# System and method for automated geometric shape deviation modeling for additive manufacturing

A method for providing a 3D shape model for a new process in an additive manufacturing system includes receiving an unoptimized 3D model of one or more shapes based on a first process (P1); receiving deviation information based on P1; baseline deviation modeling of P1 based on the received unoptimized 3D model and the received deviation information to thereby generate an optimized 3D model for the one or more shapes based on P1; applying a transfer function that connects deviation information of P1 to deviation information of a second process (P2); receiving deviation information based on P2; adaptively modeling the deviations of P2 based on the transformed baseline model and the deviation information of P2; generating an optimized 3D model for the one or more shapes based on P2; and generating a compensation plan for additively manufacturing the one or more shapes between P1 and P2.

## Latest Purdue Research Foundation Patents:

**Description**

**CROSS-REFERENCE TO RELATED APPLICATIONS**

The present patent application is related to and claims the priority benefit of U.S. Provisional Patent Application Ser. No. 62/671,190 filed May 14, 2018, the contents of which are hereby incorporated by reference in its entirety into the present disclosure.

**STATEMENT REGARDING GOVERNMENT FUNDING**

This invention was made with government support under CMMI 1544841 and CMMI 1544917 awarded by the National Science Foundation. The government has certain rights in the invention.

**TECHNICAL FIELD**

This specification relates to additive manufacturing (AM), and particularly to a system and method for predicting shape deviation error and compensating therefor when changing printing processes or shapes of an object to be manufactured.

**BACKGROUND**

This section introduces aspects that may help facilitate a better understanding of the disclosure. Accordingly, these statements are to be read in this light and are not to be understood as admissions about what is or is not prior art.

AM has gained a significant amount of interest in recent years. To manufacture a complex object, traditionally several methods have been employed, e.g., fixed tool manufacturing such as injection molding. However, as the tools age, shape deviations are introduced into the parts until such deviations are no longer acceptable. At such point, the tools will need to be repaired or replaced. With the advent of additive manufacturing and three-dimensional (3D) printing, objects can be manufactured layer-by-layer. As such, shape deviation can be accounted, as discussed in the U.S. Pub. App. No. 20160320771 for Huang. In the '771 publication Huang discloses a method of compensating for deviation to thereby minimize shape deviation during manufacturing by an AM machine. The method includes obtaining a deviation model for an AM machine, receiving a 3D model of an object to be built using the AM machine, predicting shape deviation for the object using the deviation model applied to the 3D model, minimizing the predicted deviation by selecting an amount of deviation compensation, and modifying the 3D model based on the selected amount of deviation compensation. However, Huang and others have not addressed changes and shifts in process resulting in a priori changes in deviation model and/or changes in shapes (e.g., from a circular cylinder shape to a pentagonal cylinder shape), or more intelligently based on real-time data.

Furthermore, the statistical model of shape deviations used by Huang to derive compensation and thereby modifications to the CAD model is rendered difficult by at least three reasons that result from the nature and capability of AM for one-of-a-kind manufacturing. First, objects that are to be manufactured using an AM process have myriad shapes and complex geometries. Second is the spectrum of AM processes and variations in those processes that can yield a broad range of complicated deviation patterns. Third, there typically exist a small collection of shape deviation data that can possibly be used to specify statistical deviation models for a large number of AM processes, and a significant amount of time is needed to develop these statistical models for new shapes.

Therefore, there is an unmet need for a novel approach to compensate for geometric shape deviation and minimize deviation during manufacturing by an AM machine that result from changes in process and/or shape by a priori or real-time adjustments to the 3D model of the object to be printed.

**SUMMARY**

A method performed by a computer system including a processor and a memory for providing a 3D shape model for a new process in an additive manufacturing (AM) system is disclosed. The method includes receiving an unoptimized 3D model of one or more shapes based on a first process. The method further includes receiving deviation information based on the first process. The deviation information of the first process defines AM deviations of the one or more shapes, both in-plane and out-of-plane, based on the first process. The method also includes baseline deviation modeling of the first process based on the received unoptimized 3D model and the received deviation information. Furthermore, the method includes generating an optimized 3D model for the one or more shapes based on the first process based on the initial optimizing step. The method then includes applying a transfer function that connects deviation information of the first process to deviation information of a second process to the optimized 3D model to thereby transform the baseline model for the second process. The process then includes receiving deviation information based on the second process. The deviation information of the second process defines AM deviations, both in-plane and out-of-plane, based on the second process. The method then includes adaptively modeling the deviations of the second process based on the transformed baseline model and the deviation information of the second process. Thereafter, the method includes generating an optimized 3D model for the one or more shapes based on the second process and based on the adaptive modeling of deviations step. The method then includes generating a compensation plan for additively manufacturing the one or more shapes between the first and the second processes.

Another method performed by a computer system including a processor and a memory for providing a 3D shape model for a new shape in an additive manufacturing system is disclosed. The method includes receiving an unoptimized 3D model for a first shape based on a process. The method further includes receiving deviation information based on the process, where the deviation information of the process defines AM deviations, both in-plane and out-of-plane, for the first shape based on the process. The method then includes baseline deviation modeling of the first process based on the received unoptimized 3D model and the received deviation information. The method also includes generating an optimized 3D model for the first shape based on the process based on the initial optimizing step. Furthermore, the method includes receiving an unoptimized 3D model of the second shape based on the process. In addition, the method includes applying a delta function to the baseline deviation model for the first shape, generated optimized 3D model for the first shape, observed deviation data for the second shape, and to the received unoptimized 3D model of the second shape. The delta function connects the baseline deviation model for the first shape to an unoptimized 3D model of a second shape based on the process. The method then includes adaptively modeling deviations of the second shape based on the applied delta function and the received unoptimized 3D model of the second shape. The method then includes generating an optimized 3D model for the second shape based on the adaptive modeling of deviations step. Thereafter, the method includes generating a compensation plan for additively manufacturing the second shape based on the process.

