advantages, and others, mean that AM is attractive for building parts such as aeronautical components and components of medical devices.

AM techniques also come with disadvantages, however. One such disadvantage is the appearance of defects in parts built with powder bed AM techniques such as electron beam melting (EBM) and laser blown metal powder techniques, also known as laser metal deposition.

In EBM, a layer of powdered material is spread across a build area. An electron beam is moved across the material to selectively melt the areas corresponding to a 2D slice of the part to be built. If the electron beam has insufficient energy at a particular point, the powder at that point may either not melt, or may not melt sufficiently to fuse to the material around it or to the layer below it. A hole is therefore formed in the layer. If the surface on which the layer of powdered material is spread is defective, a hole in the layer can also be formed. As successive layers are built, the hole can propagate through the object as material above the hole fails to fill it. This can lead to defects appearing in the form of tunnels or chimneys through the object. Such holes, and the defects that propagate from them, are known as lack-of-fusion (LOF) defects.

In laser blown metal powder techniques, metal powder is blown coaxially to a laser beam which is moved across areas corresponding to a 2D slice of the part to be built. The laser beam melts the powder to build up the part. As with EBM, LOF defects can also occur in laser blown metal powder techniques if the laser beam has insufficient energy at a particular point.

Another type of defect that can occur in parts built with powder bed AM techniques such as those described above is a keyhole defect. Keyholes form when the beam vaporises - rather than simply melts - powder at a certain point. The vaporised powder forms a bubble which results in rapid solidification, thereby forming a keyhole defect.

A further type of defect that can occur in parts built with powder bed AM techniques such as those described above is a shrinkage crack. Shrinkage cracks can occur when melted material solidifies, resulting in material contraction. If the contraction incurs a stress that is larger than the tensile strength of the material, the contraction will crack open the material, causing a shrinkage crack.

In fields in which AM is typically used, for example the manufacture of aeronautical components, the high standards required of components can mean that if defects are present in an object, the object cannot be used. As AM of an object can be slow (for example, in the order of days), and the materials used costly, wastage of parts due to the presence of defects is highly undesirable. Although some defects can be corrected once an object is built (for example by hot isostatic pressing), this adds a further step to the manufacturing process. Other defects, for example those that reach the surface of the part, may not be able to be corrected once the object is built. It is therefore generally desirable to be able to identify defects as an object is being built, either to enable their correction or to enable the build to be stopped before more time or material is wasted.

In some fields, and in some parts, a certain level of defects may be tolerated in a certain area of the object. For example, areas of an object that are not subject to stress in use may contain defects without affecting the safety of a device of which the object is a part. However, if it is known that defects are present in the object, but it is not known where those defects are located, the object cannot safely be used. Further, the presence of defects in a certain area of an object may indicate a particular maintenance requirement for the system or apparatus producing the object. Once again, if it is known that defects are present in the object without knowing where those defects are located, such a maintenance requirement cannot be diagnosed. It would therefore generally be desirable to be able to detect where defects are located in an object.

One method of monitoring the appearance of such defects in an object is to analyse images of the object taken as it is being built. In particular, certain AM systems include an infrared (IR) or nearinfrared (NIR) camera which can be used to take images of each layer after it has been formed and before new material is added. In such images, defects in the layer appear brighter than their surroundings (for example, in a greyscale image, they may appear grey or white rather than black). However since other areas of the layer, including its boundaries, may also appear bright (for reasons explained below), it is difficult to distinguish holes from these other areas. Figure 1 shows an area of a NIR image 100 of a layer of an object being manufactured by EBM, in which defects 110 and a boundary 120 appear brighter than the body 130 of the object and its surrounding powder 140. Attempts to detect defects in the layer based on determining pixels having a brightness (or greyscale value) above a certain threshold therefore fail, as the brightness of pixels representing a hole may be very similar to the brightness of other pixels that do not represent a hole. Of course, if the image analysed is a negative of the original image, then defects will appear darker than their surroundings, and the word bright in the above discussion can be replaced with the word dark.

One method of correcting defects during the AM process is to re-melt a layer, or part of a layer, in which a defect is detected, by applying a higher energy to the layer than is used to form the layer. This approach, however, adds to the time taken to build an object, since the method must wait for a defect to be detected and then carry out a further step on the layer in which it is detected before it can move on to manufacturing the next layer.

It is generally believed that if too much energy is supplied to the material being used to manufacture the object, it will swell, causing irreparable distortion to the object. Thus, a second method of correcting defects during the AM process is to incrementally increase the energy supplied to each layer built subsequent to a layer in which a defect is detected, keeping any increase in energy low relative to the energy supplied to the layer in which a defect is detected. For example, the energy supplied to the layer built immediately after the layer in which a defect is detected may be increased by 5% relative to the energy used to build the layer in which the defect is detected, and the energy supplied to the layer built after that may be increased by 10% relative to the energy used to build the layer in which the defect was detected. This approach, however, is not very effective in fixing detected defects. In other words, it has a fairly low probability of correcting a defect. An object of at least certain methods and systems disclosed herein is to address one or more of these problems.

SUMMARY

The invention is defined in the claims.

[First Aspect]

According to a first aspect of the present disclosure, there is provided a method of boundary detection for an object manufactured by additive manufacturing (AM), the method comprising: receiving a first, infrared or near-infrared, image of a first layer of an object manufactured by AM, the first image having a plurality of pixels; slicing a computer-aided design model representing the object into a plurality of slices, at least one slice corresponding to a layer of the object as manufactured by AM; selecting a first slice of the computer-aided design model, the first slice corresponding to the first layer of the object; outputting at least the first slice of the computer-aided design model as a second, computer-aided design, image having a border representing a boundary of the object; correlating the second image with the first image; and identifying as a pixel that represents at least part of a boundary of the object any pixel of the first image that corresponds to the border of the second image.

As discussed above, in an IR or NIR image of an object being manufactured by AM, defects may appear brighter than their surroundings. As the boundary of the object in such an image also appears brighter than its surroundings -the mechanism behind this is explained in more detail below with reference to Figures 9a and 9b - previous attempts to identify defects based on their relative brightness have resulted in false-positives; identifying defects in areas which in fact are only a boundary of an object. The above method, in allowing pixels representing the boundary of the object to be identified, therefore allows such false-positives for defects in a boundary area to be reduced. This can lead to more accurate subsequent detection of defects in an object, and more accurate determination of their location. In turn, this allows some of the disadvantages identified in the Background section, above, to be addressed. For example, wastage of time, energy and materials can be reduced if it is determined that the build should not continue, or that the defects can be corrected in-build, or that the defects are sufficiently minor or are located in an acceptable area of the object such that the object can be kept without further processing to correct defects.

[Further Steps]

The method may further comprise removing from the first image any pixels that are identified as representing at least part of the boundary of the object. The method may further comprise removing from the first image any pixels that are identified as being outside the boundary of the object, or any pixels outside the border of the second image.

This can mean that if the IR or NIR image is subsequently analysed for pixels representing defects, pixels that do not represent the object need not be analysed. This can lead to a reduction in processing time, or in processing power required to perform the analysis.

The method may further comprise receiving a calibration image and, after receiving the first image, correcting the first image based on the calibration image. The calibration image may be an image of a projection of an energy beam onto a base plate of an AM apparatus in a predetermined pattern.

By correcting the first image based on the calibration image, alignment of the first image and the second image can be improved, so that the substantially all pixels representing a boundary of the object can be identified. In this way, a subsequent determination of the presence and location of defects in the object need not determine pixels representing a boundary of the object as pixels representing defects.

[Terms Used]

The first image may be an image taken by an IR or NIR camera. The first image may be an image taken of the layer as the object is being manufactured.

The computer-aided design (CAD) model may be the model used to manufacture the object. The boundary of the object may be an edge of the object. The boundary of the object may be powder adjacent to the object. The boundary of the object may comprise powder adjacent to the object. The boundary of the object may comprise a surface of the object. The boundary of the object may comprise any surface that the object was designed to have. The boundary of the object may comprise any surface of the object that is represented by the CAD model. Slicing the CAD model into a plurality of slices may comprise processing a first file representing the CAD model to convert the model into a plurality of slices and outputting these slices as a second file. The plurality of slices may be a plurality of cross-sections of the CAD model.

