FAILURE DETECTION FOR ADDITIVE MANUFACTURING SYSTEMS

Abstract

This invention teaches a quality assurance system for additive manufacturing. This invention teaches a multi-sensor, real-time quality system comprising sensors, integrated hardware, and data processing algorithms that are Lagrangian (i.e., in a moving reference frame) or Eulerian (i.e., in a fixed reference frame) with respect to reference frames of the associated input measurements. The quality system for additive manufacturing is capable of measuring actual process state variables associated with an additive manufacturing process, i.e., those process variables that define a possible process space within which the process is considered to nominally operate. The process state variables can also be correlated with part structure or microstructure and thus be helpful in identifying specific regions within the part that are likely to contain defects.

Description

BACKGROUND OF THE INVENTION

[0001] Additive manufacturing, the sequential buildup or production of a part through the combination of material input and applied energy, takes many forms and currently exists in many specialized applications and embodiments. Additive manufacturing can be achieved using countless different processes that involve the formation of a three-dimensional part of virtually any shape. Common to these different processes are the sintering, curing, or melting of liquid, powdered, or granular raw material, a layer-by-layer application of ultraviolet light, a high-energy laser, or an electron beam. Unfortunately, known processes for determining the quality of a part produced in this way are limited. Conventional quality assurance tests generally involve destruction of the part.While destructive testing is an accepted method for evaluating the quality of a part, as it allows for the detailed examination of various internal areas of the part, such tests cannot be applied to a production part for obvious reasons. Therefore, there is considerable interest in ways to non-destructively verify the integrity of a part produced through additive manufacturing.

[0002] US 2015 / 0 048 064 A1 describes a direct metal laser melting (DMLM) system for improving build parameters of a DMLM component, comprising a confocal optical system configured to measure at least one of a melt pool size and a melt pool temperature. The DMLM system further comprises a computing device configured to receive at least one of the melt pool size and the melt pool temperature from the confocal optical system. Furthermore, the DMLM system comprises a controller configured to control operation of a laser device based on at least one build parameter.DE 20 2010 010 771 U1 discloses a laser melting device in which a component is produced by successively solidifying individual layers. The melting region can be detected by a sensor device with regard to its dimensions, shape, and / or temperature, and sensor values for evaluating component quality can be derived therefrom. The laser melting device contains a storage device in which the sensor values detected for evaluating component quality can be stored together with coordinate values localizing the sensor values in the component, and a visualization device by which the stored sensor values can be displayed in a two- or multi-dimensional representation relative to their detection location in the component. WO 2007 / 147 221 A1 discloses a device for selective laser powder processing that has a feedback controller for improving the stability of the selective laser powder processing process.A signal reflecting a geometric size of the melt zone is used in the feedback controller to adjust the scanning parameters (e.g., laser power, laser spot size, scan speed, etc.) of the laser beam to maintain the geometric size of the melt zone at a constant level. The signal reflecting the geometric size of the melt zone is displayed for monitoring the selective laser powder processing process. DE 694 09 669 T2 discloses a laser sintering device comprising a detector for detecting the temperature of the powder at a movable detection point. The device also comprises a laser control device responsive to a detection signal indicative of the temperature from the detector for controlling the power of the laser beam.WO 2012 / 037694 A2 discloses systems for using optical interferometry in connection with material modification processes such as surgical laser or welding applications. An imaging optical source that generates imaging light. A feedback controller controls at least one processing parameter of the material modification process based on an interferometry output generated using the imaging light. US 2005 / 0251282 A1 discloses an apparatus for applying packaging material to workpieces. A machine vision system having at least one camera is operatively connected to a computer that controls a material application system so that the system can detect the position and orientation of nearby workpieces to which the material is to be applied. US 2012 / 0098164 A1 discloses a method of two-photon stereolithography using a photocurable material.A light beam is focused on a focal area of the material to induce two-photon absorption and thus polymerization of the material. The beam is scanned across the material according to a preselected pattern, focusing the beam on different, preselected areas to trigger polymerization of the material in the preselected areas. OVERVIEW OF THE INVENTION

[0003] The invention is described in the appended set of claims. The present invention generally relates to methods and systems for non-destructively characterizing structural integrity of a part manufactured by an additive manufacturing process. For example, some embodiments relate to quality control procedures observing the manufacture of metal parts using additive manufacturing techniques. More specifically, embodiments relate to monitoring thermal emissions during an additive manufacturing process to identify microstructural defects created during the additive manufacturing process.

[0004] The described embodiments relate to a major subcategory of additive manufacturing, which involves the use of an energy source embodied as a moving field of high thermal energy. When this thermal energy causes physical melting of the supplied material, these processes are widely known as melting processes. In melting processes, the material, which is fed stepwise and sequentially by the energy source, is melted in a manner similar to melt-welding.

[0005] When the feed material is in the form of powder layers, after each incremental layer of powder material is sequentially fed to the part being manufactured, the heat source melts the incrementally fed powder by melting regions of the powder layer, creating a moving melt region, referred to hereafter as a melt pool, so that these regions, upon solidification, become part of the previously incrementally fed, melted, and solidified layers below the new layer, resulting in the production of the part. Since additive manufacturing processes can be lengthy and involve any number of passes of the melt pool, it can be difficult to prevent even small variations in the size and temperature of the melt pool, which is used to densify the part.It should be noted that additive manufacturing processes are typically performed by a computer numerical control (CNC) machine because high movement speeds of the heating element and complex shapes are required to form a three-dimensional structure.

[0006] One way to measure and characterize the final quality of a part is to add one or more sensors to an additive manufacturing tool that provide process measurements during the additive manufacturing process. The process measurements may be performed by sensors configured to precisely monitor a temperature of the melt pool as it produces the part. In this way, any variation in the temperature of the melt pool during manufacturing can be recorded and characterized. In some embodiments, temperature deviations that exceed a certain limit can be recorded for later analysis, which may lead to the determination of whether or not a part meets a set of quality assurance standards.The analysis may involve collecting data from multiple sensors to determine specific cooling rates of the material during the manufacturing process. State variables may be obtained from the aforementioned sensor measurements (e.g., measurements characterizing the current state or the temporal evolution of physical process behavior) and used to determine the presence of any microstructural deviations or even cracks as a result of the deviations recorded by the sensors. Example state variables include cooling rates, heating rates, peak temperatures, and phase transition information, which may be assigned to different regions for specified sections on each layer of a part produced by the additive manufacturing process.

[0007] According to embodiments of the present invention, methods and systems are provided for determining the quality of a part produced by additive manufacturing. The quality assurance system may monitor the additive manufacturing process in real time using a number of different sensors. The quality assurance system may operate by calculating process state variables from sensor readings taken in both a Lagrangian and Eulerian reference frame during an additive manufacturing process. The process state variables may then be used to identify areas of the part likely to contain microstructural defects. In some embodiments, the additive manufacturing process may include the creation of a control region, which may be destructively inspected without causing damage to the part.

[0008] In some embodiments, an additive manufacturing process may be performed by at least performing the following steps: depositing a layer of metallic material; melting a region of the layer of metallic material to form a part manufactured by the additive manufacturing process with a heat source that scans over the region of the layer of metallic material to melt the region; monitoring an amount of energy emitted by the scanning heat source with a first optical sensor that follows a path along which the heat source scans the region to provide a first set of information; monitoring a predetermined portion of the region of the layer of metallic material with a second optical sensor to provide a second set of information;and subsequent to melting the region of the layer of metallic material, determining whether the information sets indicate that the region falls within a known-good range of a base data set associated with the part being produced by the additive manufacturing process by: correlating data contained in the second information set with data contained in the first information set, the correlated data being collected from the first and second information sets while the heat source passed through the specified portion of the region; calculating a plurality of state variables using at least a portion of the first information set and at least a portion of the second information set; and comparing the plurality of state variables with a plurality of ranges to which state variables of the known-good range of the base data set are associated.