Yet another method performed by a computer system including a processor and a memory for providing a 3D shape model for a new process and a new shape in an additive manufacturing (AM) system is disclosed. The method includes receiving an unoptimized 3D model of a first shape based on a first process. The method further includes receiving deviation information based on the first process. The deviation information of the first process defines AM deviations of the first shape, both in-plane and out-of-plane, based on the first process. The method also includes baseline deviation modeling of the first process based on the received unoptimized 3D model and the received deviation information. Furthermore, the method includes generating an optimized 3D model for the first shape based on the first process based on the initial optimizing step. The method then includes applying a transfer function that connects deviation information of the first process to deviation information of a second process to the optimized 3D model to thereby transform the baseline model for the second process. The process then includes receiving deviation information based on the second process. The deviation information of the second process defines AM deviations, both in-plane and out-of-plane, based on the second process. The method then includes adaptively modeling the deviations of the second process based on the transformed baseline model and the deviation information of the second process. Thereafter, the method includes generating an optimized 3D model for the first shape based on the second process and based on the adaptively modeling deviations step. The method then includes receiving an unoptimized 3D model of a second shape based on the second process. Furthermore, the method includes applying a delta function to the generated optimized 3D model for the first shape under the second process, observed deviation data for the second shape under the second process, and to the received unoptimized 3D model of the second shape under the second process. The delta function connects the optimized 3D model of the first shape to an unoptimized 3D model of a second shape based on the second process. The method also includes second adaptively modeling deviations of the second shape based on the applied delta function and the received unoptimized 3D model of the second shape under the second process. The method then includes generating an optimized 3D model for the second shape based on the second process and based on the adaptively modeling deviations step, and generating a compensation plan for additively manufacturing the second shape based on the second process.

**BRIEF DESCRIPTION OF DRAWINGS**

*a ***1** and Shape **2**) for manufacture.

*b ***1** and Shape **2**) for manufacture.

*c **a ***1***b. *

*d *

*e *

*f **a ***1***b. *

*g **c ***1***f. *

*a *

*b *

*a*, 4*b*, 4*c*, 4*d*, 4*e*, 4*f*, 4*g*, and 4*h *

*a*, 8*b*, 8*c*, and 8*d *

*a*, 10*b*, 10*c*, and 10*d **a**b**c**a**d*

*a *and 11*b *

**DETAILED DESCRIPTION**

For the purposes of promoting an understanding of the principles of the present disclosure, reference will now be made to the embodiments illustrated in the drawings, and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of this disclosure is thereby intended.

In the present disclosure, the term “about” can allow for a degree of variability in a value or range, for example, within 10%, within 5%, or within 1% of a stated value or of a stated limit of a range.

In the present disclosure, the term “substantially” can allow for a degree of variability in a value or range, for example, within 90%, within 95%, or within 99% of a stated value or of a stated limit of a range.

In the present disclosure, processes A, B, and C refer to a baseline process of a stereolithography machine (Process A), a new setting of the stereolithography machine (Process B), and the deviation of cavities under a particular setting of the stereolithography machine (Process C), respectively.

A novel system and method to compensate for geometric shape deviation and to minimize deviation during manufacturing by an additive manufacturing (AM) machine resulting from changes in process and/or shape of an object is disclosed by a priori or real-time adjustments to a 3D model of the object to be printed.

Referring to *a***10** according to one embodiment of the present disclosure is provided. The system **10** includes two processes (Process A and Process B) and two shapes (Shape **1** and Shape **2**) for manufacture. A known model **12** is initially provided for the first shape (Shape **1**) under the first process (Process A). The flow in the system **10** includes learning a deviation model f_{1,B }for the first shape (Shape **1**) under the second process (Process B) using knowledge of the deviation model f_{1,A }for the first shape (Shape **1**) under the first process (Process A), as shown in Block **14**. Next the flow of the system **10** includes specifying a deviation model f_{2,B }for the second shape (Shape **2**) under the second process (Process B) using deviation knowledge (process and shape) of all of the models for the previously manufactured products, in an automated manner, as shown in Block **16**.

Referring to *b***20** according to another embodiment of the present disclosure is provided. The system **20** also includes two processes (Process A and Process B) and two shapes (Shape **1** and Shape **2**) for manufacture. A known model **22** is initially provided for the first shape (Shape **1**) under the first process (Process A). The flow in the system **20** includes learning a deviation model f_{2,A }for the second shape (Shape **2**) under the first process (Process A) using knowledge of the deviation model f_{1,A }for the first shape (Shape **1**) under the first process (Process A), as shown in Block **24**. Next the flow of the system **20** includes specifying a deviation model f_{2,B }for the second shape (Shape **2**) under the second process (Process B) using deviation knowledge (process and shape) of all of the models for the previously manufactured products, in an automated manner, as shown in the Block **26**.

The systems **10** and **20** shown in *a *and 1*b *

Referring to *c***100** according to the present disclosure capable of carrying out some of the steps shown in *a ***1***b*. In particular, *c ***12** and **14** in *a ***24** and **26** in *b**c ***100** shown in *c ***102**. This input file provides a conventional model for various shapes for a particular AM process. For example, a cylindrical shape may be presented as stacks of 2D circles formed (or, in the case of a 3D printer, printed) vertically to form the 3D cylinder. The input file in Block **102** then is a representation of a circle. The system **100** also provides an input file in Block **104** directed to the deviation model for the first process (Process A). This input file in Block **104** is generally constructed to capture deviation of the AM machine over a period of time. In particular, an AM machine may begin its life forming parts with some deviations. In addition, the AM machine thereafter may introduce additional deviations due to wear. The input file in Block **104** provides a model for such deviations (i.e., initial and time-dependent variations) from the desired shapes based on measurements of the parts formed by the AM machine (data based on measurements, not shown, are implicitly included in Blocks **102** and **104**). For example, with respect to various boundaries, the input file in Block **104** defines the manufacturing quality (i.e., deviation) in both in-plane and out-of-plane with respect to the demanded output for the first process (Process A). An example of this input file is one that contains computer code for the general statistical model that can be fit given the data by executing the code in a particular programming language (e.g., R, Python, etc.). Thus, exemplarily, the input file is computer code based on the shaped provided in *d***102** and **104**) are provided to a modeling block in Block **106** to specify the baseline deviation model of the first process (Process A). Using an optimization technique that includes validation of the resulting compensation, the baseline modeling block optimizes the 3D model for the first process (Process A). This change in the 3D model ensures that for the first process (Process A) printing an object defined by the optimized output 3D model will result in minimum deviation from the original object. This optimization is repeated for a class of shapes (e.g., cylinders, cubes, . . . ) as shown in Block **106**. The optimization process of Block **106** specifies a new model for each shape in the class that when formed by the AM machine will result in minimum deviation from the desired shape. An example of a compensated shape is provided in *e*_{new,A}=f_{A}(input,compensation), where f_{new,A }is the new compensated model of the original model (Block **102**) incorporating the deviation data (compensation data) of Block **104**.