The first slice corresponding to first layer of the object may be a slice from which the first layer is built. The first slice may be a slice having the same dimensions as the first layer.

Correlating the second image with the first image may comprise overlaying the second image on the first image. A pixel of the first image that corresponds to the border of the second image may be a pixel of the first image that overlaps the border of the second image.

[Second Aspect]

According to a second aspect of the present disclosure, there is provided a computer-readable storage medium storing instructions that are arranged, when executed by a computer, to cause the computer to carry out the method of the first aspect.

[Third Aspect]

According to a third aspect of the present disclosure, there is provided an additive manufacturing system arranged to perform boundary detection in an object manufactured in the system, the system comprising: an additive manufacturing apparatus arranged to manufacture an object by manufacturing a plurality of layers of the object; an infrared or near-infrared camera arranged to take a first image of a first layer of the plurality of layers of the object, the first image having a plurality of pixels; a boundary detection module communicatively connected to the camera and to the storage device, and arranged to receive the first image from the camera, the boundary detection module further arranged to: slice a computer-aided design model representing the object into a plurality of slices, at least one slice corresponding to a layer of the object as manufactured by additive manufacturing; select a first slice of the computer-aided design model, the first slice corresponding to the first layer of the object; output at least the first slice of the computer-aided design model as a second, computer-aided design, image having a border representing a boundary of the object; correlate the second image with the first image; and identify as a pixel that represents at least part of a boundary of the object any pixel of the first image that corresponds to the border of the second image.

[Pixel Removal]

The boundary detection module may additionally be arranged to remove from the first image any pixels that are identified as representing at least part of the boundary of the object. The boundary detection module may be arranged to remove from the first image any pixels that are identified as being outside the boundary of the object, or any pixels outside the border of the second image.

[Calibration]

The infrared or near-infrared camera may be arranged to take a calibration image. The apparatus may be arranged to project an energy beam onto a base plate of the apparatus in a predetermined pattern. The calibration image may be an image of the projection of the energy beam onto the base plate of the apparatus. The boundary detection module may further be arranged to receive the calibration image and, after receiving the first image, correct the first image based on the calibration image.

[AM Apparatus]

The AM apparatus arranged to manufacture an object by manufacturing a plurality of layers of the object may be a hot bed AM apparatus. In other words, the AM apparatus may be arranged to manufacture the object on a bed that is heated. The apparatus may be a powder bed AM apparatus. In other words, the AM apparatus may be arranged to manufacture an object from successive layers of powdered material. The apparatus may be an EBM AM apparatus. In other words, the AM apparatus may be arranged to manufacture an object by electron beam melting (EBM). The apparatus may be a laser blown powder (otherwise known as laser metal deposition) apparatus. In other words the AM apparatus may be arranged to manufacture an object by a laser blown powder technique (also known as laser metal deposition).

[Placement of Camera]

The IR or NIR camera may be axially displaced with respect to an energy beam used to melt a material to produce a layer of the object.

[Arrangement of System]

The additive manufacturing system may be an additive manufacturing machine. The AM apparatus, IR or NIR camera, and boundary detection module may all form part of the additive manufacturing machine.

In this way, images from the camera can be processed locally and in real-time in the hardware of the AM machine, to give immediate feedback on the build.

The AM apparatus and IR or NIR camera may be parts of an AM machine, and the boundary detection module may be separate from the AM machine.

The communicative connection between the IR or NIR camera and the boundary detection module may be any of: a wired connection, such as Ethernet, USB or Firewire; a wireless connection, such as Bluetooth, Wi-Fi, cellular network or infrared; or a manual transfer using a computer-storage medium, such as CD, DVD, Blu-Ray, Memory Card or USB flash drive; according to the requirements of the particular implementation.

In this way, data can be gathered from the AM machine and processed elsewhere.

The boundary detection module may comprise a processor. The processor may be arranged to execute arithmetic processing in accordance with a program or other types of instructions. The boundary detection module may additionally comprise a storage device. The storage device may comprise, for example, volatile memory such as RAM and/or non-volatile memory such as a hard disk, and may be arranged to store program information and/or program data.

[Fourth Aspect]

According to a fourth aspect of the present disclosure, there is provided a method of defect detection for an object manufactured by additive manufacturing, the method comprising: (a) performing the method of the first aspect; (b) selecting a region of the first image, the region consisting of a plurality of pixels; (c) determining the brightness of a first pixel in the region; (d) determining the brightness of each of the other pixels in the region; (e) calculating the average brightness of each of the other pixels in the region; (f) calculate the difference between the brightness of the first pixel and the average brightness; (g) repeating steps (c) to (f) for each of the pixels in the region; and (h) identifying any pixel having a difference in brightness from the average brightness above a threshold value as representing at least part of a defect.

As discussed above, in IR or NIR images of a layer of an object being manufactured by AM, defects in the layer appear brighter than their surroundings. Since, however, other areas of the layer may also appear bright, attempts to detect holes in the layer based on comparing pixel brightness to a threshold value are unreliable. In the above method, by comparing each pixel to the pixels surrounding it, defects can be more reliably identified than in methods relying on comparing pixel brightness with a threshold value. As discussed above, by more reliably identifying defects, and determining their position, time and energy can be saved by using the results to determine one or more of the following: that a build should not be finished as the severity and location of defects is unacceptable; that the identified defects can be corrected during the build; that although defects are present, their location is acceptable with regard to the object's intended use, such that it need not be discarded or subjected to further treatment; that the process used to manufacture the object needs to be modified to reduce defects in subsequent parts; or that the apparatus or system used to manufacture the object needs to be maintained to reduce defects in subsequent parts.

[AM Technique]

The infrared or near infrared image may be an image of a layer of an object being manufactured by hot bed AM.

In hot bed AM, the bed on which the object is built is heated and therefore the cooling of a layer after it is manufactured takes longer than in AM techniques which take place in a chamber at a temperature closer to room temperature. Further, some areas of an object will cool faster than others, based on the structure of the object. After a layer is built, some areas will therefore be than others, and there will be heat transitions across the object. Some areas will be brighter than others, even when those areas do not represent defects. Thus, it is very difficult to detect defects based on a cut-off, or threshold, for brightness, above which a pixel is considered to represent a defect, because such an approach will either miss defects (if the threshold is set too high) or produce false positives for defects (if the threshold is set too low). The above method, in which each pixel is compared to the pixels around it, and not to a set threshold value, is therefore particularly useful in hot bed AM manufacturing, although its use is not limited to hot bed AM.

The infrared or near infrared image may be an image of a layer of an object being manufactured by powder bed AM.

In powder bed AM, an object is manufactured by selectively melting a powdered material. Defects such as holes, keyholes and cracks are more common in such AM techniques than in other kinds of AM, and the above method is therefore particularly useful in powder bed AM, although its use is not limited to such manufacturing techniques.

The infrared or near infrared image may be an image of a layer of an object being manufactured by electron beam melting (EBM).

In EBM, an electron beam is used selectively to melt a material to manufacture an object. By comparison to AM techniques using a laser as the energy beam, such as laser metal fusion, it is harder in EBM methods to monitor where the energy radiation beam hits a material used for building an object. This is because in laser AM, unlike in EBM, reflected light from the laser can be used to determine where the laser beam has struck. The above method allows for the identification of defects, even where it cannot otherwise accurately be determined where an energy/radiation beam has or has not hit a material used for building an object.

The infrared or near infrared image may be an image that is taken by a camera that is axially displaced with respect to an energy beam used to melt a material to produce a layer of the object.

In AM methods in which the energy beam is not co-axial with the camera, it is more difficult to capture an image of the melt pool on the object than in methods which allow for the camera to be mounted behind the lens used to direct the energy beam. In AM methods in which the energy beam is not co-axial with the camera, therefore, there is an increased need for methods such as the above method which allow for the detection of defects in a layer (when compared with methods in which images of the melt pool can be captured while an object is being built).

[Order and Repetition of Steps]

Step (a) may be performed before steps (b) to (g).