[0009] In some embodiments, an automated additive manufacturing apparatus for producing a part on a powder bed comprises at least the following elements: a heat source; a processor; a scanning head configured to direct energy received from the heat source toward a powder layer provided on the powder bed in a pattern specified by the processor and corresponding to a shape of the part; a first optical sensor configured to determine a temperature associated with a specified portion of the part; and a second optical sensor configured to receive light emitted through the scanning head from a portion of the powder layer melted by the energy of the heat source;wherein the processor is configured to receive sensor data from the first and second optical sensors during a generative manufacturing process to characterize a quality of different portions of the part;

[0010] In some embodiments, an additive manufacturing method for determining a baseline data set for manufacturing a part comprises at least the following steps: collecting temperature data recorded by a plurality of sensors for a respective plurality of layers deposited during a plurality of additive manufacturing operations for manufacturing a part, wherein a first portion of the plurality of additive manufacturing operations is performed using setpoint parameter ranges and a second portion of the additive manufacturing operations predominantly uses non-setpoint parameter ranges, wherein the non-setpoint parameter ranges are areas expected to produce undesirable material defects in the part;Performing metallurgical testing at a specified location of each of the parts, wherein the specified location corresponds to a location on the part at which a field of view of a first optical sensor of the plurality of sensors remains fixed during each of the plurality of additive manufacturing operations and a field of view of a second optical sensor of the plurality of sensors periodically passes the location; Categorizing the sensor data collected by the plurality of sensors into target and non-target data ranges; and Creating a baseline data set for the part that includes process limits for the sensor data that have been shown to result in the part having acceptable material properties.

[0011] It should be noted that the aforementioned process is used in this description for exemplary purposes only and processes described herein may, with some modifications, also be used for other additive manufacturing processes, including all of the following: selective heat sintering, selective laser sintering, direct metal laser sintering, selective laser melting, fused deposition, and stereolithography. BRIEF DESCRIPTION OF THE ILLUSTRATIONS

[0012] The disclosure will be readily understood from the following detailed description taken in conjunction with the accompanying drawings, wherein like reference numerals describe similar structural elements. Fig. Figure 1 is a flowchart showing how process measurements and data relate to quality. Fig. Figure 2A shows a simplified version of the flowchart shown in Fig. 1 is shown. Fig. Figure 2B shows a possible process domain based on process data represented by state variables instead of process inputs. Fig. Figure 3 shows a schematic view of an exemplary additive manufacturing process based on a moving region of high thermal energy that creates a melt pool or a highly thermally influenced region. Fig. 4A to Fig. 4B show perspective views of an additive manufacturing system using a scanning laser beam and sensors, where the sensors are used to perform process measurements. Fig. Figure 5 shows a schematic view showing the collection of Lagrangian and Eulerian data on a control area or control region. Fig.Figure 6 shows a schematic view illustrating the radiation observation factor between a differential element and a disk, both lying in parallel planes but with an offset of their respective central axes. Fig. Figure 7 shows a graph of the effectiveness of a silicon photodiode as a function of the wavelength of the light incident on it. Fig. Figure 8A is a flowchart illustrating a process for creating a base parameter set for manufacturing a part according to an embodiment of the present invention. Fig. 8B is a flowchart illustrating a process for classifying a quality of a production-level part based on the generated base parameter set according to an embodiment of the present invention. Fig.Figure 9 is a logic flowchart and decision tree for accepting a production product based on process data and analysis of a control area. DETAILED DESCRIPTION OF THE SPECIFIC EMBODIMENTS

[0013] Embodiments of the present invention relate to methods and systems for performing quality assurance control during additive manufacturing processes.

[0014] Additive manufacturing, or the step-by-step and sequential assembly or construction of a part through the combination of material input and applied energy, takes many forms and already exists in many specific applications and embodiments.

[0015] 3D printing, or additive manufacturing, is any of several processes for creating a three-dimensional part of virtually any shape from a 3D model or from an electrical data file generated from a scan of a model or a 3D CRD rendering. These different processes each involve the sintering, solidification, or melting of a liquid, powder, or granular raw material, a layer-by-layer application of ultraviolet light, a high-energy laser, or an electron beam.

[0016] An electron beam process (EBF3) was developed at NASA Langley Research Laboratory. It uses a solid wire as a feedstock in a vacuum environment, and where possible, in gravity-free space capsules. The process is notable for its economical use of raw material. A focused, high-energy electron beam is converted and generates a molten pool on a metallic surface, into which the wire stock is fed, guided along a coded deposition path. It has been used to manufacture components ranging in size from fractions of an inch to several Ful, limited only by the size of the vacuum chamber and the quantity and composition of the available wire feedstock.

[0017] Selective heat sintering (SHS) uses thermoplastic powders that are fused by a heated print head. After each layer is fused, it is lowered by a movable base plate, and a layer of fresh thermoplastic powder is added to prepare for the next pass through the print head.

[0018] Selective laser sintering (SLS) uses a high-energy laser to fuse thermoplastic, metallic, and ceramic powders. This is also a scanning technology, with the laser path for each layer derived from a 3D model program. During the manufacturing process, the part is lowered by a movable support by exactly the thickness of a powder layer, keeping the laser focus at the powder level.

[0019] Direct metal laser sintering (DMLS), which is almost identical to selective laser sintering, has been used with almost every metal and alloy.

[0020] Selective laser melting (SLM) has been used for titanium alloys, chromium-cobalt alloys, stainless steels, and aluminum. Instead of sintering, the material is completely melted using a high-energy laser to produce highly dense components in a layered structure.

[0021] Fused deposition modeling (FDM) is an extrusion process in which a heated nozzle melts and extrudes small drops of material that instantly harden upon tracing a pattern. The material is fed as a thermoplastic filament or metal wire wound on a spool and dispensed through the feed nozzle. The nozzle position and flow are computer-controlled in three dimensions.

[0022] One way to measure and characterize the quality of a metal part produced using an additive manufacturing process is to add a number of temperature-characterizing sensors to an additive manufacturing tool set. These sensors record and characterize the heating and cooling that occur during the formation of each layer of the part. This recording and characterization can be accomplished by sensors configured to precisely record the temperature of portions of each layer undergoing heating or cooling at any given time during the manufacturing process.When a heat source along the lines of a laser generates the heat necessary to fuse each layer of added material, the heated portion of the layer can take the form of a molten pool, whose expansion and temperature can be recorded and characterized by the sensors. To determine the quality of each layer of the part, real-time or post-process analysis of the recorded data can be performed. In some embodiments, the recorded temperatures for each part can be compared and contrasted with temperature data collected during manufacturing of parts with acceptable material properties. In this way, the quality of the part can be determined based on the characterization of each temperature variation that occurred during manufacturing of the part.

[0023] These and other embodiments are described below with reference to the Fig. 1 to Fig. 9; however, those skilled in the art will readily appreciate that the detailed description given herein with reference to these figures is for exemplary purposes only and is not to be construed as limiting.

[0024] Fig. Figure 1 shows a block diagram describing how QUALITY 100 relates to different components of an additive manufacturing process. QUALITY 100 is defined as the ability of a manufactured part or item to meet PERFORMANCE REQUIREMENTS 101 of a larger system of which it is a component.

[0025] These PERFORMANCE REQUIREMENTS 101 are functions of the development system (e.g., aircraft, automobile, etc.), but they include important PROPERTIES 102 of the part that must be met. Examples of such properties include, but are not limited to: a physical dimension of the part, a surface roughness and surface finish of the part, a static tensile strength, thermophysical properties (e.g., density, thermal conductivity, etc.), life and dynamic endurance properties such as fatigue resistance, impact resistance, fracture toughness, etc.

[0026] The properties 102 of a part manufactured from any substance are determined by the material structure 103 of the material from which the part is made, as well as the defect distribution 104 of any anomalies, defects, or other imperfections that exist in the part—whether on its surface or in its volume. Both the material structure 103 and the defect distribution 104 are a function of the process conditions 105 under which the part is manufactured.

[0027] PROCESS CONDITIONS 105 can be given by CONTROLABLE PROCESS INPUTS 106, UNCONTROLABLE PROCESS VARIATIONS 107 and ENVIRONMENTAL INFLUENCES 108. The result of the CONTROLABLE PROCESS INPUTS 106, the UNCONTROLABLE PROCESS VARIATIONS 107 and the ENVIRONMENTAL INFLUENCES 108 is the combination of physical behaviors that occur during the manufacturing or additive manufacturing process, which is referred to as PHYSICAL PROCESS BEHAVIOR 109.

[0028] Associated with each PHYSICAL PROCESS BEHAVIOR 109 may be one or more variables that can be used to either directly or indirectly measure the current status of the manufacturing process. These are called PROCESS STATUS VARIABLES 110. These are status variables in the original sense of the definition, meaning that complete knowledge of these PROCESS STATUS VARIABLES 110 completely describes the current status of the manufacturing process. PROCESS STATUS VARIABLES 110 may, for example, include a rate at which various regions of the part are heated or cooled. The cooling rate may be extrapolated by measuring a temperature of a surface of one or more regions of the part during its cooling. In some embodiments, temperature data may be determined optically via sensors such as a pyrometer, an infrared camera, and / or a photodiode.The temperatures obtained in this way can also be used to predict the values of other state variables, such as the times at which solidification or melting occurs. The state variables can also include a peak temperature reached for a specific region or section of the material used to manufacture the part.