The optimization process taking place in Block **106** is a recursive process, as shown in Blocks **108** and **110**. In Block **108**, a model deviation file is generated based on the input file of Block **104** that if applied to the input file of Block **102** will result in minimization of the deviations from the desired shapes of the input file of Block **102**. Application of this model deviation file (Block **108**) is tested against the input file of Block **102** according to a predetermined criterion in the decision block of Block **110**. The model validation data in block **108** captures the discrepancies between the observed manufactured product, and the computer-aided design model for the shape that is manufactured. The input file: deviation model in block **104** is a computer function, file, or object that describes the predictions of shape deviations for a product that is to be manufactured. The predetermined criterion can be based on an acceptable standard deviation, or other statistical parameters such as least mean squares (which is a statistical approach seeking to minimize the error between a desired shape and the shape in the recursive process), known to a person having ordinary skill in the art.

Once the baseline deviation model for one class of shapes under the first process (Process A) has been optimized, as signified by Block **112**, the system **100** of the present disclosure proceeds to the next block. In Block **114**, the system **100** uses a transfer function that connects a second process (Process B) to the first process (Process A). The system **100** also uses an input file directed to the deviation model for the second process (Process B), similar to the deviation model for the first process (Process A) shown in Block **104**, to develop a new model for the second process (Process B). The transfer function (or otherwise referred to as transfer compensation) is expressed as T=Orig Component(Δ_{B}−f_{A}(input,compensation))^{2}, wherein Δ_{B }is the observed deviation data for products manufactured under process B. The deviation model for process B is generated via this transfer function, as provided in Block **114**. With the transfer function of Block **114** and the input file for observed data for new process (Process B) of Block **116** as inputs, the system **100** then proceeds to an adaptive modeling of deviation shown in Block **118** that is similar to the modeling of the baseline deviation of Block **106**. Similarly, as discussed above, the optimization process taking place in Block **118** is a recursive process, as shown in Blocks **120** and **122**. In Block **120**, a model deviation file is generated based on the input file of Block **116** that if applied to the output of the transfer function will result in minimization of the deviations from the desired shapes of the output of the transfer function block **114**. Application of this model deviation file (Block **120**) is tested against the output of the transfer function **114** according to a predetermined criterion in the decision block of Block **122**. The model validation data in block **120** captures the discrepancies between the observed manufactured product, and the output of the transfer function block **114** for the shape that is manufactured. The observed data for the new process is a computer file that describes the shape deviation for a product that is to be manufactured under the new process. The predetermined criterion can be based on an acceptable standard deviation, or other statistical parameters such as least mean squares (which is a statistical approach seeking to minimize the error between a desired shape and the shape in the recursive process), known to a person having ordinary skill in the art. The system **100** then outputs an optimized 3D output model for the new process (Process B).

Once the adaptive deviation model for one class of shapes under the new process (Process B) has been optimized, as signified by Block **124**, the system **100** of the present disclosure proceeds to the next block (Block **126**), where a compensation plan is generated for going from the first process (Process A) to the second process (Process B) for a variety of shapes, accounting for variations in i) deviations due to changes in the AM machine as well as ii) changes for going from one process to another process. It should be noted that these shapes all belong to a common class, e.g., cylinders with different nominal radii.

Now referring to *f***200** for determining an optimized 3D model for a new shape is provided using the same process (Process X). The blocks in system **200** according to the present disclosure provide capability of carrying out some of the steps shown in *a *and 1*b***14** and **16** in *a ***22** and **24** in *b***200** is to calculate an optimized 3D model for a first shape (Shape **1**) given a deviation model for the first shape (Shape **1**) under the process (Process X) and an optimized 3D model for a second shape (Shape **2**) under the same process (Process X). Similar to the system **100** of *c***200** of FIG. if starts on the left side by receiving an unoptimized 3D model for a first shape (Shape **1**) for the process (Process X), as shown in Block **202**. This input file provides a conventional model for the first shape (Shape **1**) for a particular AM process (Process X). For example, a cylindrical shape may be presented as stacks of 2D circles formed (or, in the case of a 3D printer, printed) vertically to form the cylinder. The input file in Block **202** then is a representation of a circle. The system **200** also provides an input file in Block **204** directed to the deviation model for the first shape (Shape **1**) for the process (Process X). This input file in Block **204** is generally constructed to capture deviation of the AM machine over a period of time. In particular, an AM machine may begin its life forming parts with some deviations. In addition, the AM machine thereafter may introduce additional deviations due to wear. The input file in Block **204** provides a model for such deviations (i.e., initial and time-dependent variations) from the desired shape (Shape **1**) based on measurements of the parts formed by the AM machine (data based on measurements, not shown, is implicitly included in Blocks **202** and **204**). For example, with respect to various boundaries, the input file in Block **204** defines the manufacturing quality (i.e., deviation), both in-plane and out-of-plane, with respect to the demanded output for the process (Process X) for the first shape (Shape **1**). An example of this input file is one that contains computer code for the general statistical model that can be fit given the data by executing the code in a particular programming language (e.g., R, Python, etc.). Thus, exemplarily, the input file is computer code based on the shapes provided in *d***202** and **204**) are provided to a modeling block in Block **206** to specify the baseline deviation model of the first shape (Shape **1**). Using an optimization technique that includes validation of the resulting compensation, the baseline modeling block optimizes the 3D model for the first shape (Shape **1**). This change in the 3D model ensures that for the first shape (Shape **1**), manufacturing an object defined by the optimized output 3D model will result in minimum deviation from the original object. The optimization process of Block **206** specifies a new model for the first shape (Shape **1**) that, when formed by the AM machine, will result in minimum deviation from the desired shape (Shape **1**). An example of a compensated shape is provided in *e***1** can be expressed as f_{new,A}=f_{A}(input, compensation), where f_{new,A }is the new compensated model of the original model (Block **202**) incorporating the deviation data (compensation data) of Block **204**.