By first identifying a plurality of pixels that represent at least part of a boundary of the object, and then performing the remaining steps of the method, less processing time or power may be required than if the boundary is not identified before defect detection is performed.

Steps (b) to (g) may be repeated for each pixel in the first image.

In this way, each pixel of the image can be analysed to determine whether it represents a defect.

Steps (a) to (g) may be repeated for a plurality of IR or NIR images, each IR or NIR image being an image of a layer of the object being manufactured by AM.

[Noise]

The method may comprise identifying as noise any group of a predetermined number or fewer adjacent pixels having a difference in brightness from the average brightness that is at or above a threshold value. For example, the method may comprise identifying as noise any group of seven or fewer adjacent pixels having a difference in brightness from the average brightness that is at or above a threshold value. The method may comprise identifying any group of a predetermined number or fewer adjacent pixels having a difference in brightness from the average brightness that is at or above a threshold value as not representing a defect. For example, the method may comprise identifying any group of seven or fewer adjacent pixels having a difference in brightness from the average brightness that is at or above a threshold value as not representing a defect.

The present inventor has found that defects are generally of at least a particular size. In the first image of the layer, therefore, groups of adjacent pixels that appear to represent a feature that is smaller than this particular size can therefore be determined not to represent a defect. These anomalous pixels may be caused by noise in the image, for example. By identifying such groups of pixels as not representing defects, the method allows for a reduction in false-positives in defect detection, giving more reliable defect detection than in methods that do not have this feature.

The predetermined number may be different according to the particular process and system used to manufacture the object, as well as the resolution of the first image and the distance from which it was taken. However, the present inventor has found that defects in parts built by one particular system using EBM are rarely so small that they are represented by seven or fewer pixels in an image taken of a layer by a near-infrared camera in the system. Thus, the present inventor has found that groups of seven or fewer pixels in images taken during this process can be determined not to represent defects.

[Number of Pixels in Comparison Region]

The region may consist of four pixels. The region may consist of nine pixels. The region may be square. The region may be bounded by four sides. Each side may be two pixels long. Each side may be three pixels long.

The present inventor has found that the accuracy of defect detection increases as the number of pixels in the region increases. However, the computing power required for analysis also increases as the number of pixels in the region to be analysed increases. Since, as found by the present inventor, the increase in computing power as the area of the region increases is greater than the increase in the accuracy of detection as the area increases, the present inventor has found that an area of 3 by 3 pixels represents an acceptable balance of detection accuracy and computing power required. This is as opposed to, for example, a region of 5 by 5 pixels, in which accuracy of detection is not greatly increased relative to an area of 3 by 3 pixels, whereas computing power required is almost tripled. If computing power is of secondary concern, the region can consist of conceivably any number of pixels.

[3D Modelling]

The method may comprise determining coordinates of each pixel representing at least part of a defect. The CAD model may be a first CAD model and the method may further comprise converting the coordinates of each pixel representing at least part of a defect into a second CAD model. The method may further comprise combining the first and second CAD models to produce a third CAD model of the object and the identified defects in the object. The second and third CAD models may be 3D models. The second and third CAD models may be models in a CAD file format.

Such a CAD model of the object with its identified defects can be visually inspected through a CAD file viewer. This not only makes decision-making based on the location of defects in the object more intuitive than if the defects were not displayed in a 3D model, but also allows for further elimination of false-positive identification of defects. In particular, artefacts that may appear in an image of a single layer to represent defects can show themselves in a 3D model not, in fact, to have characteristics of a defect. For example, artefacts in the image caused by dirt on the glass shielding a camera taking the IR or NIR images can appear as lines through the model because they appear in substantially the same place on each image.

[Terms Used]

The defect may be a hole in at least one layer. The defect may be a tunnel through a plurality of layers of the object.

The defect may be a lack of fusion (LOF) defect. A LOF defect may be a hole in at least one layer. A LOF defect may be at least a point in a layer of the object at which a material used to manufacture the object has not been heated to a sufficiently high temperature to melt. A LOF defect may be at least a point in a layer of the object at which a material used to manufacture the object has not been heated to a sufficiently high temperature to fuse to surrounding material. A LOF defect may be at least a point in a layer of the object at which a material used to manufacture the object has not been heated to a sufficiently high temperature to fill a void present in the preceding layer.

The defect may be a keyhole. A keyhole may be a hole in at least one layer. A keyhole may be at least a point in a layer of the object at which a material used to manufacture the object been heated such that it vaporises. A keyhole may be at least a point in a layer of the object in which a bubble is present.

The defect may be a shrinkage crack. A shrinkage crack may be at least a point in a layer of the object at which a material used to manufacture the object has cracked. A shrinkage crack may be at least a point in a layer of the object at which a stress has split the material.

[Fifth Aspect]

According to a fifth aspect of the present disclosure, there is provided a computer-readable storage medium storing instructions that are arranged, when executed by a computer, to cause the computer to carry out the method of the fourth aspect.

[Sixth Aspect]

According to a sixth aspect of the present disclosure, the additive manufacturing system of the third aspect is additionally arranged to perform defect detection in the object manufactured in the system.

In the sixth aspect, the system comprises a defect detection module that is arranged to: (i) select a region of the first image, the region consisting of a plurality of pixels; (ii) determine the brightness of a first pixel in the region; (iii) determine the brightness of each of the other pixels in the region; (iv) calculate the average brightness of each of the other pixels in the region; (v) calculate the difference between the brightness of the first pixel and the average brightness; (vi) repeat steps (ii) to (v) for each of the pixels in the region; and (vii) identify any pixel having a difference in brightness from the average brightness above a threshold value as representing at least part of a defect.

[CAD Model]

The defect detection module may additionally be arranged to determine coordinates of each pixel representing at least part of a defect. The CAD model may be a first CAD model and the defect detection module may further be arranged to convert the coordinates of each pixel representing at least part of a defect into a second CAD model. The defect detection module may further be arranged to overlay the second CAD model on the first CAD model to create a third CAD model.

[Visual Display]

The system may further comprise an output device communicatively connected to the defect detection module. The output device may be arranged to display at least part of the third CAD model. The output device may be a display module. When the AM apparatus and IR or NIR camera are parts of an AM machine, and the defect detection module is separate from the AM machine, the output device may also be separate from the AM machine. Alternatively, when the system is an AM machine, the output device may be part of the AM machine.

[Seventh Aspect]

According to a seventh aspect of the present disclosure, there is provided a method of in-build defect correction for an object manufactured by additive manufacturing, the method comprising: depositing a first layer of material; manufacturing a first layer of the object by supplying energy to the first layer of material at a first intensity; depositing a second layer of material; and, responsive to the detection of at least one defect in the first layer of the object, supplying energy to the second layer of material at a second, predetermined, intensity, wherein the second intensity is greater than the first intensity and is an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured by supplying energy to at least one further layer of material at the second, predetermined, intensity.

By supplying energy to the second layer of material at an intensity greater than the intensity of the energy supplied to the first layer of material (in which the at least one defect is detected), the second layer of material can be supplied with sufficient energy to prevent it from following the form of the at least one defect in the first layer of the object when the second layer solidifies. This prevents the defect from propagating through the layers of the object to form a tunnel. In some cases, by supplying energy to the second layer of material at an intensity greater than the intensity of the energy supplied to the first layer of material (in which the at least one defect is detected), the defect in the first layer of the object can be fixed, since the material of the first layer re-melts and fills the hole.

As mentioned above, it is generally believed that if too much energy is supplied to the material being used to manufacture the object, it will swell, causing irreparable distortion to the object. It would therefore be understood by a person skilled in the art that supplying energy to the second layer of material at an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured at this intensity would cause swelling. The present inventor has found, surprisingly, that this is not the case. In particular, the method of the seventh aspect can correct defects in a first layer of an object by supplying increased energy to only one subsequent layer of material. Although a relatively high intensity is used, counterintuitively, the method can reduce swelling in an object in which this method is used relative to in an object in which the intensity is incrementally increased through several layers, by only a small amount. In other words, although it is known that supplying too much energy to an object being manufactured by AM can cause swelling, the present inventor has surprisingly found that by supplying energy at a relatively great intensity to only a single layer in response to the detection of a defect in a first layer, swelling can be reduced relative to approaches that increase intensity by a relatively small amount but over a number of layers.