[0029] Thus, it is a way to determine the QUALITY 100 of a part produced by an additive manufacturing process by measuring parameters during the additive manufacturing process that can be used to obtain the PROCESS STATUS VARIABLES 100. This results in a CORRELATION BETWEEN PROCESS STATUS VARIABLES AND POST-PROCESS QUALITY 111 and forms the basis for some embodiments of the present invention, which are described herein.

[0030] In Fig.2, a more concise reformulation of the above description is as follows: By measuring PROCESS STATUS VARIABLES 200, it is possible to understand the current status of the process; by knowing the current status of the process, it is possible to understand the PHYSICAL PROCESS BEHAVIOR 201 and classify it as NOMINAL 202 or NON-NOMINAL 203. This classification is based on the existence of a POSSIBLE PROCESS SPACE 204 defined in the system of PROCESS STATUS VARIABLES 200, where the process is, by definition, in a NOMINAL 202 state if it lies within the boundaries of the POSSIBLE PROCESS SPACE 204. It should be noted that, although the POSSIBLE PROCESS SPACE 204 appears two-dimensional, in fact, significantly more variables can contribute to the definition of the POSSIBLE PROCESS SPACE 204.Thus, according to some embodiments of the present invention, the state variables, rather than the input variables used to control the system, are used to define the possible process space. The input variables, such as laser energy and scan speed, which are typically used to define the possible process space, result in physical process behaviors, such as melt pool temperature. In some embodiments, other physical process behaviors include, for example, a melt pool temperature gradient, melt pool volume, a natural frequency of the melt pool, vaporization at the melt pool, spectral emission of the melt pool, such as infrared emission of the melt pool and optical emission of the melt pool, and others. Both intrinsic (i.e., dependent on the melt pool volume) and extrinsic (i.e.,Physical behaviors (independent of the volume of the melt pool) are included within the scope of physical process behaviors encompassed by the present invention. As described in more detail herein, these physical process behaviors can be measured to obtain process variables, also called state variables.

[0031] This process quality control method can be applied to a wide range of manufacturing processes. However, with a focus on additive manufacturing processes, it is useful to consider the class of additive manufacturing processes in which a moving molten, plasticized, or otherwise thermally affected region migrates across the surface of the part being manufactured. The material being added is either pre-placed, as in the case of a powder bed process, or it can be added to the molten, plasticized, or otherwise thermally affected region.

[0032] Fig.Figure 3 illustrates the key physical phenomenon that occurs during an additive manufacturing process, as described above. The substrate 300 is the part being built. An energy source 301 impinges on the surface of the substrate 300. The energy source 301 has a moving velocity 302, expressed by the symbol v, and generates a molten, plasticized, or otherwise thermally affected region 303 as it moves along a path on the substrate 300. Immediately following the moving thermally affected region 303, a region 304 of the substrate 300 results that has undergone a thermal cycle, i.e., has been thermally affected and cooled back to a nominal temperature of the substrate 300.For example, in the case of a laser sintering process performed on a pre-placed powder bed, the thermally cycled region 304 corresponds to the trace of powder melted and solidified, or sintered, by the moving energy source 301. This thermally cycled region 304 will generally have a cross-section 305 below the surface of the substrate 300.

[0033] Given a comprehensive energy balance for the moving energy source 301, radiated and conducted energy will occur, which also provides excellent indications and information regarding the key physical phenomena occurring in the thermally affected region 303. For example, the thermal conduction 306, indicated by the heat flux Q, will result in a heat flow from the thermally affected region 303 and the region 304 that has undergone a thermal cycle. This flow will be generally perpendicular to the outlines of the profile 305 of the thermally affected region below the surface of the substrate 300.In addition, radiation signals and information 307 are emitted, which may be a form of optical radiation, or, in the case that the additive manufacturing process in question is carried out in a controlled atmosphere and not in a vacuum, acoustic radiation. Finally, back-reflected signals 308 may be present, which are 100% collinear with the incident energy source 301 or may occur at small offset angles. In the case that the influencing energy source is, for example, a laser operating in the near-infrared, the backscattered signal 308 could be optical radiation that passes back through the laser optics but does not interfere with the incident beam because the incident beam is in the near-infrared.These signals and others that can indicate the status of the equipment and the status of the process together form the PROCESS STATUS VARIABLES that define the current status of the PHYSICAL PROCESS BEHAVIOR that determines QUALITY.

[0034] Fig.4A is a schematic diagram illustrating a quality control system 400 according to an embodiment of the present invention. The quality control system 400 may be used in conjunction with additive manufacturing processes where the moving heat source is a laser and the material supply is accomplished either via the sequential pre-placement of layers of metallic powder to form a volume of powder 401, which also includes a powder bed 402 as described, or the material supply may be accomplished by selectively placing powder directly into the molten region created on the part by the moving laser. The volume of powder 401 comprises a plurality of separate fabrication regions 403 that are constructed.In the case of the illustrated embodiment, build is accomplished by applying the heat source to the material build regions 403, causing the deposited powder in these regions to melt and then solidify as a part having the desired geometry. The different regions 403 may be different sections of the same part, or they may represent three completely different parts, as shown.

[0035] As in Fig.4A, a control region 404 is provided. The control region 404 is a standardized volume element, referred to as a control region, which allows for a sample to be provided for each manufacturing step and represents a small and controllable, yet still representative, portion of the material, which can be destructively tested for its metallurgical integrity, physical properties, and mechanical properties. For each deposited layer, the control region 404 also includes a layer of the material that is simultaneously deposited with the layer in the separate regions 403. An optical sensor 405, e.g., a pyrometer, is provided that directly monitors the control region 404.For clarity, the optical sensor 405 is represented here as a pyrometer, although it will be apparent to those skilled in the art that other optical sensors may also be used. The pyrometer 405 is fixed with respect to the powder bed 402 and collects radiation from a fixed portion of the volume of powder 401, i.e., the control region 404.

[0036] In the case where the additive manufacturing process includes a scanning laser impinging on a powder bed 402, the laser source 406 emits a laser beam 407, which is deflected by a partially reflecting mirror 408. The partially reflecting mirror 408 can be configured to reflect only those wavelengths of light associated with the wavelengths of the laser beam 407, while allowing other wavelengths of light to pass through the partially reflecting mirror 408. After the laser beam 407 is deflected by the mirror 408, it reaches the scan head 409. The scan head 409 can include an internal X-deflection, a Y-deflection, and focusing optics. The deflected and focused laser beam 407 leaves the scanning head 409 and forms a narrow, hot, and migrating melt pool 410 in the different manufacturing regions 403, which are melted or sintered layer by layer.The scan head 409 may be configured to maneuver the laser beam 407 at high speeds across a surface of the bulk powder 401. Note that in some embodiments, the laser beam 407 may be activated and deactivated at specific intervals to prevent heating of the portions of the bulk powder 401 over which the scan head 409 would otherwise direct the laser beam 407.

[0037] The melt pool 410 emits optical radiation 411, which travels back through the scan head 409 and passes the partially reflecting mirror 408 to be collected by the optical sensor 412. The optical sensor 412 collects optical radiation from the migrating melt pool 410 and thereby images different portions of the volume of powder 401 as the melt pool 410 traverses the volume of powder. A sampling frequency of the optical sensor 412 will generally determine how many data points can be acquired as the melt pool 410 scans across the volume of powder 401. The optical sensor 412 can take many different forms, including a photodiode, an infrared camera, a CCD array, a spectrometer, or any other optically sensitive measurement system. For example,When a spectrometer is used, data regarding the chemical content of the melt pool can be obtained, providing insight into the materials / species that also evaporate from the melt pool, or additionally, the materials / species that remain in the melt pool. In addition to the pyrometer 405 and the optical sensor 412, the quality control system 400 may include another optical sensor 413. The optical sensor 413 may be configured to obtain optical information over a large field of view 414 to enable real-time monitoring of substantially the entire volume of the powder 401. Like the optical sensor 412, the optical sensor 413 may take many different forms, including a photodiode, an infrared camera, a CCD array, and the like.By adding the optical sensor 413, which continuously monitors the entire volume of powder 401, to the quality monitoring system 400, the monitoring system 400 receives an additional set of sensor data including Eulerian data for each point on the volume of powder 401. In configurations where the optical sensor 413 is configured to distinguish relative amounts of emitted heat, the readings from the pyrometer 405 can be used to calibrate the optical sensor 413 so that heat readings can be continuously recorded across the entire surface of the volume of powder 401 and analyzed for irregularities. Furthermore, quantitative temperature information can be measured at all areas of the volume of powder 401 using the optical sensor 413.