The optimization process taking place in Block **206** is a recursive process, as shown in Blocks **208** and **210**. In Block **208**, a model deviation file is generated based on the input file of Block **204** that, if applied to the input file of Block **202**, will result in minimization of the deviations from the desired shapes of the input file of Block **202**. Application of this model deviation file (Block **208**) is tested against the input file of Block **202** according to a predetermined criterion in the decision block of Block **210**. The model validation data in Block **208** captures the discrepancies between the observed, manufactured product, and the computer-aided design model for the shape that is to be manufactured. The input file: deviation model in block **204** is a computer function, file, or object that describes the predictions of shape deviations for a product that is to be manufactured. The predetermined criterion can be based on an acceptable standard deviation, or other statistical parameters such as least mean squares (which as discussed above is a statistical method seeking to minimize the error between a desired shape and the shape in the recursive process), known to a person having ordinary skill in the art.

Once the baseline deviation model for the first shape (Shape **1**) under the process (Process X) has been optimized—thus, implicitly including the baseline deviation model, as signified by Block **212**, the system **200** of the present disclosure proceeds to the next block. In Block **214**, the system **200** uses an unoptimized 3D model of a second shape (Shape **2**) for the process (Process X). This input file, similar to the input file of Block **202**, provides a conventional model for the second shape (Shape **2**) for the process (Process X). For example, a cube may be presented as stacks of 2D squares formed (or, in case of a 3D printer, printed) vertically to form the cube. The input file in Block **214** then is a representation of a square. Utilizing the input file of Block **214**, the system **200** then determines a difference model between the first shape (Shape **1**), e.g., a circle, and the second shape (Shape **2**), e.g., a square, as shown in Block **216** for the process (Process X). This model (output of Block **216**) and the output of Block **212** (optimized model for the first shape (Shape **1**)) are then fed to an adaptive modeling block of Block **218** in order to generate the model for the second shape (Shape **2**) as the system **200** proceeds to Block **218** which is similar to the modeling of the baseline deviation of Block **206**. Similarly, as discussed above, the optimization process taking place in Block **218** is a recursive process, as shown in Blocks **220** and **222**. In Block **220**, a model deviation file is generated based on the input file of Block **214** that if applied to the changed input file of Block **212** (i.e., the input files of Blocks **212** and **214** changed by the difference model of Block **216**) will result in minimization of the deviations from the desired shapes of the input file of Block **214**. Block **216** captures the difference between the observed deviation data for the new shape (Shape **2**) and the predictions for the first shape (Shape **1**); and this difference is referred to herein as the delta function. It should be noted that block **220** captures the difference between the predictions for the new shape (Shape **2**) and the observed data for the new shape (Shape **2**). Application of this model deviation file (Block **220**) is tested against the changed version of input files of Blocks **212** and **214** according to a predetermined criterion in the decision block of Block **222**. The predetermined criterion can be based on an acceptable standard deviation, or other statistical parameters such as least mean squares (which as discussed above is a statistical method seeking to minimize the error between a desired shape and the shape in the recursive process), known to a person having ordinary skill in the art. The new model then outputs an optimized 3D output model for the second shape (Shape **2**) under the process (Process X).

Once the adaptive deviation model for the second shape (Shape **2**) under the process (Process X) has been optimized, as signified by Block **224**, the system **200** of the present disclosure proceeds to the next block (Block **226**), where a compensation plan is generated for going from the first shape (Shape **1**) to the second shape (Shape **2**) for the process (Process X), accounting for variations in i) deviations due to changes in the AM machine as well as ii) changes for going from one shape to another shape.

In both *c *and 1*f ***100** and **200** for new process and new shape, respectively), the output includes an optimized 3D output for the new process, an optimized 3D output for the new shape, and the corresponding compensation plan that will yield minimum deviation in either case.

A combination of *c *and 1*f**a ***124** (see *c***216** and **218** as an optimized 3D output for a new process (Process B) to be used for a new Shape (Shape **2**). The remainder of *f *

As discussed above, the optimization techniques described above include a variety of different techniques known to a person having ordinary skill in the art. Examples include least means squares and a class of Bayesian extreme learning machines (ELMs). According to one embodiment, the methodology described in the present disclosure utilizes point cloud measurement data collected from a small sample of test shapes without requiring specific knowledge of AM processes. The optimization method described herein advantageously meet the stated goal of automated specification of both in-plane and out-of-plane deviation models for different shapes under a specific process and/or same shapes under multiple processes, and to do so with increased simplicity. Referring to *g***1**, the baseline deviation model f_{s,p }for a class of shapes s manufactured under an AM process p is defined as

*y*_{s,p}*=x*_{s,p}*+H*_{s,p}β_{s,p}+ϵ_{s,p }

where y_{s,p }contains the observed deviations, x_{s,p }contains the compensations for each point, β_{s,p }contains the parameters to be estimated, ϵ_{s,p }contains the random error (i.e., random variables representing high-frequency deviation components), and H_{s,p }is a function of the inputs defined as the N_{s,p′}×M_{s,p′→p }matrix whose

(*i,m*) entry is *g*(α_{0,m}^{(s,p′)}*+z*_{i}(θ_{i}*,r*_{l}_{i}^{nom}(⋅),*x*_{l}_{i}(θ_{i}))^{T}α_{m}^{(s,p′)})

for a certain activation function g.