The method can be used to correct defects in-build without introducing significant delays into the manufacturing process. In particular, the method works by increasing (relative to the energy supplied to the first layer) the intensity of the energy supplied to a layer deposited subsequent to the first layer, without requiring that the first layer be re-melted. Thus, the second layer of material can be deposited while any processing required to detect the at least one defect in the first layer is performed. In other words, in contrast to a method in which the first layer is re-melted in order to correct defects in that layer, in the method of the seventh aspect there is no need to wait for the results of defect detection before the manufacturing process can continue. Instead, the results of defect detection will usually be available by the time the second layer of material has been deposited, which is the first time at which the method of the seventh aspect requires the intensity to be increased above normal levels used in manufacturing the part if a defect is detected.

[Ranges for Intensity on Second Layer]

The second, predetermined, intensity may be at least 1.3 times the first intensity. The second, predetermined, intensity may be at least two times the first intensity. The second intensity may be at least three times the first intensity. The second intensity may be at least four times the first intensity. The second energy may be at least five times the first intensity. The second, predetermined, intensity may be 1.3 to five times the first intensity.

As discussed above, the present inventor has discovered that by supplying energy to the second layer of material at a much higher intensity than the intensity of the energy supplied to the first layer, defects in the first layer can either be fixed in the first layer or at least prevented from propagating through into the second layer. As also discussed above, however, supplying too much energy to the material being used to manufacture an object by additive manufacturing can lead to swelling. The present inventor has discovered that by selecting a second intensity that is within the range of 1.3 to five times the first intensity, defect correction can be achieved without unacceptable levels of swelling.

[Third Layer]

The step of supplying energy to the second layer of material at the second intensity may comprise manufacturing a second layer of the object. The method may further comprise, after this step of manufacturing a second layer of the object by supplying energy to the second layer of material at the second intensity, depositing a third layer of material on the second layer and manufacturing a third layer of the object by supplying energy to the third layer of material at substantially the first intensity. The third layer of material may be directly deposited on the second layer. In other words the third layer of material may be deposited on the second layer without intermediate layers.

In this way, the method of the seventh aspect can correct defects in a first layer of an object by supplying increased energy to only one subsequent layer of material. After the second layer is manufactured at the second intensity, the subsequent (third) layer is manufactured at substantially the first intensity. In other words, after correcting defects in the first layer by increasing the intensity to manufacture the second layer, the method returns to manufacturing layers at the first intensity. This reduces the risk of swelling that would occur if the intensity were increased over too many subsequent layers. In particular, it allows for the method to be repeated for each layer in which a defect is detected, while minimising the risk that the cumulative increased intensity applied to the layers of the object is so high as to cause swelling in the object.

[Order and Repetition of Steps]

The method of the seventh aspect may be performed on a defect detected using the method of the fourth aspect. In other words, the method of the seventh aspect may comprise detecting at least one defect in the first layer of the object by carrying out the method of the fourth aspect.

By using the defect detection method of the fourth aspect, the defect correction method of the seventh aspect can benefit from the effects identified above as being associated with the defect detection method of the fourth aspect.

The method may be repeated for subsequent layers of the object manufactured by additive manufacturing. The method may comprise manufacturing an nth layer of an object by supplying energy to an nth layer of material at a first intensity; depositing an (n + l/th layer of material on the nth layer of the object; and, responsive to the detection of at least one defect in the nth layer of the object, supplying energy to the (n + l/th layer of material at a second, predetermined, intensity, wherein the second intensity is greater than the first intensity and is an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured by supplying energy to at least one further layer of material at the second, predetermined, intensity. The method may comprise, for any - nth - layer in which a defect is detected, carrying out the steps of depositing an (n + l/th layer of material on the nth layer of the object; and supplying energy to the (n + I/th layer of material at a second, predetermined, intensity, wherein the second intensity is greater than the first intensity and is an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured by supplying energy to at least one further layer of material at the second, predetermined, intensity. For example, the method may comprise depositing a fourth layer of material on the third layer of the object; and, responsive to the detection of at least one defect in a third layer of the object, supplying energy to the fourth layer of material at the second intensity.

By carrying out the steps of depositing an (n + l)th layer of material on the nth layer of the object; and supplying energy to the (n + l/th layer of material at a second intensity, wherein the second intensity is at greater than the first intensity and is an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured by supplying energy to at least one further layer of material at the second, predetermined, intensity for any - nth - layer in which a defect is detected, the method can minimise defects in an object manufactured by AM.

[Terms Used]

The material may be powdered material. The method of in-build defect correction for an object manufactured by additive manufacturing may be for an object manufactured by powder bed additive manufacturing. The method may be for an object manufactured by EBM. The method may be for an object manufactured by a laser blown metal powder technique, also known as laser metal deposition.

As used herein, the term intensity means the power transferred per unit area of the layer. The second intensity can be achieved, for example, by increasing the energy of a beam used to supply energy to the second layer (relative to the energy of the beam for the first layer). It can, additionally or alternatively, be achieved by focussing the beam used to supply energy to the second layer (relative to the energy of the beam for the first layer), or by slowing the passage of the beam across the second layer (relative to the first layer).

The second, predetermined, intensity can be determined experimentally and may be taken from a look-up table or calculated in real-time.

As used herein, the term swelling means a deformation to an object or part of an object caused by the energy supplied to successive layers being greater than the energy that can be lost by the layers through cooling over a given time. This deformation results in the object having a height and/or width that is greater than the expected height or width of the object if the layers were supplied with energy at the first energy density. The at least one further layer of material may be a layer directly on top of the layer at to which energy is supplied at the second, predetermined, intensity responsive to the detection of at least one defect in the first layer of the object.

The second layer of material may be a layer of material deposited directly on the first layer of the object (in which the at least one defect was detected). The second layer of material may be a layer of material deposited on at least one intermediate layer of the object (between the first layer in which the at least one defect was detected and the second layer). The method may comprise, subsequent to manufacturing a first layer of the object by supplying energy to the first layer of material at a first intensity, and before depositing a second layer of material and supplying energy to the second layer of material at a second intensity, depositing at least one intermediate layer of material on the first layer of the object.

When the second layer of material is a layer of material deposited directly on the first layer of the object, the method has a greater probability of correcting defects in the first layer than if there are one or more intermediate layers between the first layer of the object and the second layer of material.

However, the present inventor has discovered that when the second intensity is at great enough (for example at least two times the first intensity), the penetration depth of the energy through the layers is sufficient to correct defects in layers below the layer immediately below the second layer. Thus, when the second layer of material is a layer of material deposited on an intermediate layer of the object, defects in the first layer can nevertheless be corrected. This approach allows for a reduction in the number of layers of material to which energy is supplied at the second intensity relative to an approach in which energy is always supplied at the second intensity on a layer immediately above a layer in which a defect is detected. Further, this approach can accommodate processing times for defect detection of the order of the time taken to manufacture two or more layers of an object, without introducing additional delays. In particular, if the processing time to perform defect detection in the first layer of the object is greater than the time taken to manufacture a further layer of the object on top of the first layer, the approach can allow for the results of the defect detection to be calculated before supplying energy to the second layer at the second intensity, even if that means that intermediate layers have been formed between the first layer and the second layer in the meantime.

[Eighth Aspect]

According to an eighth aspect of the present disclosure, there is provided a computer-readable storage medium storing instructions that are arranged, when executed by a computer, to cause the computer to carry out the method of the seventh aspect.

[Ninth Aspect]

According to a ninth aspect of the present disclosure, there is provided an additive manufacturing system arranged to correct defects in an object manufactured by the system, the system comprising: an additive manufacturing apparatus arranged to manufacture an object by manufacturing a plurality of layers of the object, wherein manufacturing each of the plurality of layers comprises depositing a layer of material and supplying energy to the layer of material; and an instruction module arranged to instruct the additive manufacturing apparatus to: deposit a first layer of material; manufacture a first layer of an object by supplying energy to the first layer of material at a first intensity; deposit a second layer of material; and, responsive to receipt by the instruction module of an indication that at least one defect has been detected in the first layer of the object, supply energy to the second layer of material at a second, predetermined, intensity, wherein the second intensity is greater than the first intensity and is an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured by supplying energy to at least one further layer of material at the second, predetermined, intensity.