[0038] Fig.4B shows an alternative arrangement in which a second pyrometer 415 may be arranged to monitor a different control region 416. By including a second pyrometer, the control region 416 can be used to capture conditions corresponding to an out-of-parameter thermal burst, i.e., when deviations occur that exhibit a temperature gradient beyond the known good performance parameters while the melt pool 410 is passing through the control region 416. In some embodiments, the analysis can be switched entirely to the control region 416 when a given circumstance requires it. In this way, a deviation that occurs in the wrong place and / or at the wrong time no longer impacts the ability to characterize the part via the analysis of a control region.In some embodiments, the accuracies of a first and a second pyrometer may be quite different. For example, the first pyrometer 405 may have a significantly higher sensitivity to temperatures than pyrometer 415. Other differences between the pyrometers, such as the size of the base area scanned by each pyrometer, are possible.

[0039] While both Fig. 4A and Fig.4B show and demonstrate the use of a control region, it should be understood that in some cases, once the manufacturing process is properly understood, one or more of the pyrometers may instead be focused on a portion of one of the manufacturing regions 403. While such a configuration precludes the destructive analysis of a portion of a production part, once the process is well understood, confidence in the described thermal analysis may be high enough to accept a part without a destructive analysis of a control region and for production runs in which thermal bursts do not exceed a predetermined limit. In yet another embodiment, when multiple parts are manufactured simultaneously, one of the parts may assume the role of the control region.In this way, one of a number of parts having the same size and geometry can be analyzed to gain additional insight into the temperature characteristics experienced by the other parts and to more accurately predict the core structure of the other parts manufactured in parallel.

[0040] When the melt pool 410 passes through the control area 404, both the Eulerian pyrometer 405 (i.e., the pyrometer 405 monitors a fixed section of the region of the metallic material being additively processed, taking measurements in a stationary reference frame) and the Lagrangian optical sensor 412 (i.e., the optical sensor 412 monitors the area where the laser energy impinges, taking measurements in a moving reference frame) observe the same area in space. The signals from the Eulerian pyrometer 405, the Lagrangian optical sensor 412, and the optical sensor 413 are present in the control area, which is a condition for the control area. Thus, the calibration of the sensor measurement data can be performed when the melt pool overlaps the control area.In an embodiment in which a narrow-focus Eulerian photodetector only receives radiation from the area of the control region provided in conjunction with the control region (not shown), calibration of the optical sensor 412 may be performed when the melt pool overlaps with the control region.

[0041] In some embodiments, a narrowly focused photodiode is directed toward the control region. In these embodiments, the photodiode collects spectral emissions from the control region, which is transformed into a molten pool when the laser source passes through the control region. The spectral emissions may be ultraviolet, visible, or infrared, depending on the temperature of the molten pool. In some applications, multiple photodiodes may be used to collect spectral emission across a plurality of spectral bandwidths. The photodiode may be used to collect the spectral emissions, and these measurements may be correlated with status variables such as the size of the molten pool, the temperature of the molten pool, the molten pool temperature gradient, and the like. One skilled in the art will recognize many variations, modifications, and alternatives.

[0042] In Fig.5, the control region 500 is shown for a given slice. The Lagrangian optical sensor 412 will operate at a finite sampling rate as the beam scans the control region, and it will collect data at discrete sampling positions 501. The Eulerian pyrometer will examine a fixed field of view 501 that lies within the larger control region 500. In general, a set of Lagrangian optical sensor measurement data 503 (which can be assumed to be within the field of view of the optical sensor 412) will be obtained that lies within the field of view of the Eulerian pyrometer 502. This will be true on a slice-by-slice basis. Thus, both the Lagrangian and Eulerian measurements are available in the control region 500. Furthermore, the control region 500 becomes the target of subsequent destructive examination.This provides an opportunity to further correlate microstructural and even mechanical property data with the relationships obtained layer by layer during manufacturing.

[0043] Although the Lagrangian optical measurements 503 are shown smaller than the field of view of the Eulerian pyrometer 502, this is not necessary for the present invention. In some embodiments, an optical image sensor may be used as the optical sensor 412 to obtain an image of the control region as well as other regions. In these embodiments, process status variables, such as the melt pool size, may be obtained using the data collected by the optical sensor 412. One skilled in the art will recognize many variations, modifications, and alternatives.

[0044] Generally speaking, the signal from a Lagrangian optical sensor will be a function of the optical or infrared energy emitted from the melt pool and collected by the scanner optics into the optical sensor. This will be subject to several factors that result in an overall transfer function that establishes a connection between the emitted radiation at the source and the measured signal at the detector. In the most general case, the transfer function can be represented as: T=T{ε,dA,F(x,y),ρmirror,σsensor}

[0045] Here, ε is the radiance of the area of the melt pool which is irradiated, dA is the area of the melt pool which is irradiated and is assumed to be small compared to the area of the output lens of the scanning unit, F(x,y) is an imaging factor which relates the small area of the melt pool to the area of the output lens on the scanner, ρ mirroris the wavelength-dependent reflectivity of the mirror that splits the sensor signal while allowing the primary laser energy to pass through and σ sensor is the wavelength-dependent sensitivity of the optical sensor with respect to the incident radiation.

[0046] Thus, the general relationship between the signal measured by the optical sensor and the energy emitted or radiated by the melt pool at a given position and at a given time is: S(x,y,t)=T(x,y)⋅Eweldpool(x,y,t)

[0047] The imaging factor can be approximated as shown in Fig.6. The small area of the melt pool 600 is represented by dA1 and the area of the exit lens of the scan head 601 is represented by A2. However, the melt pool is generally not located directly below the exit lens of the scan head 601 in the plane of a top layer of the powder bed and is displaced by a distance 602 in the plane, which is represented by a. The working height 603 is the distance from the exit lens of the scan head 601 to the powder bed and is represented by h. The exit lens of the scan head 601 has a radius 604, which is represented by r.

[0048] The imaging factor is expressed by the following mathematical relationship, which was originally derived by Hamilton and Morgan: Fd1−2=12[1−Z−2R2(Z2−4R2)1 / 2] where H=ha,R=ra,and Z=1+R2+H2

[0049] The variable a can also be related to the x and y positions on the plane of the powder bed. If we assume that the position directly below the center of the scan head's exit lens is the origin of a coordinate system in the plane of the powder bed, then the variable a is related to the x and y positions of the melt pool using the following formula: a=x2+y2

[0050] These x and y positions can be obtained sequentially from the drive signal that controls the beam deflection within the scan head. In high-speed laser scanners, for example, these x and y positions can be controlled by mirrors driven by high-frequency response galvanometers.

[0051] The reflectivity of the mirror is defined in a range of wavelengths over which the mirror reflects radiation within that wavelength with a high degree of reflectance, and outside that range the mirror is essentially transparent. Thus, the mirror reflectance will be very high for radiation radiating from the melt pool and passing back through the scan head's output lens for an observation window of frequencies defined as follows: ωMIN<ωRADIATION<ωMAX

[0052] The sensitivity of the optical sensor depends specifically on the type of sensor used. A typical efficiency curve for a silicon photodiode, for example, is shown in Fig. 7. The curve describes the efficiency with which light is converted into electricity. Thus, the y-axis of Fig. 7 the conversion efficiency. The x-axis of Fig.7 is the frequency of the incident radiation collected by the photodiode.

[0053] Consequently, it can be seen that the transfer function as described in formula 1 is actually based on the knowledge of the various factors defined in formulas 3-5 and the sensitivity of the sensor as described in Fig.7. This makes it possible to perform a transformation that brings the radiation collected at any position in the powder bed or plane of the part into a reference frame of any other region of the part, with which a comparison can be made. More specifically, in this invention, such a comparison is performed between the control region and any other region of the part. A concrete example will now be discussed that will further explain how such a transfer function can be used to effectively compare a control region located immediately below the center of the scan head's output lens with any region in the part or powder bed plane.