In step **2** of *g*

*{circumflex over (T)}*_{s,p′→p}*=H*_{s,p′→p}β_{s,p′→p}+ϵ_{s,p′→p},

where the deviation model for the new process is expressed as

*y*_{s,p′}*=x*_{s,p′}*+H*_{s,p′→p}{circumflex over (β)}_{s,p′→p}*+H*_{s,p′}β_{s,p}+ϵ_{s,p′}.

Next in Step **3** of *g*

*y*_{s′,p}*=x*_{s′,p}*+H*_{s′,p,G}β_{s,p}*+H*_{s′,p,L}β_{s′,p}*+ϵs′,p *

with the local deviation feature denoted by δ_{s′}=H_{s′,p,L}β_{s′,p}.

Finally, in Step **4** of *g*

*y*_{s′,p′}*=x*_{s′,p′}*+H*_{s,p′→p,G}{circumflex over (β)}_{s,p′→p}*+H*_{s,p′,G}β_{s,p}*+H*_{s′,p′,L}β_{s′,p′}+ϵ_{s′,p′}.

To demonstrate this approach, a case study is presented herein based on modeling in-plane deviations for an irregular polygon shown in *a **b *

In the present disclosure, in-plane deviations refer to 2D deviations of a 3D shape with a negligible vertical height. For such a shape, the top and bottom surface deviations are about the same. A polar coordinates representation of the in-plane deviation is shown in

The methodology used in the present disclosure is based on in-plane deviations of shapes with negligible heights, and out-of-plane deviations of shapes with negligible widths. In both cases, each point on a product is identified by an angle θ. The nominal radius function for a manufactured product 1 is the function r_{l}^{nom}:[0,2π]→ that defines a computer model. For a point θ on a product (denoted as 1), the radius is denoted by r_{l}^{obs}(θ), and its deviation is defined as Δ_{l}(θ)=r_{l}^{obs}(θ). This representation enables consistent in-plane and out-of-plane deviation modeling. The compensation plan x_{l}(⋅) for a product 1, according to one embodiment of the present disclosure is a function x_{l}:[0;2π]→ that specifies the addition and subtraction of material based on original computer model. The observed radius for point θ on product 1 under compensation plan x_{l}(⋅) is denoted by r_{l}^{obs}(θ,x_{l}(θ)), and the corresponding deviation is Δ_{l}(θ,x_{l}(θ))=r_{l}^{obs}(θ,x_{l}(θ))−r_{l}^{nom}(θ).

In order to show the applicability of the system and method according to the present disclosure, case studies of a Bayesian ELM deviation modeling methodology for different shapes and AM processes are now provided. Initially, implementation of the first step for the baseline shape of cylinders with small heights manufactured under a single setting of a stereolithography machine, as discussed with respect to *c**a*-4*h*

In the first case study, we specify a baseline model for in-plane cylinder deviations using four uncompensated cylinders (shape class **1**) of nominal radii 0.5″, 1″, 2″, and 3″ manufactured under the same stereolithography process (Process A). We have about 4000 points on these products, with the independent variables for each point i defined as

*z*_{i}(θ_{i}*,r*_{l}_{i}^{nom}(⋅))=(θ_{i}*,r*_{l}_{i}^{nom}(θ_{i}))^{T}.

We proceed to fit the baseline Bayesian ELM deviation model. The posterior predictions for deviations from our model are summarized in

Next, in order to model in-plane cylinder deviations under a different setting introduced by a change in the stereolithography calibration, the previously specified baseline deviation model is transferred to cylinders manufactured under a new process (Process B) of the stereolithography machine that introduces overcompensation. The distinct and complicated features of this new process are immediate upon inspection of the deviations in

Next, in-plane deviations of circular cavities are examined. The deviations of cavities in additively manufactured products are effectively generated under a new type of AM process C, thus the second step of our methodology can enable deviation modeling for such components of a product. We illustrate this using three circular cavities of nominal radii 0.5″, 1.0″, and 1.5″ contained in cylinders of nominal radii 1″, 2″, and 3″, respectively. We have about 3000 points on these products. The posterior predictions obtained for these products, summarized in

Next out-of-plane deviations of vertical semi-cylinders are studied. Out-of-plane deviations are also generated under a new type of AM process D in comparison to in-plane deviations provided in the prior art. Specifying out-of-plane deviation models is of great importance for AM and is especially challenging due to the complications introduced by interlayer bonding effects that exist in the vertical direction of a 3D product. Once again, we address this challenge in deviation modeling by means of the second step of our methodology. To illustrate, we consider four vertical semi-cylinders of nominal radii 0.5″, 0.8″, 1.5″, and 2″. We have about 4000 measurements on the product. *a*, 8*b*, 8*c*, and 8*d *

Next, results from transferring the baseline cylinder model to new shapes and processes are discussed. The in-plane deviations of polygons under a process are first examined. First, results from in-plane deviations (dots) for three squares, and the posterior predictive means (solid lines) and 95% central posterior predictive intervals (dashed lines) obtained by our transfer of the cylinder deviation model to these polygons are presented in *a**a*, 10*b*, and 10*c *

Next, we look at a hexagonal cavity of size 1.8″ that is contained in a cylinder of small height and nominal radius 3″. This constitutes a product from a shape class **2** manufactured under the process C. Just as before, our method yields a deviation model with good predictive performance for this shape as shown in *d. *

The last case study includes two free-form shapes of small height manufactured under process B. Modeling these deviations is complicated by the geometric complexity of the shapes. Our connectable ELM structures have a broad scope of application and can immediately accommodate both of these free-form shapes. *a *and 11*b **a *and 11*b *