[Instruction Module]

The instruction module may comprise a processor. The processor may be arranged to execute arithmetic processing in accordance with a program or other types of instructions. The instruction module may additionally comprise a storage device. The storage device may comprise, for example, volatile memory such as RAM and/or non-volatile memory such as a hard disk, and may be arranged to store program information and/or program data. The processor may be communicatively connected to the storage device and to the additive manufacturing apparatus, and arranged to carry out instructions received from the storage device to instruct the additive manufacturing apparatus to: deposit a first layer of material; manufacture a first layer of an object by supplying energy to the first layer of material at a first intensity; deposit a second layer of material; and, responsive to a detection of at least one defect in the first layer of the object, supply energy to the second layer of material at a second, predetermined, intensity, wherein the second intensity is greater than the first intensity and is an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured by supplying energy to at least one further layer of material at the second, predetermined, intensity.

[Defect Detection]

The system may additionally be arranged to perform defect detection and to send to the instruction module indication that at least one defect has been detected in the first layer of the object. The system may comprise a defect detection module arranged to perform defect detection and to send to the instruction module indication that at least one defect has been detected in the first layer of the object. The defect detection module may be the defect detection module of the sixth aspect.

Optional features of each aspect are also optional features of each other aspect, with changes of terminology being inferred by the skilled addressee where necessary for these to make sense. For example, when applying the feature of the third aspect that step (a) may be performed before steps (b) to (g) to the sixth aspect, the skilled addressee would interpret this feature as the processor is arranged to carry out the instructions to perform the steps defined in the fifth aspect before steps (i) to (vii). As another example, when applying the feature of the fourth aspect that the infrared or near infrared image may be an image of a layer of an object being manufactured by hot bed AM to the seventh aspect, the skilled addressee would interpret this feature as the object manufactured by additive manufacturing may be manufactured by hot bed AM.

BRIEF DESCRIPTION OF THE DRAWINGS

Specific embodiments will be described below by way of example only and with reference to the accompanying drawings, in which:

Figure 1 shows an area of a NIR image of a layer of an object being manufactured by EBM;

Figure 2 shows a flow diagram of a method of boundary detection for an object manufactured by AM;

Figure 3a shows an image of a grid of electron beam dots projected onto a base plate for calibration of images of layers of an object built on the base plate;

Figure 3b shows a close-up of part of the image of the grid;

Figure 3c shows the close-up of part of the image after correction to remove distortion;

Figure 4a shows a raw NIR image of a layer of a group of objects including the object shown in Figure 1;

Figure 4b shows the image after calibration;

Figure 5 shows a CAD image of a slice of a CAD model representing the object;

Figure 6a shows the CAD image overlaid on the NIR image;

Figure 6b shows the resulting image with pixels representing the boundary of the objects of the group and pixels that are outside this boundary removed;

Figure 7 shows a flow diagram of a method of defect detection for an object manufactured by AM;

Figures 8a-8e shows five areas of an NIR image of an object, and the brightness distribution of the pixels in each of those areas;

Figure 9a shows a schematic representation of an object as it is being built by EBM on a base plate;

Figure 9b shows the temperature profile across the object and the surrounding powder at successive time intervals after a layer of the object has been built;

Figure 10 shows a region to be analysed of an NIR image, and a pixel to be analysed within the region;

Figure 11 shows the image of Figure 6b with identified defects highlighted;

Figure 12 shows a CAD 3D model of the object and the identified defects in the object;

Figure 13a shows a close-up of defects in the CAD 3D model of the object of Figure 1;

Figure 13b shows a close-up of part of the model with an artefact produced by dirt on the glass in front of a camera taking the images from which the defects are identified;

Figure 14 shows a flow diagram of a method of in-build defect correction for an object manufactured by AM; and

Figure 15 shows a schematic diagram of an AM system arranged to perform boundary detection, defect detection and defect correction in an object manufactured by the system.

SPECIFIC DESCRIPTION OF CERTAIN EXAMPLE EMBODIMENTS

Figure 2 shows a flow diagram of a method 200 of boundary detection for an object manufactured by AM.

As discussed above, being able to identify pixels representing a boundary of an object in such an image can prevent these pixels from being mistaken for pixels representing defects in the object during defect detection performed on the image. It was mentioned above that the brightness of pixels representing a boundary of an object could cause such pixels to be confused with pixels representing defects in the object (since these pixels are also bright). The cause of the brightness of pixels representing at least part of the boundary of an object in an IR or NIR image of a layer of the object being manufactured by AM - hereinafter called boundary pixels will now be described with reference to Figures 9a and 9b.

Figure 9a shows a model used for a simulation of the cooling of an object during EBM. Figure 9b shows results of the simulation in the form of a number of graphs of the temperature across the central line of the object shown in Figure 9a. The graphs show the temperature across the central line at successive time intervals after the object and surrounding powder have been heated to

1800°C and 625°C respectively and then allowed to cool for a few seconds (simulating the heating and cooling of an object and powder during EBM).

As can be seen from Figure 9b, in this simulation, 2.4s after heating of the object and powder the very edges of the object are at a higher temperature than the part of the object just inward of the edges. This effect is exaggerated as the time after heating the part increases.

The body of the object has higher thermal conductivity than the surrounding powder. At the end of heating, the high temperature gradient between the object and the powder causes the powder adjacent the object to absorb heat energy quickly. Afterwards, however, since the powder has a lower thermal conductivity than the body of the object, it will cool down more slowly than the object. In an IR or NIR image taken after a layer has been manufactured (i.e. after the object has been heated), the boundary between the object and the powder will therefore be brighter than the surrounding region.

With reference once more to Figure 2, the method 200 of boundary detection for an object manufactured by AM will now be described in more detail. In the particular example shown in Figure 2, the AM method is EBM, and the object being manufactured is a mug being manufactured as a group of mugs (an image of a 3D model of the mug is shown in Figure 12). In the example of Figure 2, the method is carried out on an additive manufacturing system 1500 which will be described in more detail below with reference to Figure 15.

In other examples, the AM method can be any method in which a material is heated to manufacture a layer, and the object being manufactured may be a part or component of conceivably any sort that can be built using the AM method.

Returning to the example shown in Figure 2, the steps of the method 200 will now be described.

The method 200 of this example is carried out on a NIR image. In other examples, the image can be an infrared image (or any other image in which the brightness of the pixels represents thermal radiative energy emitted by an object). In the present example, the NIR image is of a layer of four mugs being manufactured by EBM. The skilled reader will understand the EBM manufacturing process, which will therefore not be described further here. The NIR image is captured by a NIR camera within the EBM apparatus. The NIR image is taken after a layer of the mugs has been built and before further powdered material for building the next layer of the mugs has been raked across the build surface.

In the present example, the NIR camera used to take the image is off-centre with respect to the mugs being built (because it is not on the same axis as the electron beam gun used to build the mugs). Because of the positioning of the camera, and because of distortion caused by the camera lens, images taken with the camera will be slightly skewed.

In the present example, the NIR image is corrected using a calibration image before it is processed to identify the pixels representing the boundary of the objects. This calibration image in this example is an image of the base plate of the EBM apparatus, captured by the NIR camera before building of the mugs is begun. For the taking of the calibration image, the electron beam gun is caused to project a grid of dots onto the base plate, which in this embodiment are 3mm apart. The calibration image 300 used with this example is shown in Figure 3a.

The first step of the method in the present example is therefore receiving 20 the calibration image 300. In examples in which it is not desired to correct the NIR or IR image based on a calibration image, this step is omitted.

As the spacing of the dots and the angle of the lines of dots with respect to the edges of the baseplate is known, the calibration image 300 is resized and rotated to have a set number of pixels between the dots in the image (e.g. 40 pixels, in this example) and a set angle of the lines of dots with respect to the edges of the image. Figure 3b shows part 310 of the calibration image 300 before such resizing and rotation. Figure 3c shows this part 310 after the resizing and rotation.