[0054] In general, the radiation flux collected at the output lens of the scan head is related to the flux emitted from the melt pool via the imaging factor, as shown in formulas 2-4. The mirror will have minimum and maximum cutoff frequencies, which determine the window of frequencies within which the radiation is allowed to pass through to the photodiode collector. The photodiode collector will have a conversion efficiency as shown in Fig. 7, where the average conversion efficiency is related to the cut-off frequencies of the mirror by the following equation: faverage=f(ωMIN)+f(ωMAX)2

[0055] The effectiveness can be determined at the respective section frequencies from Fig.7. Thus, the overall transfer function, which relates the energy radiated by the melt pool at a point (x,y) and at a time t to the electrical signal measured by the sensor (in this case the photodiode), can be expressed in first order as follows: S(x,y,t)=T(x,y)⋅Eweldpool(x,y,t)=ε⋅faverage⋅Fd1−2(x,y)∗Eweldpool(x,y,t)

[0056] Here S(x,y,t) is the sensor signal of the emitted radiation when the melt pool is at position (x,y) at time t, and E weld pool (x,y,t) is the actually emitted radiation in energy per time per area, which is radiated from the location (x,y) at time t. Therefore, formula 7 must be transformed to E weld pool be resolved in order to accurately compare the energy emitted at a given position with the energy emitted at another position: Eweldpool(x,y,t)=S(x,y,t)ε⋅faverange⋅Fd1−2(x,y)

[0057] The purpose of formula 9 is to normalize the measured optical signal in order to more accurately compare data taken at different (x,y) locations in the powder bed or plane of the part being systematically manufactured layer by layer.

[0058] Fig. 8A is a flowchart illustrating a process 800 for obtaining a base parameter set for manufacturing a part according to an embodiment of the present invention. Referring to Fig.8A, the method includes collecting and analyzing overlapping Eulerian and Lagrangian sensor data during one or more additive manufacturing processes using setpoint parameter ranges 801. In some embodiments, the overlapping portion of the sensor data relates to material that is distinct and different from the part being manufactured (sometimes this portion may serve as a control region), while in other embodiments, the overlapping sensor data relates to a portion of the part itself. In cases where the overlapping sensor data is located within the part itself, it may be necessary to remove the portion of the part if verification of the microstructural integrity of that portion is desired without destroying the part.The Eulerian and Lagrangian sensor data can be collected from a variety of sensors, such as pyrometers, infrared cameras, photodiodes, or the like. The sensors can be arranged in a variety of different configurations; however, in a specific embodiment, a pyrometer can be configured as an Eulerian sensor, focused on a fixed portion of the part, and a photodiode or other optical sensor can be configured as a Lagrangian sensor, following the path of the heating element as it scans across the part.

[0059] Data collection begins with testing setpoint parameter ranges (i.e., those parameters or control inputs that are likely to result in, or have resulted from, acceptable microstructure and / or acceptable mechanical properties and / or acceptable defect structures for a particular metal being used). In some embodiments, a user may begin with more or less precise parameter ranges during the establishment of setpoint parameter ranges. It should be understood that beginning with a more precise setpoint parameter range may reduce the number of iterations required to obtain a representative number of data points that fall within the part's setpoint parameter ranges.When using a control region, it should be noted that the Lagrangian data can be transformed for the range of the control region using the transfer function as presented in Formula 9.

[0060] Once a representative number of data points corresponding to the presence of acceptable material properties in the part has been collected, additional additive manufacturing processes are performed using non-nominal parameter ranges. During these manufacturing processes, the overlapping Eulerian and Lagrangian sensor data are collected and analyzed (802). Similar to the data collection method used for nominal data collection, the sensors can focus on the same section of the part used to collect nominal data. The Lagrangian data is again transformed using Formula 9. Non-nominal parameter ranges are such parameter ranges (e.g., laser power, scan speed, etc.).) that have been verified to result in unacceptable microstructure and / or unacceptable mechanical properties and / or unacceptable defect structures, obtained by subsequent destructive analysis of the control area or equivalent regions of the structure. Off-target data collection may include multiple part manufacturing processes to determine boundaries or thresholds at which a part is known to be defective. Off-target data collection may also include test runs in which the laser power is periodically decreased and increased, with otherwise target parameters being used, to characterize the effect that temporary parameter deviations may have on the manufacture of a part.As described in detail below, the collection and analysis of process sensor data during a set of manufacturing processes employing non-setpoint parameter conditions can be used to establish process limits for the process sensor data. Embodiments of the present invention measure characteristics of the process (i.e., process sensor data) in addition to measuring characteristics of the manufactured part.

[0061] In 803, one or more sections of the part where the Eulerian and Lagrangian sensor data overlap (i.e., the control region) are analyzed to enable the generation of a baseline data set. In general, there are three types of analyses that can be applied to a control or equivalent region of the part. First, the microstructure can be examined in detail. This includes, but is not limited to, analyses of grain size, grain boundary orientation, chemical composition in a macroscopic or microscopic scale, condensate size and distribution in the case of age-hardening alloys, and grain size of precursors that may have formed first, with clear indication of such premature graining. The second category of evaluations that can be performed are mechanical property testing procedures.These include, but are not limited to, analyses of hardness / microhardness, tensile properties, elongation / ductility, fatigue behavior, impact strength, fracture toughness, as well as measurements of crack growth, thermomechanical fatigue, and creep fatigue. The third set of evaluations, which can be performed on a control area or equivalent regions of the component, is the characterization of defects and anomalies. This includes, but is not limited to, the analysis of pore shape, size, or distribution; the analysis of crack size or distribution; evidence of direct melt inclusions (i.e., those formed during gas atomization of the powder itself), other inclusions inadvertently introduced during the additive manufacturing process; and other known welding defects such as lack of fusion.It should also be noted that in certain cases, a position of the control area or focus of the pyrometer can be adjusted to obtain a more accurate representation of the respective critical section of the part.

[0062] Once both the process sensor data (Eulerian and transformed Lagrangian data) and the post-process data (microstructural, mechanical, and defect-related characterizations) have been collected, it is possible in step 804 to use a wide variety of outlier detection schemes 804 and / or classification schemes that separate the data into nominal and non-nominal conditions. Likewise, the process conditions that result in a particular set of post-process data can be characterized, with the associated process data being collected while the sample is being manufactured. This process data, both Eulerian and Lagrangian, can be associated and correlated with post-process part characterization data. This allows a connection to be established between clear post-process conditions and the process properties in the form of process data that give rise to these post-process conditions.More specifically, properties extracted from process data can be directly linked and correlated with properties obtained from post-process analysis. In some embodiments, data collected during manufacturing using target parameter ranges differs from data collected during manufacturing using non-target parameter ranges, such as two different cluster diagrams. One skilled in the art will recognize many variations, modifications, and alternatives.

[0063] Once such properties are established and correspond to both what was described in real time and what was described post-process, a process window may be defined in step 805 based on the process boundaries of both Eulerian and Lagrangian data, which correspond to target conditions, i.e., conditions that have been verified to result in acceptable microstructure and / or acceptable mechanical properties and / or acceptable defect structures, derived by post-process destructive analysis of the control area or equivalent regions of the part.The practical significance of achieving this state thus arises from the fact that the process can be defined as being within a target regime, advantageously via actual process measurements that are directly linked to the physical behavior occurring during the additive manufacturing process, and advantageously in contrast to defining such a process window using ranges of machine settings or variables contained in a process parameter set that are additionally extracted from the process. In other words, embodiments of the present invention differ from conventional systems that only specify process parameters.Embodiments of the present invention determine the process data for both target parameter ranges 801 and non-target ranges 802, which provide a "process fingerprint" for a known set of conditions. Given this established baseline data set, it is possible, for each material of interest and each set of process conditions, to accurately predict whether the resulting manufacturing result of a known-good product will exhibit the desired metallurgical and / or mechanical properties.

[0064] It should be understood that the specific steps as described in Fig.8A, provide only a single method for generating a basic parameter set for manufacturing a part according to an embodiment of the present invention. According to alternative embodiments, other workflows may also be performed. For example, for alternative embodiments of the present invention, the steps discussed above may be performed in a different order. Furthermore, the individual steps described in Fig. 8A, include multiple substeps performed in multiple sequences corresponding to the individual steps. Furthermore, depending on the particular application, additional steps may be added or removed. One skilled in the art will recognize many variations, modifications, and alternatives. In the following, attention is specifically drawn to the use of a process window in a production environment.

[0065] Fig. 8B is a flowchart illustrating a process 806 for classifying a quality of a production step part based on the obtained base parameter set, according to an embodiment of the present invention. Fig. Figure 8B shows a process 806 describing the use of the base data set in a build scenario. The base data set can be created using the method described in Fig. 8A is illustrated.

[0066] A block 807 represents the collection of long-range data from (x,y) locations distributed across the manufacturing plane and corresponding to Eulerian data of a specified section within the manufacturing plane during an additive manufacturing process. In a particular embodiment, the Lagrangian data may be collected from a photodiode and the Eulerian data may be collected from a pyrometer. The specified section may be a control region or a region of the part that is subsequently removed from the part for testing. In some embodiments, the Lagrangian data may be collected from all sections in the manufacturing plane, while the Eulerian data is collected only in a specified section of the control region, but the present invention is not limited to these embodiments.In other embodiments, a subset of all possible sections is used to collect the Lagrangian data. The Lagrangian data is collected in a specified section of the control region as the melt pool passes through the section of the control region. One skilled in the art will recognize many variations, modifications, and alternatives.