Referring to **1000** for analyzing data and performing other analyses described herein, and related components. The system includes a processor **1086**, an AM printer **1020**, a user interface system **1030**, and a data storage system **1040**. AM printer **1020**, the user interface system **1030** and the data storage system **1040** are communicatively connected to the processor **1086**. Processor **1086** can be communicatively connected to network **1050** (shown in phantom), e.g., the Internet or a leased line, as discussed below. The imaging described in the present disclosure may be obtained using imaging sensors **1021** (e.g., 3D scanners) and/or displayed using display units (included in user interface system **1030**) which can each include one or more of systems **1086**, **1030**, **1040**, and can each connect to one or more network(s) **1050**. The digital scanner of the imaging sensor **1021** can be employed to capture visual representations of a physical object. The 3D scanner of the imaging sensor **1021** can include cameras or sensors employed to generate dense point clouds or polygon meshes used in 3D modeling. Additionally, it can be implemented using technology including but not limited to: 3D digitizers; laser scanners; structured light scanners; laser pulsed scanners; white light scanners; (Light Detection And Ranging) LIDAR; digital cameras; and structured light 3D cameras. As an example, the system **1000** can be employed to perform 3D printing followed by 3D scanning. Processor **1086**, and other processing devices described herein, can each include one or more microprocessors, microcontrollers, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), programmable logic devices (PLDs), programmable logic arrays (PLAs), programmable array logic devices (PALs), or digital signal processors (DSPs).

Processor **1086** can implement processes of various aspects described herein. Processor **1086** can be or include one or more device(s) for automatically operating on data, e.g., a central processing unit (CPU), microcontroller (MCU), desktop computer, laptop computer, mainframe computer, personal digital assistant, digital camera, cellular phone, smartphone, or any other device for processing data, managing data, or handling data, whether implemented with electrical, magnetic, optical, biological components, or otherwise. Processor **1086** can include Harvard-architecture components, modified-Harvard-architecture components, or Von-Neumann-architecture components.

The phrase “communicatively connected” includes any type of connection, wired or wireless, for communicating data between devices or processors. These devices or processors can be located in physical proximity or not. For example, subsystems such as the AM printer **1020**, user interface system **1030**, and data storage system **1040** are shown separately from the data processing system **1086** but can be stored completely or partially within the data processing system **1086**.

The AM printer **1020** can include one or more devices configured to provide digital content records to the processor **1086**. The processor **1086**, upon processing of digital content can provide the same to the AM printer **1020**, to generate an output shape.

The user interface system **1030** can include a mouse, a keyboard, another computer (connected, e.g., via a network or a null-modem cable), or any device or combination of devices from which data is input to the processor **1086**. The user interface system **1030** also can include a display device, a processor-accessible memory, or any device or combination of devices to which data is output by the processor **1086**. The user interface system **1030** and the data storage system **1040** can share a processor-accessible memory.

In various aspects, processor **1086** includes or is connected to communication interface **1015** that is coupled via network link **1016** (shown in phantom) to network **1050**. For example, communication interface **1015** can include an integrated services digital network (ISDN) terminal adapter or a modem to communicate data via a telephone line; a network interface to communicate data via a local-area network (LAN), e.g., an Ethernet LAN, or wide-area network (WAN); or a radio to communicate data via a wireless link, e.g., WiFi or GSM. Communication interface **1015** sends and receives electrical, electromagnetic or optical signals that carry digital or analog data streams representing various types of information across network link **1016** to network **1050**. Network link **1016** can be connected to network **1050** via a switch, gateway, hub, router, or other networking device.

Processor **1086** can send messages and receive data, including program code, through network **1050**, network link **1016** and communication interface **1015**. For example, a server can store requested code for an application program (e.g., a JAVA applet) on a tangible non-volatile computer-readable storage medium to which it is connected. The server can retrieve the code from the medium and transmit it through network **1050** to communication interface **1015**. The received code can be executed by processor **1086** as it is received, or stored in data storage system **1040** for later execution.

Data storage system **1040** can include or be communicatively connected with one or more processor-accessible memories configured to store information. The memories can be, e.g., within a chassis or as parts of a distributed system. The phrase “processor-accessible memory” is intended to include any data storage device to or from which processor **1086** can transfer data, whether volatile or nonvolatile; removable or fixed; electronic, magnetic, optical, chemical, mechanical, or otherwise. Exemplary processor-accessible memories include but are not limited to: registers, floppy disks, hard disks, tapes, bar codes, Compact Discs, DVDs, read-only memories (ROM), erasable programmable read-only memories (EPROM, EEPROM, or Flash), and random-access memories (RAMs). One of the processor-accessible memories in the data storage system **1040** can be a tangible non-transitory computer-readable storage medium, i.e., a non-transitory device or article of manufacture that participates in storing instructions that can be provided to processor **1086** for execution.

In an example, data storage system **1040** includes code memory **1041**, e.g., a RAM, and disk **1043**, e.g., a tangible computer-readable rotational storage device such as a hard drive. Computer program instructions are read into code memory **1041** from disk **1043**. Processor **1086** then executes one or more sequences of the computer program instructions loaded into code memory **1041**, as a result performing process steps described herein. In this way, processor **1086** carries out a computer implemented process. For example, steps of methods described herein, blocks of the flowchart illustrations or block diagrams herein, and combinations of those, can be implemented by computer program instructions. Code memory **1041** can also store data, or can store only code.

Various aspects described herein may be embodied as systems or methods. Accordingly, various aspects herein may take the form of an entirely hardware aspect, an entirely software aspect (including firmware, resident software, micro-code, etc.), or an aspect combining software and hardware aspects These aspects can all generally be referred to herein as a “service,” “circuit,” “circuitry,” “module,” or “system.”