Next, a NIR image 400 is received 21. The NIR image is shown in Figure 4a. In this example, the NIR image 400 is corrected 22 using the calibration image 300 by applying the same rotation and resizing to the NIR image 400 as was applied to the calibration image 300 to achieve the set spacing and angle of the dots. The corrected NIR image 410 is shown in Figure 4b. In examples in which it is not desired to correct the NIR or IR image based on a calibration image, this step is omitted.

With reference once more to Figure 2, the step of slicing 23 a CAD model representing the mugs into a plurality of slices is carried out. In this example, the CAD model is in .stl file format, and is the model from which the mugs are built. In other examples, the CAD model may be in other CAD file formats. In this example, each slice of the CAD model corresponds to a layer of the mugs as built. An appropriate slicing method will be known to the skilled reader and will therefore not be described further here.

Next, the slice of the model that corresponds to the layer in the NIR image 400 is selected 24 and is output 25 as a CAD image 500. This image 500 of the slice is shown in Figure 5. In this example, the slice corresponds to the layer shown in the NIR image 400 in that the layer was built from this slice.

As can be seen from Figure 5, the image 500 of the slice has a border 501. This border represents the boundary of the mugs.

To identify the boundary pixels of the NIR image 400, the CAD image 500 is correlated 26 with the NIR image 400. In this example, the two images are correlated 26 by overlaying the CAD image 500 on the NIR image 400. The resulting combined image 600 is shown in Figure 6a. Any pixel of the NIR image 400 that is under the border of the CAD image 500 is identified 27 as being a boundary pixel.

In this example method 200, a further step of removing 28 from the combined image 600 any pixels outside the border of the CAD image 500 is performed. The resulting image 610 is shown in Figure 6b. Since in any defect detection method performed on an image, the region of the image not representing the object need not be analysed, removing 28 the area of the image that is outside the border of the CAD image 500 gives an image 610 in which only areas of interest for defect detection are shown.

With brief reference now to Figure 7, a method 700 of defect detection for an object manufactured by AM will be described.

As discussed above, pixels representing defects (hereinafter called defect pixels) are difficult to detect in an IR and NIR image of a layer of an object manufactured by AM. Defect pixels may represent holes in the object, and therefore - as discovered by the present inventor -appear bright either because the holes exhibit higher emissivity than the rest of the object due to their larger surface area per unit area, or because as holes they emit black body radiation. However, although the defect pixels appear brighter than pixels representing a relatively cool area of the object, in an IR or NIR image taken after a layer of the object has been manufactured, the object may not be cool and thus pixels other than the defect pixels may also be bright. It is thus difficult to distinguish defect pixels from pixels representing warmer or hotter areas of the object.

With reference now to Figures 8a to 8e, this difficulty in detecting defects will be elaborated on.

Figures 8a to 8e show NIR images of areas of a layer of an object, and accompanying graphs showing pixel brightness (with the brightness of the pixels being on the x-axis, and the number of pixels being on the y-axis). The standard deviation (SD) of the pixels is also shown. In particular, Figure 8e shows a NIR image of area of a layer of an object in which there is a transition between hotter and cooler. In this image, there is little contrast in brightness between one area and another; the heat transition appears in the image as a gradient. This area of the layer may contain defects, but they are not visible in the image as the contrast of a defect pixel with the background is not well-defined. Such defects would be hard to detect based on a threshold method of detection (in which pixels above a certain brightness threshold are identified as representing defects), due to the broad distribution of background brightness. In areas of IR or NIR images such as that shown in Figure 8e, it is therefore particularly difficult to detect defect pixels. The present method of defect detection 700 does not rely on comparing pixel brightness to a threshold brightness and thus allows for more accurate detection of defect pixels than such a method.

Returning now to Figure 7, the method of defect detection 700 of the present example will now be described in more detail.

In the example shown in Figure 7, the first step of the method of defect detection 700 is to create the NIR image 610 shown in Figure 6b, by performing the steps of the method 200 of boundary detection as described above. In other examples, the identification of boundary pixels may be performed after the other steps of the method of defect detection, or may be performed concurrently with the other steps.

Returning now to the discussion of the present example method of defect detection 700, after identifying any boundary pixels in the manner described above, a region 101 of the NIR image 610 is selected 70. This is the region 101 to be analysed for defects first. The region 101 is shown in Figure 10. As can be seen from Figure 10 the region 101 consists of a number of pixels. In this example method 700, the region 101 is a square of 64 pixels; that is, the region is 8 pixels by 8 pixels. In other examples, the region can be made up of more or fewer pixels, for example having between two and five pixels per side.

Next, a particular pixel 111 within the region 101 is selected 71. For the purposes of this description, this pixel 111 will be called the potential defect pixel 111. The brightness of the potential defect pixel 111 is determined 71. In the present example, the image 610 is a greyscale image, and determining the brightness of the potential defect pixel 111 is performed by identifying its greyscale value.

The brightness of each of the other pixels in the region 101 is then determined 72. Again, in this example, the determination 72 is performed by identifying the greyscale values of each of the other pixels in the region 101.

Next, the average brightness (i.e., in this example, the average greyscale value) of each of the pixels of the region 101 apart from the potential defect pixel 111 is calculated 73.

The difference between the average brightness and the brightness of the potential defect pixel 111 is then calculated 74. In this example, this means determining the difference in the average greyscale value of each of the pixels of the region 101 apart from the potential defect pixel 111 and the greyscale value of the potential defect pixel 111. In this way, the brightness of the potential defect pixel 111 can be compared to the brightness of the pixels around it; in contrast to methods in which the brightness of the potential defect pixel 111 is compared to a threshold value. Thus, the method 700 allows for a determination of the relative brightness of the potential defect pixel 111 and its surrounding pixels.

Next, a second potential defect pixel 112 is analysed. In this example, this means repeating 75 the above steps of selecting 71 a pixel within the region 101 (in this case the second potential defect pixel 112), determining 72 the greyscale value of each of the other pixels in the region 101, calculating 73 the average greyscale value of these pixels, and calculating 74 the difference between the average greyscale value and the greyscale value of the potential defect pixel (the second potential defect pixel 112).

The steps of the above paragraph are repeated taking each pixel of the region 101 as a potential defect pixel. In this example, therefore, the steps are repeated a further 62 times (as there are 64 pixels in the region 101).

To identify 76 a defect pixel 1101, the difference in brightness (i.e. difference in greyscale value) of each potential defect pixel 111, 112 is compared to a threshold value. If the difference in brightness (in this example, difference in greyscale value) is greater than that threshold value for one or more of the potential defect pixels 111, 112, that or those pixels is identified 76 as a defect pixel 1101. The threshold value can depend on factors such as the temperature to which the object is heated, the shape of the object being built, the wavelength of the radiation detected by the camera, the length of the exposure, the camera aperture and the sensitivity of the camera sensor. Thus, it is determined experimentally for a particular setup. It can, however, be determined without undue experimentation.

In some examples, the method can end after this step, when only a very small area of an image is to be analysed for defect pixels 1101. In this example defect detection method 700, however, the above steps of selecting a pixel and comparing its brightness (greyscale value) to an average brightness (greyscale value) of the pixels surrounding it are repeated for each pixel in the NIR image 610. In this way, defects can be detected across the whole of the layer. In other examples, the above steps can be carried out for just a subset of pixels in an IR or NIR image. This approach can be used, for example, when only defects in a certain area of the object are of interest.

Returning to the present example defect detection method 700, the method 700 continues with a step of identifying 78 noise in the image 610. Based on the inventor's analysis of defects, it has been determined that defects are usually represented by a group of at least a certain number of defect pixels 1101. Thus, any group of bright pixels below this size can be determined not to be a defect and in fact to be noise. In this particular example, with its particular NIR image 610 resolution, groups of nine or fewer bright pixels generally represent noise and not a defect. Thus, to identify 78 noise in the image 610, any group of seven or fewer adjacent pixels having a brightness above a threshold brightness are identified 78 as noise, and not defect pixels 1101. The number seven has been selected so as not to miss smaller defects.