[0067] Block 808 describes a verification process that may be performed to determine whether the Eulerian and Lagrangian data collected at the specified section are free of data points that lie outside the target baseline dataset (i.e., within the region defined by the baseline dataset). The same classification and outlier detection scheme used during baseline generation in process 800 may be used to perform this verification. In other words, this step proves that overlapping Eulerian and Lagrangian sensor readings collected during an actual production run correspond to overlapping Eulerian and Lagrangian sensor readings collected under normal conditions as part of the baseline dataset.

[0068] Block 809 describes the comparison of Lagrangian data collected at one or more (x,y) positions with Lagrangian data collected in the defined section. In some embodiments, the Lagrangian data collected at each of the (x,y) positions is compared with the Lagrangian data collected at the defined section associated with the control area. In this way, a set of Lagrangian process data associated with sections or the entire manufacturing level can be compared with a set of process data of a control area area. This step can be performed subsequently at block 808 if it is confirmed that the Lagrangian data of the defined section in the production run was within the range of the target conditions described in the base data set. Accordingly, the embodiment described in Fig.8B, compares the Lagrangian data set associated with some or all areas of the manufacturing level with the Lagrangian data set of the control area, verifying that the process data are within the limits of the base data set.

[0069] Once the verification and comparison according to blocks 808 and 809 has been successfully completed for all desired measurement points in the part, the entire part can also be classified as being within the limits of the target base data set via logical reasoning in the optional block 810.

[0070] [in block 811 can provide useful verification of part quality / conformity to the baseline data set. Block 811 describes additional verification that is performed to ensure that no anomalies occur in the manufacturing Lagrangian signal that are not present in the baseline. For example, short thermal anomalies and / or highly localized anomalies may physically represent some variation in powder sintering, the presence of a foreign object in the powder layer, a fluctuation in laser energy, melting at a highly localized [plane, or the like. [indicator of an anomaly may then be communicated to a system operator accordingly. In response to the indicator, a quality engineer may request that the part undergo additional testing to determine whether the short-term anomaly [influences part performance.The verification process in 811 may differ from that performed in 808, primarily because the time span of the respective verification processes may differ significantly. Furthermore, different thresholds may be used to obtain the appropriate filter function. For example, the verification process may be performed for each collected data point that exceeds a reasonably significant threshold, while the process in 808 only considers a small number of data points (e.g., with a reduced sampling rate) and uses a significantly lower threshold for irregular measurements. In some embodiments, block 811 may be performed optionally and is thus not necessary for the present invention. In some embodiments, the order of the verification processes in 808 and 811 is changed according to the respective application.In some embodiments, the verification process in 811 may be performed using data from a sensor different from the sensor used in block 808. For example, the sensor associated with the verification may be a high-speed camera that measures temperature data thousands of times per second. This high-speed sensor may have lower accuracy than a sensor associated with block 808, especially if it is designed to capture significant but transient deviations from the baseline data set.

[0071] Finally, block 812 describes an optional step. This optional step may be performed if general confidence in the manufactured production part is still in doubt. In such a case, the material corresponding to the specified section may be destructively tested to ensure that the post-process metallurgical, mechanical, or effect-related properties of the manufactured control region are within the same limits as those of a target baseline control region. In some embodiments, the destructive testing just described may be performed only periodically or, in some special cases, not at all.

[0072] It should be noted that as part of the process for producing manufactured parts, the computer numerical control (CNC) unit can be used to control the additive manufacturing tool set and can also be responsible for performing certain actions based on the aforementioned sensor data. For example, multiple thresholds can be set up and correlated with various actions performed by the CNC device. For example, a first threshold can initiate the recording of a parameter deviation, a second threshold can prompt the system to alert a tool set user, while a third threshold can be configured to abort the production of the part.

[0073] Conversely, if one of these conditions is not met and if the (x,y) position of the Lagrangian data is known, the particular section of the product or production run can be categorized as "non-nominal" or potentially suspect, i.e., as possibly having unacceptable microstructure, unacceptable mechanical properties, or unacceptable defect distributions.

[0074] Therefore, the Fig. 8A to Fig.8B Embodiments of the present invention as they correspond to the use of Eulerian and Lagrangian process data in a production run. Here, the link to the baseline data, more precisely the baseline data taken from a control section manufactured under nominal conditions known to generate acceptable post-process properties, and the methodology under which the Eulerian and Lagrangian process data converge in the control section, is set with the situation in connection with the process properties used for the manufacturing process in order to be able to classify a manufacturing process as nominal. This includes, for example, a representative of the baseline obtained using process properties known to generate an acceptable microstructure and / or acceptable mechanical properties and / or acceptable defect distributions.

[0075] It should be understood that the specific steps as described in Fig. 8B, provide a single way to classify a quality of a production section based on the generated basic parameter set, according to an embodiment of the present invention. Other sequences of steps may also be performed according to alternative embodiments. For example, alternative embodiments of the present invention may process the steps described above in a different order. Furthermore, the individual steps described in Fig. 8B, contain multiple substeps that can be performed in sequences corresponding to the individual steps. Furthermore, depending on the specific applications, additional steps may be added or removed. One skilled in the art will recognize many variations, modifications, and deviations.

[0076] Fig.Figure 9 briefly describes the logical flow of the decision process described in this invention for determining whether a given production part should be accepted as nominal based on real-time process data. In this flowchart, there are four boxes, all four of which generally must be met to classify a build or production run as acceptable based on real-time process data (both Eulerian and Lagrangian). It should be noted that the subsequent analysis and risk management logic can be applied to accept parts that fail to meet all of these conditions. The first step of the decision tree 900 determines whether or not the portion of the part containing overlapping Eulerian and Lagrangian sensor data (e.g.,the control region) is within the nominally known-to-be-good region that represents the baseline data set. If the control region of the production run does not have Eulerian and Lagrangian process data to meet this requirement, then the production run is marked as potentially questionable. The second step of decision tree 901 determines whether or not the Lagrangian data collected at one or more (x,y) points correspond to Lagrangian data collected in the control region for the same part. In some embodiments, as discussed above, the Lagrangian data may be collected for each point of interest in the manufacturing plane or for a subset of the locations in the manufacturing plane.For example, a plan of the part can be overlaid with the manufacturing plane to use only the Lagrangian data for positions corresponding to the geometry of the part being manufactured. In other embodiments, the Lagrangian data can be collected for portions of the manufacturing plane that correspond to the laser path, or for portions of the manufacturing plane that correspond to the laser path when the laser is on. One of ordinary skill in the art will recognize many variations, modifications, and alternatives.

[0077] The third step of the decision tree decides whether or not the post-process properties measured at the control region and associated with the part lie within the nominally well-known range of post-process properties measured at control regions and corresponding to the baseline data set. The final, fourth step of the decision tree decides whether or not significant anomalies occur in the collected Lagrangian data at the (x,y) positions, such as properties of the part's Lagrangian data being visible there that are not visible in the baseline Lagrangian data. This additional and final step is necessary because outlier detection and classification are based on properties. It is possible that new properties not originally contained in the baseline data set may appear in the Lagrangian data over time.