Furthermore, various aspects herein may be embodied as computer program products including computer readable program code stored on a tangible non-transitory computer readable medium. Such a medium can be manufactured as is conventional for such articles, e.g., by pressing a CD-ROM. The program code includes computer program instructions that can be loaded into processor **1086** (and possibly also other processors), to cause functions, acts, or operational steps of various aspects herein to be performed by the processor **1086** (or other processor). Computer program code for carrying out operations for various aspects described herein may be written in any combination of one or more programming language(s), and can be loaded from disk **1043** into code memory **1041** for execution. The program code may execute, e.g., entirely on processor **1086**, partly on processor **1086** and partly on a remote computer connected to network **1050**, or entirely on the remote computer.

Those having ordinary skill in the art will recognize that numerous modifications can be made to the specific implementations described above. The implementations should not be limited to the particular limitations described. Other implementations may be possible.

## Claims

1. A method performed by a computer system including a processor and a memory for providing a 3D shape model for a new process in an additive manufacturing (AM) system, comprising:

- receiving an unoptimized 3D model of one or more shapes based on a first process;

- receiving deviation information based on the first process, the deviation information of the first process defines AM deviations of the one or more shapes, both in-plane and out-of-plane, based on the first process;

- baseline deviation modeling of the first process based on the received unoptimized 3D model and the received deviation information and performing an initial optimization to minimize deviation under the first process for the one or more shapes, wherein the initial optimization step of the baseline deviation modeling of the first process is automatically performed using a Bayesian extreme learning machine architecture;

- generating an optimized 3D model for the one or more shapes based on the first process based on the initial optimization;

- applying a transfer function that connects deviation data of the first process including the baseline deviation model, the unoptimized 3D model, and the AM deviations for the one or more shapes to deviation information of a second process to thereby generate a baseline model for the second process for the one or more shapes;

- receiving deviation information based on the second process, the deviation information of the second process defines AM deviations, both in-plane and out-of-plane, based on the second process;

- adaptively modeling the deviations of the second process based on the generated baseline model and the deviation information of the second process and performing a secondary optimization to minimize deviation under the second process for the one or more shapes, wherein the secondary optimization step of the adaptive deviation model of the second process is automatically performed using a Bayesian extreme learning machine architecture;

- generating an optimized 3D model for the one or more shapes based on the second process and on the adaptive deviation modeling step; and

- generating a compensation plan for additively manufacturing the one or more shapes between the first and the second processes.

2. The method of claim 1, the baseline deviation modeling includes: initial optimizing of an initial deviation model that can be applied to the received unoptimized 3D model based on the received deviation information for the first process to generate minimum deviation when additively manufacturing the one or more shapes based on the first process.

3. The method of claim 2, the optimization of the baseline deviation modeling is based on a least mean squares deviation minimization.

4. The method of claim 1, the adaptive deviation modeling includes:

- secondary optimizing of a deviation model that can be applied to the generated baseline model for the second process based on the received deviation information for the second process to generate minimum deviation when additively manufacturing the one or more shapes based on the second process.

5. The method of claim 4, the optimization of the adaptive modeling of deviation is based on a Bayesian extreme learning machine architecture.

6. The method of claim 4, the optimization of the adaptive modeling of deviation is based on a least mean squares deviation minimization.

7. A method performed by a computer system including a processor and a memory for providing a 3D shape model for a new shape in an additive manufacturing (AM) system, comprising:

- receiving an unoptimized 3D model of a first shape based on a process;

- receiving deviation information based on the process, the deviation information of the process defines AM deviations, both in-plane and out-of-plane, for the first shape based on the process;

- baseline deviation modeling of the first process based on the received unoptimized 3D model and the received deviation information and performing an initial optimization to minimize deviation under the process for the first shape, wherein the initial optimization step of the baseline deviation modeling of the process is automatically performed using a Bayesian extreme learning machine architecture;

- generating an optimized 3D model for the first shape based on the process based on the initial optimization;

- receiving an unoptimized 3D model of a second shape based on the process;

- applying a delta function to the generated optimized 3D model for the first shape, observed deviation data for the second shape, and to the received unoptimized 3D model of the second shape, the delta function connects the optimized 3D model of the first shape to the unoptimized 3D model of the second shape based on the process, the delta function connects deviation data of the first shape including the baseline deviation model, the unoptimized and optimized 3D models of the first shape, and the AM deviations for the first shape to the unoptimized 3D model of the second shape and the AM deviations of the second shape based on the process;

- adaptively modeling deviations of the second shape based on the applied delta function and the received unoptimized 3D model of the second shape and performing a secondary optimization to minimize deviation under the process for the second shape, wherein the secondary optimization step of the adaptive deviation model under the process is automatically performed using a Bayesian extreme learning machine architecture;

- generating an optimized 3D model for the second shape based on the process and on the adaptive deviation modeling step; and

- generating a compensation plan for additively manufacturing the second shape based on the process.

8. The method of claim 7, the baseline deviation modeling includes: initial optimizing of an initial deviation model that can be applied to the received unoptimized 3D model based on the received deviation information for the first shape based on the process to generate minimum deviation when additively manufacturing the first shape based on the process.

9. The method of claim 8, the optimization of the baseline deviation modeling is based on a least mean squares deviation minimization.

10. The method of claim 7, the adaptive deviation modeling step includes:

- secondary optimizing of a deviation model that can be applied to the output of the delta function to generate minimum deviation when additively manufacturing the second shape based on the process.

11. The method of claim 10, the optimization of the adaptive modeling of deviation is based on a Bayesian extreme learning machine architecture.

12. The method of claim 10, the optimization of the adaptive modeling of deviation is based on a least mean squares deviation minimization.