An image 1100 showing pixels 1101 identified 76 as representing at least part of a defect (without pixels identified 78 as noise) is shown in Figure 11. The defect pixels 1101 are shown in white.

With reference once more to Figure 7, in this example defect detection method, the above steps from identifying and removing boundary pixels to identifying 76 defect pixels 1101 and noise 78 - are repeated 79 for NIR images representing each layer of the mugs. In this way, any defects in the mugs can be detected for each layer of the mugs.

Next, the 3D coordinates of each defect pixel 1101 are determined 80 either as Cartesian or polar coordinates.

The 3D coordinates of the defect pixels 1101 are then converted 81 into a 3D defect model, showing the location of the defect pixels 1101. This is output as a CAD file. In this example, the file format is .stl.

Finally, the 3D defect model is overlaid on the CAD model representing the mugs. The resulting model can be visually inspected through a CAD file viewer. An image 1200 of a mug from such a viewer is shown Figure 12. A close-up 1310 of part of the model is shown in Figure 13a. As can be seen from this figure, defects 110 in the form of tunnels are present in the mug. Visual inspection of the defect model overlaid on the model representing the mug allows for artefacts which have been identified by the above process as defects to be reviewed to check that they do in fact have characteristics typical of defects. For example, Figure 13b shows a close up 1320 of a straight line 1321 through the model. This kind of straight line 1321 can be identified as a defect by the above process, but visual inspection allows for it to be determined in fact to be an artefact caused by dirt on the glass shielding the camera taking the NIR images (as defects 110 are typically not straight lines).

With reference now to Figure 14, a method 1400 of in-build defect correction for an object manufactured by additive manufacturing will now be described. As with the methods of boundary 20 and defect detection 70 described above with reference to Figures 2 and 5 respectively, the object manufactured in the method 1400 of in-build defect correction is manufactured by EBM. Once again, the object being manufactured is a mug (an image of a 3D model of the mug is shown in Figure 12). The skilled reader will understand, of course, that the method 1400 of in-build defect correction described herein is applicable to the manufacture of any object that can be manufactured by a powder-bed manufacturing technique. In the example shown in Figure 14, the method 1400 is carried out on the additive manufacturing system 1500 that is described in more detail below with reference to Figure 15.

In this example, the first step of the method 1400 is to deposit 140 a first layer of powdered material. This is done by raking the material across a build area, in a way that would be known to the skilled addressee. In this example, the material is titanium, although the method can be used with any other material that can be used to manufacture an object by powder-based additive manufacturing.

The second step is to supply energy 141 to the first layer of powdered material. In this example, this is done using an electron beam that is selectively scanned across the surface of the first layer of powdered material. In other examples, this step could be carried out using a laser beam that is similarly scanned across the surface of the first layer of powdered material. The energy is supplied 141 at an intensity that is appropriate for manufacturing a layer of an object using the powdered material. The specific intensity required is dependent on the apparatus used to carry out the method, the object that is being manufactured and the material that it is being manufactured from. It can be determined experimentally for these particular conditions, without undue experimentation.

Next, defect detection 142 is performed on the first layer. This is done, in this example, using the method described above with reference to Figure 2. In other examples, it can be done using conceivably any method that is able to detect a defect in a layer.

In this example, the processing required for this defect detection is performed concurrently with the next step of the method, which is to deposit 143 a second layer of powdered material. Depositing 143 the second layer of powdered material is carried out in the same way as the above-described deposition 140 of the first layer of powdered material. In this example, the results of the defect detection 142 are available as soon as the second layer of powdered material has been deposited 143.

If a defect has been detected, the method proceeds to the step of supplying 144 energy to the second layer of powdered material. This is done as described in relation to the step of supplying 141 energy to the first layer of powdered material, except that the energy supplied 144 to the second layer of powdered material is supplied with - in this example - two times the intensity of the energy supplied 141 to the first layer. This can be achieved by reducing the hatching line offset, the focus offset or speed function of the beam of energy that is scanned across the surface of the second layer of powder. In this example, the speed function (the relation of speed and beam current) is reduced, so that the electron beam passes more slowly across the second layer of powdered material than it did across the first layer of powdered material, thereby imparting more energy to the second layer of powdered material.

The step of supplying 144 energy to the second layer of powdered material at two times the intensity of the energy supplied to the first layer of powdered material results in the correction of defects in the first layer of the object - or at least the prevention of their propagation into subsequent layers. It also results in the manufacture of a second layer of the object.

In this example, once the second layer of the object has been built, the above steps are repeated 145. In other words, the method 1400 returns to the step of depositing 140 a layer of powdered material. This deposition 140 of a third layer is carried out in substantially the same manner as the deposition 140 of the first layer of powdered material. The remaining steps of the method 1400 are also the same, with the above description of steps in relation to the first layer of powdered material and of the object being applied in relation to the third layer of powdered material and of the object, and with the above description of steps in relation to the second layer of powdered material and of the object being applied in relation to the fourth layer of powdered material and of the object.

With reference now to Figure 15, an additive manufacturing system arranged to perform boundary detection, defect detection and in-build defect correction in an object manufactured in the system in other words, to perform the above-described methods - will now be described. In this example, the system is an EBM machine 1500. The EBM machine 1500 has an additive manufacturing apparatus in the form of an EBM apparatus 151, a camera in the form of an NIR camera 152, and an integrated computer 156. The system has boundary detection module 154, a defect detection module 153 and an instruction module 155. In this example, the boundary detection module 154, defect detection module 153 and instruction module 155 form part of the computer 156. In this example, the boundary detection module 154, defect detection module 153 and instruction module 155 are embodied in a storage device in the form of RAM and a hard disk and a processor. They are shown in Figure 15 as separate modules for ease of understanding. In other examples, the boundary detection module 154, defect detection module 153 and instruction module 155 can each be a dedicated hardware module or can take the form of software. Returning to the description of the EBM machine 1500 shown in Figure 15, the computer 156 also has an output device in the form of a visual display 158.

The boundary detection module 154 and defect detection module 153 are communicatively connected via one or more communications busses 157. The boundary detection module 154 and NIR camera 152 are also communicatively connected via the one or more communications busses 157. The defect detection module 153 is communicatively connected to the instruction module 155 via the one or more communications busses 157. Finally, the defect detection module 153 and visual display 158 are also communicatively connected via the one or more communications busses 157.

In other examples, any or all of the boundary detection module, defect detection module, instruction module and visual display can be in a separate location from the AM apparatus and camera. For example, the AM apparatus and camera can be contained in an AM machine, while the boundary detection module, defect detection module, instruction module and visual display can be part of a computer that is not integrated with the AM machine. In such examples, the communications busses 157 are accordingly replaced with any of: a wired connection, such as Ethernet, USB or Firewire; a wireless connection, such as Bluetooth, Wi-Fi, cellular network or infrared; or a manual transfer using a computer-storage medium, such as CD, DVD, Blu-Ray, Memory Card or USB flash drive; according to the requirements of the particular implementation.

The EBM apparatus 151 is of a standard sort that will not be described further here. In this example, the NIR camera 152 is located above the build area of the EBM apparatus 151 and off-axis with respect to the electron beam of the EBM apparatus 151. This means that calibration is required of images taken by the NIR camera 152.

The boundary detection module 154 is arranged to perform the method 200 of boundary detection for an object manufactured by AM described above with reference to Figure 2. In this example, this means that the storage device contains instructions that, when executed by the processor, carry out the above-described boundary detection method 200.

The defect detection module 153 is arranged to perform the method 700 of defect detection for an object manufactured by AM that is described above with reference to Figure 7. In this example, this means that the storage device contains instructions that, when executed by the processor, carry out the above-described defect detection method 700.

The instruction module 155 is arranged to instruct the EBM apparatus 151 to deposit 140 a first layer of powdered material, manufacture a first layer of an object by supplying 141 energy to the first layer of powdered material at a first intensity and deposit 143 a second layer of powdered material in the manner described above in relation to Figure 14. The instruction module 155 is further arranged to receive an indication from the defect detection module 153 that a defect has been detected, and responsive to this indication, to instruct the EBM apparatus 151 to supply 144 energy to the second layer of powdered material at the second intensity, also as described above in relation to Figure 14. In this example, the storage device contains instructions that, when executed by the processor, carry out these steps.