[0078] In order to effectively perform classification or implement an outlier detection scheme to compare data from a production run with data from a baseline dataset, properties must first be extracted from the real-time data. As an exemplary embodiment of this invention, assume that the Eulerian sensor is a multicolor pyrometer and that the Lagrangian sensor is a silicon photodiode. Furthermore, in the exemplary embodiment, the heat source is a scanning laser or electron beam, and material feeding is accomplished via pre-placement of the powder between sintering runs. The table below describes properties extracted from the corresponding Eulerian and Lagrangian process data so that an effective comparison can be made between the properties in the baseline dataset and the properties during the production run. Eulerian properties Lagrange properties Scan peak temperature: Photodiode RMS: When the laser or electron beam passes directly through the field of view of the pyrometer, this property represents the maximum temperature of the process that occurs during the high-speed deflection of the laser or electron beam through the field of view. In the case of a laser-based process, the photodiode signal represents the backscattered radiation emitted by the melt pool and collected back through the optics and beam splitter. This property represents the root mean square value of this signal intensity after the transformation correction according to Equation 9 has been applied. Scan heating rate: Photodiode standard deviation: When the laser or electron beam passes directly through the field of view of the pyrometer, this property represents the maximum heating rate of the process, which occurs during the high-speed deflection of the laser or electron beam through the field of view. In the case of a laser-based process, the photodiode signal represents the backscattered radiation emitted by the melt pool and collected back through the optics and beam splitter. This property represents the standard deviation of this Signal intensity after the transformation correction according to equation 9 has been performed. Scan cooling rate: Photodiode frequency spectrum: When the laser or electron beam passes directly through the field of view of the pyrometer, this property represents the maximum cooling rate of the process, which occurs during the high-speed deflection of the laser or electron beam through the field of view. In the case of a laser-based process, the photodiode signal represents the backscattered radiation emitted by the melt pool and collected back through the optics and beam splitter. This property represents the frequency spectrum of this signal intensity after the transformation correction according to Equation 9. Bulk peak temperature: Photodiode skew: When the laser or electron beam is not within the field of view of the pyrometer, the material will still exhibit a thermal background profile and this property is the peak temperature associated with this thermal background profile. In the case of a laser-based process, the photodiode signal represents the backscattered radiation emitted by the melt pool and collected back through the optics and beam splitter. This property represents the skewness of this signal intensity after the transformation correction according to Equation 9 has been applied. Bulk heating rate: Photodiode curvature: If the laser or electron beam is not within the field of view of the pyrometer, the material will still exhibit a thermal background profile and this property is the maximum heating rate associated with this thermal background profile. In the case of a laser-based process, the photodiode signal represents the backscattered radiation emitted by the melt pool and collected back through the optics and beam splitter. This property represents the kurtosis of this signal intensity after the transformation correction according to Equation 9 has been applied. Bulk cooling rate: When the laser or electron beam is not within the field of view of the pyrometer, the material will still exhibit a thermal background profile and this property is the maximum cooling rate associated with this thermal background profile.

[0079] All of these properties can also be averaged across a given layer. Furthermore, the Lagrangian data collected over the same region of the control domain as the Eulerian data can be considered a separate property, even though they are a subset of all Lagrangian properties.

[0080] Additionally, there are other possibilities regarding classification and outlier detection schemes. Some of these are listed in the following table, although it should be understood that a variety of possible schemes can also be implemented and still fall within the spirit and intent of this invention. Possible classification schemes and outlier detection methods Mahalanobis Distance (MD): This is a good procedure because it provides accurate knowledge of covariances in a multivariate space and provides a simple, non-subjective interpretation in which the squared MD distances are fitted into a chi-square distribution and the critical value of the chi-square distribution determines the outlier limit of the MD distances at a given confidence level. Extreme value statistics: For any individual property, or set of properties, for example, in contrast to the chi-square distribution (but with further specification of a given confidence level), the generalized extreme value distribution can be used and a similar analysis of the outliers can be performed. Arbitrary limits: Instead of algorithmically defined limits, user-defined limits based on experimental experience can be used. In some cases, this is a suitable alternative, but it is accompanied by the risk of human subjectivity and the difficulty of assessing the true-false-negative / false-positive ratio.

[0081] Even if all steps from the Fig. 8A , Fig. 8B and Fig.9 are strictly followed and the part is deemed acceptable based on process data and real-time data, the question still remains whether the process data can fully describe the quality of the component. For this purpose, in addition to the steps described in the Fig. 8A, Fig. 8B and Fig. 9, further correlations with other physically independent, subsequent, non-destructive testing methods are required. Over time, these additional tests can be outsourced or assigned to an irregular periodic sample test to ensure that the real-time data taken during the process still sufficiently captures the physics of the process to allow any conclusion about the quality, as is the case in the Fig. 8A , Fig. 8B and Fig. 9 is written.

[0082] Nonetheless, any other non-destructive inspection method, such as ultrasound or X-ray, will have its own sensitivity, resolution, accuracy, probability of detection, and false positive and false negative ratios, which are generally different from those of real-time process measurements. Thus, as with any other inspection method, a residual risk remains that cannot be eliminated without destroying the component and fully investigating its microstructure, mechanical properties, and defect distributions. Indeed, such periodic, fully destructive inspections of actual manufactured parts are required on an occasional and periodic basis.Such studies serve to further strengthen the validity of correlations between process data, real-time data, and part quality. However, these studies are very expensive and time-consuming and must therefore be kept to an absolute minimum. The specific design of such periodic collection of both non-destructive and destructive studies to continuously verify the validity of process product acceptance depends on the respective additive manufacturing application. For example, different requirements will apply to medicine and aerospace than to the automotive or energy generation sectors, and similarly to any other field in which additive manufacturing is used to produce functional structural components.

[0083] In the preferred embodiment described herein, there are other ways in which the present invention can be used to support additive manufacturing processes based on additive manufacturing. The following table lists different scenarios that regularly arise during production and, in particular, how the present invention addresses quality aspects and issues for each of these scenarios. scenario Solving quality problems Movement of machine tools from one physical position to another. Here, the quality issue relates to the requalification of a machine after it has been physically moved and / or partially disassembled to enable the movement. The present invention provides a specific platform-independent method for directly addressing such a quality concern: i) immediately before disassembly and movement, a new baseline data set is generated, which documents the machine status as well as the control section produced by the machine at position 1; ii) this baseline data set serves as the initial baseline data set for the requalification of the machine in physical position 2; iii) The methodology from Fig. 9 is applied to ensure that the machine in its new configuration is capable of producing samples that are consistent with the previously recorded baseline taken immediately before dismantling and moving the machine; iv) If this is the case, the Machine classified as requalified without any further effort required. machine to another, where both machines are of the same type and model, but physically separate machines with different histories and perhaps different maintenance statuses. Even though two machines are of the same type and model, they may have differences in conditions, maintenance, etc., which may result in different production results.Using the methodology described in the present invention, the following procedure can be used to determine whether two machines are identical: i) collect a new baseline dataset, or apply a pre-existing baseline dataset to machine 1; ii) collect a new baseline dataset, or apply a pre-existing baseline dataset to machine 2; iii) compare all aspects of the two baselines using a classification and / or outlier detection scheme as mentioned above; iv) if the baselines do not belong to the same population, adjust one or the other machine until the process data, confirmed by additional control section runs, are identical, or can be statistically grouped within the same population. Manufacturing processes that are transferred from one machine type to another. This is generally the most complicated transition, as highly diverse scanning strategies and local scanning parameters can occur for additive manufacturing machines. Values such as laser energy, laser spot size, laser scanning speed, and line spacing / overlap are insufficient to fully describe the differences between one machine and another. Thus, collecting process data, as described in the present invention, is critical in this case to minimize experimental repetitions and ensure that two parts manufactured on two different machines also have similar microstructure and properties.The method for achieving this using the described invention is identical to that described above for the transfer of parameters between two identical machines, but with the following modifications: i) the settings of the respective working parameters of the two machines will generally differ; ii) scanning strategies should be in an area where they are affected by. settings can be controlled, are set as identically as possible; iii) adjustment of the scanning strategies on the target machine may be necessary to obtain results similar to the source machine, and iv) all layer sampling iterations can be performed on the control section, assuming that the Eulerian and Lagrangian data can be obtained here on both machines.

[0084] There are logical extensions and generalizations of the embodiments described above, some of which will now be described. First, the above description of the embodiments involves the use of a control region for both development and manufacturing. If there is an alternative method to obtain the microstructure, mechanical properties, or defect distributions resulting from a particular set of process conditions, then this can also be used as a replacement for the corresponding control region; that is, the control region is a recommended, but not absolutely necessary, part of this invention. In some embodiments, multiple control regions are used accordingly.

[0085] It should also be understood that the examples and embodiments described herein are for illustrative purposes only and that various modifications and changes in light thereof will be suggested to those skilled in the art and are within the spirit and scope of this application and the scope of the appended claims.