The visual display 155 is of a sort that can display the 3D defect model as shown in Figures 12,13a and 13b.

Thus, the above description discloses methods of boundary detection, defect detection and in-build defect correction for an object manufactured by additive manufacturing. The above description also discloses related systems, and ways in which to perform or carry out these methods and systems.

Claims

1. A method of boundary detection for an object manufactured by additive manufacturing, the method comprising:

receiving a first, infrared or near-infrared, image of a first layer of an object manufactured by additive manufacturing, the first image having a plurality of pixels;

slicing a computer-aided design model representing the object into a plurality of slices, at least one slice corresponding to a layer of the object as manufactured by additive manufacturing;

selecting a first slice of the computer-aided design model, the first slice corresponding to the first layer of the object;

outputting at least the first slice of the computer-aided design model as a second, computer-aided design, image having a border representing a boundary of the object;

correlating the second image with the first image; and identifying as a pixel that represents at least part of a boundary of the object any pixel of the first image that corresponds to the border of the second image.

2. The method of claim 1, wherein the infrared or near infrared image is an image of a layer of an object manufactured on a heated bed.

3. The method of claim 1 or claim 2, wherein the infrared or near infrared image is an image of a layer of an object manufactured by powder bed additive manufacturing.

4. The method of claim 2 or claim 3, wherein the infrared or near infrared image is an image of a layer of an object manufactured by electron beam melting.

5. The method of any preceding claim, further comprising:

removing from the first image any pixels that are identified as representing at least part of the boundary of the object and any pixels that are outside the border of the second image.

6. A method of defect detection for an object manufactured by additive manufacturing, the method comprising:

(a) performing the method of any preceding claim;

(b) selecting a region of the first image, the region consisting of a plurality of pixels;

(c) determining the brightness of a first pixel in the region;

(d) determining the brightness of each of the other pixels in the region;

(e) calculating the average brightness of each of the other pixels in the region;

(f) calculate the difference between the brightness of the first pixel and the average brightness;

(g) repeating steps (c) to (f) for each of the pixels in the region; and (h) identifying any pixel having a difference in brightness from the average brightness above a threshold value as representing at least part of a defect.

7. The method of claim 6 wherein step (a) is performed before steps (b) to (g).

8. The method of claim 6 or claim 7 wherein steps (b) to (g) are repeated for each pixel in the first image.

9. The method of any of claims 6 to 8, wherein steps (a) to (g) are repeated for a plurality of infrared or near-infrared images, each infrared or near-infrared image being an image of a layer of the object being manufactured by additive manufacturing.

10. The method of any of claims 6 to 9, wherein the method comprises identifying as noise any group of a predetermined number or fewer adjacent pixels having a difference in brightness from the average brightness that is at or above a threshold value.

11. The method of any of claims 6 to 10, wherein the region consists of nine pixels.

12. The method of any of claims 6 to 11, wherein the computer-aided design model is a first computer-aided design model, wherein the method comprises determining coordinates of each pixel representing at least part of a defect, wherein the method further comprises converting the coordinates of each pixel representing at least part of a defect into a second computer-aided design model, and wherein the method further comprises combining the first and second computer-aided design models to produce a third computer-aided design model of the object and the identified defects in the object.

13. A method of in-build defect correction for an object manufactured by additive manufacturing, the method comprising:

depositing a first layer of material;

manufacturing a first layer of the object by supplying energy to the first layer of material at a first intensity;

depositing a second layer of material; and responsive to the detection of at least one defect in the first layer of the object, supplying energy to the second layer of material at a second, predetermined, intensity, wherein the second intensity is greater than the first intensity and is an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured by supplying energy to at least one further layer of material at the second, predetermined, intensity.

14. The method of claim 13, wherein the second, predetermined, intensity is at least 1.3 times the first intensity.

15. The method of claim 13 or claim 14, wherein the second, predetermined, intensity is at least two times the first intensity.

16. The method of any of claims 13 to 15, wherein the second layer of material is a layer of material deposited directly on the first layer of the object.

17. The method of any of claims 13 to 16, wherein the second layer of material is a layer of material deposited on at least one, intermediate, layer of the object between the first layer and the second layer.

18. The method of any of claims 13 to 15 and 17 wherein the method comprises, subsequent to manufacturing a first layer of the object by supplying energy to the first layer of material at a first intensity, and before depositing a second layer of material and supplying energy to the second layer of material at a second intensity, depositing at least one intermediate layer of material on the first layer of the object.

19. The method of any of claims 13 to 18, wherein the step of supplying energy to the second layer of material at the second intensity comprises manufacturing a second layer of the object and wherein the method further comprises, after this step of manufacturing a second layer of the object, depositing a third layer of material on the second layer and manufacturing a third layer of the object by supplying energy to the third layer of material at substantially the first intensity.

20. The method of claim 19, wherein the third layer of material is directly deposited on the second layer.

21. The method of any of claims 13 to 20, wherein the method is repeated for at least one subsequent layer of the object manufactured by additive manufacturing.

22. The method of any of claims 13 to 21, wherein the method of the seventh aspect comprises detecting at least one defect in the first layer of the object by carrying out the method of any of claims 6 to 12.

23. A method according to any of claims 6 to 22 wherein the defect is a hole or a crack.

24. A computer-readable storage medium storing instructions that are arranged, when executed by a computer, to cause the computer to carry out the method of any of claims 1 to 23.

25. An additive manufacturing system arranged to perform boundary detection in an object manufactured in the system, the system comprising:

an apparatus arranged to manufacture an object by additive manufacturing by manufacturing a plurality of layers of the object;

an infrared or near-infrared camera arranged to take a first image of a first layer of the plurality of layers of the object, the first image having a plurality of pixels;

a storage device arranged to store instructions;

a processor communicatively connected to the camera and to the storage device, and arranged to receive the first image from the camera and instructions from the storage device, the processor further arranged to carry out instructions received from the storage device to:

slice a computer-aided design model representing the object into a plurality of slices, at least one slice corresponding to a layer of the object as manufactured by additive manufacturing;

select a first slice of the computer-aided design model, the first slice corresponding to the first layer of the object;

output at least the first slice of the computer-aided design model as a second, computer-aided design, image having a border representing a boundary of the object;

correlate the second image with the first image; and identify as a pixel that represents at least part of a boundary of the object any pixel of the first image that corresponds to the border of the second image.

26. The additive manufacturing system of claim 25, wherein the boundary detection module is additionally arranged to remove from the first image any pixels that are identified as representing at least part of the boundary of the object and any pixels that are identified as being outside the boundary of the object.

27. The additive manufacturing system of claim 25 or 26, wherein the system is additionally arranged to perform defect detection in the object manufactured in the system and the processor is additionally arranged to carry out instructions received from the storage device to:

(i) select a region of the first image, the region consisting of a plurality of pixels;

(ii) determine the brightness of a first pixel in the region;

(iii) determine the brightness of each of the other pixels in the region;

(iv) calculate the average brightness of each of the other pixels in the region; (v) calculate the difference between the brightness of the first pixel and the average brightness;

(vi) repeat steps (ii) to (v) for each of the pixels in the region; and (vii) identify any pixel having a difference in brightness from the average brightness above a threshold value as representing at least part of a defect.

28. An additive manufacturing system arranged to correct defects in an object manufactured by the system, the system comprising:

an additive manufacturing apparatus arranged to manufacture an object by manufacturing a plurality of layers of the object, wherein manufacturing each of the

J I plurality of layers comprises depositing a layer of material and supplying energy to the layer of material; and an instruction module arranged to instruct the additive manufacturing apparatus to:

5 deposit a first layer of material;

manufacture a first layer of an object by supplying energy to the first layer of material at a first intensity;

deposit a second layer of material; and responsive to receipt by the instruction module of an indication that at least

10 one defect has been detected in the first layer of the object, supply energy to the second layer of material at a second, predetermined, intensity, wherein the second intensity is greater than the first intensity and is an intensity at which swelling would occur in the object if at least one further layer of the object were to be manufactured by supplying energy to at least one further layer of material at the second, predetermined, intensity.