[0086] The present invention comprises the following aspects: 1. Automated additive manufacturing device for producing a part on a powder bed, the automated manufacturing device comprising: a heat source; a processor; a scanning head configured to direct energy received from the heat source toward a layer of powder provided on the powder bed in a pattern specified by the processor and corresponding to a shape of the part; a first optical sensor configured to determine a temperature associated with a specified portion of the part; and a second optical sensor configured to receive light emitted through the scan head from a portion of the powder layer melted by the energy of the heat source; wherein the processor is configured to receive sensor data from the first and second optical sensors during an additive manufacturing process to characterize a quality of different portions of the part. 2. The automated manufacturing apparatus according to aspect 1, wherein the processor is configured to use temperature data collected by the first optical sensor to calibrate temperature data collected by the second optical sensor. 3. The automated manufacturing apparatus according to aspect 2, wherein the processor is configured to use the temperature data collected by the first and second optical sensors to calculate state variables used to assess the success of an additive manufacturing process. 4. Additive manufacturing process, comprising: Applying a layer of metallic material; Melting a region of the layer of metallic material to form a part manufactured by the additive manufacturing process with a heat source that scans over the region of the layer of metallic material to melt the region; Monitoring an amount of energy emitted by the heat source with a first optical sensor that follows a path along which the heat source scans the area to provide a first set of information; Monitoring a defined portion of the area of the layer of metallic material with a second optical sensor to provide a second set of information; and subsequent to melting the region of the layer of metallic material, determining whether the first and second sets of information indicate that the region falls within a known good range of a base data set associated with the part produced by the additive manufacturing process, by: correlating data contained in the second information set with data contained in the first information set, the correlated data being collected from the first and second information sets as the heat source passed through the specified portion of the area; Calculating a plurality of state variables using at least a portion of the first information set and at least a portion of the second information set; and Comparing the plurality of status variables with a plurality of ranges to which status variables of the known good range of the base data set are assigned. 5. Additive manufacturing method according to aspect 4, wherein the first optical sensor comprises a photodiode and the second optical sensor comprises a pyrometer. 6. Additive manufacturing process according to aspect 4, further comprising: Destructively analyzing the specified portion of the area observed by the second optical sensor to determine whether a microstructure of the specified portion matches the layer within the well-known region. 7. The additive manufacturing method of aspect 6, wherein the defined portion is distinct and different from another portion of the region used to form the part. 8. The additive manufacturing method of aspect 7, wherein the defined portion is removed from the other portion of the region to form a channel within the part. 9. Additive manufacturing process according to aspect 8, further comprising: Following melting of the area, repeating the process until the part is substantially complete. 10. Additive manufacturing process according to aspect 9, further comprising: Collecting the status variables related to each layer of the part and identifying which sections of the part are most likely to contain a defect based on the collected status variables. 11. The additive manufacturing method according to aspect 9, wherein determining whether the part falls within the known good data set is only possible if each portion of the part monitored by the second optical sensor falls within the known good range of the base data set. 12. Additive manufacturing method according to aspect 4, wherein the second optical sensor remains stationary during the execution of the additive manufacturing method. 13. The additive manufacturing method according to aspect 4, wherein the first set of information is calibrated with reference to the temperature rise and temperature fall information from the second set of information. 14. The additive manufacturing method according to aspect 4, wherein the heat source is a laser that shares optics with the first optical sensor. 15. Additive manufacturing process comprising: Collecting temperature data recorded by a plurality of sensors for a plurality of layers, which are each deposited during a plurality of additive manufacturing operations for producing a part, wherein a first portion of the plurality of additive manufacturing operations is performed using setpoint parameter ranges and a second portion of the additive manufacturing operations predominantly uses non-setpoint parameter ranges, wherein the non-setpoint parameter ranges are those ranges that are expected to produce undesirable material defects in the part; Performing metallurgical examinations at a specified location of each of the parts, wherein the specified location corresponds to a location on the part at which a field of view of a first optical sensor of the plurality of sensors remains fixed during each of the plurality of additive manufacturing operations and a field of view of a second optical sensor of the plurality of sensors periodically passes the location; Categorizing the sensor data collected from the plurality of sensors into target and non-target data ranges; and creating a baseline data set for the part that includes process limits for the sensor data that have been shown to result in the part having acceptable material properties. 16. The additive manufacturing method of aspect 15, wherein the setpoint parameter ranges comprise operating settings for an additive manufacturing system for producing the part that are expected to produce the part with acceptable material properties. 17. The additive manufacturing method of aspect 16, wherein the non-target parameter ranges comprise operating settings for an additive manufacturing system for producing the part that are expected to produce undesirable material defects in the part. 18. The additive manufacturing method according to aspect 15, wherein the categorization of the sensor data includes calculating a plurality of status variables from the sensor data. 19. The additive manufacturing method according to aspect 18, wherein the plurality of state variables are selected from a group comprising a scan peak temperature, a scan cooling rate, a scan heating rate, a bulk peak temperature, a bulk cooling rate, and a bulk heating rate. 20. The additive manufacturing method according to aspect 15, wherein the first optical sensor comprises a pyrometer. 21. Additive manufacturing process comprising: Applying a layer of metal powder; Melting a region of the metal powder layer to form a part produced by the additive manufacturing process with a heat source scanning over the region of the metal powder layer to melt the region; Monitoring an amount of energy emitted by the scanning heat source with a first optical sensor that follows a path along which the heat source scans the area to provide a first set of information; Monitoring a fixed portion of the area using a second optical sensor to provide a second set of information, wherein the fixed portion of the area is positioned such that the path is both within and outside a field of view of the second optical sensor; and after melting the region of the metal powder layer, determining whether the first and second sets of information indicate that the region falls within a known good region of a base data set associated with the part produced by the additive manufacturing process by: Comparing at least a portion of the first and second sets of information with regions associated with the known good region of the base data set to identify one or more portions of the part that may have a manufacturing defect; and in response to the first and second sets of information indicating that the range is outside the known good range of the base data set, adjusting the additive manufacturing process. 22. Additive manufacturing method according to aspect 21, wherein the region extends over the entire cross-sectional area of the part to be manufactured. 23. The additive manufacturing method of aspect 22, wherein the second sensor captures a second set of information that is correlated with the first set of information when identifying the one or more portions of the part that may contain a manufacturing defect. 24. The additive manufacturing method according to aspect 21, wherein the heat source is a laser having the same optics as the first optical sensor.

Claims

[1] An additive manufacturing system consisting of: a scanning head (409); a building level; a heat source (406) configured to transfer energy through the scan head (409) and toward the build plane; an optical sensor (412) configured to receive light through the scan head (409), the light being emitted from a portion of a layer of metal material positioned on the build plane; and a processor configured to execute computer code that causes the additive manufacturing system to perform an additive manufacturing operation to produce a part, the additive manufacturing operation comprising: Applying the layer of metal material on the build plane; Melting the portion of the layer of metal material using the heat source (406); Monitoring an amount of energy emitted by the heat source (406) using the optical sensor (412) to generate a data set; and Comparing the data set with a known good range of a base data set to determine whether one or more parts of the part may contain a manufacturing defect; wherein the optical sensor is a first optical sensor and wherein the additive manufacturing system comprises a second optical sensor (405; 413; 415) configured to receive light emitted from the portion of the layer of metal material, and wherein the second optical sensor (405) has a fixed field of view relative to the build plane, wherein the data set is a first data set and the known good region is a first known good region, and wherein the second optical sensor (405) generates a second data set that is compared to a second known good region to determine whether one or more portions of the part may have a manufacturing defect. [2] The additive manufacturing system of claim 1, wherein the heat source comprises a laser (406). [3] An additive manufacturing system consisting of: a scanning head (409); a building level; a heat source (406) configured to transfer energy through the scan head (409) and toward the build plane to create a molten pool (410); an optical sensor (412) configured to receive light through the scanning head (409), the light being emitted from the melt pool (410); and a processor configured to execute computer code that causes the additive manufacturing system to perform an additive manufacturing operation to produce a part, the additive manufacturing operation comprising: Applying a layer of metal powder to the build surface; Creating the molten pool (410) by melting a portion of the metal powder layer using the heat source (406); Generating a data set from an output of the optical sensor (412); and Comparing the data set to a known good region of a base data set to determine whether one or more portions of the part may contain a manufacturing defect; wherein the optical sensor is a first optical sensor and wherein the additive manufacturing system comprises a second optical sensor (405) configured to receive light emitted from the portion of the metal powder layer, wherein the second optical sensor (405) has a fixed field of view relative to the build plane, wherein the data set is a first data set and the known good region is a first known good region, and wherein the second optical sensor (405) generates a second data set that is compared to a second known good region to determine whether one or more portions of the part may have a manufacturing defect. [4] The additive manufacturing system of claim 3, wherein the heat source (406) comprises a laser. [5] An additive manufacturing system consisting of: a building level; a scanning head (409) arranged to traverse the build plane to build a part; a heat source (406) configured to transfer energy through the scan head (409) and toward the build plane to create a molten pool (410); an optical sensor (412) configured to generate a data set in response to receiving light via the scanning head (409), the light being emitted by the melt pool (410); and a processor configured to compare the data set with a known good range of a base data set to determine whether one or more parts of the part may have a manufacturing defect, wherein the processor changes a parameter of the heat source (406) in response to the comparison, wherein the optical sensor is a first optical sensor and wherein the additive manufacturing system comprises a second optical sensor (405) configured to receive light emitted from the melt pool (410), and wherein the second optical sensor (405) has a fixed field of view relative to the build plane, wherein the data set is a first data set and the known good region is a first known good region, and wherein the second optical sensor (405) generates a second data set that is compared to a second known good region to determine whether one or more portions of the part may have a manufacturing defect.

Images