13. A method performed by a computer system including a processor and a memory for providing a 3D shape model for a new process and a new shape in an additive manufacturing (AM) system, comprising:

- receiving an unoptimized 3D model of a first shape based on a first process;

- receiving deviation information based on the first process, the deviation information of the first process defines AM deviations of the first shape, both in-plane and out-of-plane, based on the first process;

- baseline deviation modeling of the first process based on the received unoptimized 3D model and the received deviation information and performing an initial optimization to minimize deviation under the first process for the first shape, wherein the initial optimization step of the baseline deviation modeling of the first process is automatically performed using a Bayesian extreme learning machine architecture;

- generating an optimized 3D model for the first shape based on the first process based on the initial optimization;

- applying a transfer function that connects deviation data of the first process including the baseline deviation model, the unoptimized 3D model, and the AM deviations for the first shape to deviation information of a second process to thereby generate a baseline model for the second process for the first shape;

- receiving deviation information based on the second process, the deviation information of the second process defines AM deviations, both in-plane and out-of-plane for the first shape, based on the second process;

- firstly adaptively modeling the deviations of the second process based on the generated baseline model and the deviation information of the second process and performing a secondary optimization to minimize deviation under the second process for the first shape, wherein the secondary optimization step of the adaptive deviation model of the second process is automatically performed using a Bayesian extreme learning machine architecture;

- generating an optimized 3D model for the first shape based on the second process and on the adaptive deviation modeling step;

- receiving an unoptimized 3D model of a second shape based on the second process;

- applying a delta function to the generated optimized 3D model for the first shape under the second process, observed deviation data for the second shape under the second process, and to the received unoptimized 3D model of the second shape under the second process, the delta function connects the optimized 3D model of the first shape to the unoptimized 3D model of the second shape based on the second process, the delta function connects deviation data of the first shape based on the second process including the baseline deviation model, the unoptimized and optimized 3D models of the first shape, and the AM deviations for the first shape to the unoptimized 3D model of the second shape and the AM deviations of the second shape based on the second process;

- secondly adaptively modeling deviations of the second shape based on the applied delta function and the received unoptimized 3D model of the second shape under the second process and performing a third optimization to minimize deviation under the second process for the second shape, wherein the third optimization step of the adaptive deviation model under the second process is automatically performed using a Bayesian extreme learning machine architecture;

- generating an optimized 3D model for the second shape based on the second process and on the adaptive deviation modeling step; and

- generating a compensation plan for additively manufacturing the second shape based on the second process.

14. The method of claim 13, the baseline deviation modeling includes:

- initial optimizing of an initial deviation model that can be applied to the received unoptimized 3D model based on the received deviation information for the first process to generate minimum deviation when additively manufacturing the first shape based on the first process.

15. The method of claim 13, the first adaptive modeling deviations includes:

- optimizing of a deviation model that can be applied to the transformed baseline model for the second process based on the received deviation information for the second process to generate minimum deviation when additively manufacturing the first shape based on the second process.

16. The method of claim 15, the optimizations of the first adaptive modeling of deviations is based on a Bayesian extreme learning machine architecture.

17. The method of claim 13, the second adaptive modeling deviations include:

- optimizing of a deviation model that can be applied to the output of the delta function to generate minimum deviation when additively manufacturing the second shape based on the second process.

18. The method of claim 17, the optimizations of the second adaptive modeling of deviations is based on a Bayesian extreme learning machine architecture.

**Referenced Cited**

**U.S. Patent Documents**

9827717 | November 28, 2017 | Huang et al. |

9886526 | February 6, 2018 | Huang |

20160046076 | February 18, 2016 | Huang |

20160299996 | October 13, 2016 | Huang |

20160320771 | November 3, 2016 | Huang |

20170368753 | December 28, 2017 | Yang |

20180275634 | September 27, 2018 | Mcfarland |

20180284735 | October 4, 2018 | Cella |

20190004496 | January 3, 2019 | Blom |

**Foreign Patent Documents**

104537033 | April 2015 | CN |

105160059 | December 2015 | CN |

**Other references**

- Huang, et. al., “Predictive modeling of geometric deviations of 3D printed products Products—A Unified Modeling Approach for Cylindrical and Polygon Shapes”, 2014 IEEE International Conference on Automation Science and Engineering (CASE), Taipei, Taiwan, Aug. 18-22, 2014 (Year: 2014).
- Sabbaghi, te. al., “Predictive Model Building Across Different Process Conditions and Shapes in 3D Printing”, 2016 IEEE International Conference on Automation Science and Engineering (CASE), Fort Worth, TX, USA, Aug. 21-24, 2016 (Year: 2016).
- Sabbaghi, te. al., “Bayesian additive modeling for Quality Control of 3D Printed Products”, 2015 IEEE International Conference on Automation Science and Engineering (CASE), Gothenburg, Sweden, Aug. 24-28, 2015 (Year: 2015).
- Soria-Olivas et. al., “BELM: Bayesian extreme learning machine” IEEE Transactions on Neural Networks, vol. 22, No. 3, Mar. 2011, p. 505-509 (Year: 2011).
- Cheng, et al., “A Statistical Transfer Learning Perspective for Modeling Shape Deviations in Additive Manufacturing,” Oct. 2017, IEEE Robotics and Automation Letters, vol. 2, No. 4.
- Baumann et al., “Abstract API for 3D Printing Hardware and Software Resources,” 2017, Int J Adv Manuf Technol, 92:1519-1535.
- Cheng et al., “A Prediction and Compensation Scheme for Inplane Shape Deviation of Additive Manufacturing With Information on Process Parameters,” Feb. 8, 2018, IISE Transactions, vol. 50, No. 5, 394-406.
- Chowdhury, “Artificial Neural Network Based Geometric Compensation for Thermal Deformation in Additive Manufacturing Processes,” Nov. 2, 2016, Master's Thesis.
- Chowdhury et al., “Part Build Orientation Optimization and Neural Network-Based Geometry Compensation for Additive Manufacturing Process,” Mar. 2018, Journal of Manufacturing Science and Engineering, vol. 140, 031009:1-15.
- Chowdhury et al., “Artificial Neural Network Based Geometric Compensation for Thermal Deformation in Additive Manufacturing Processes,” Jun. 27, 2016, Proceedings of the ASME 2016 International Manufacturing Science and Engineering Conference.
- Huang , “An Analytical Foundation for Optimal Compensation of Three-Dimensional Shape Deformation in Additive Manufacturing,” Jun. 2016, Journal of Manufacturing Science and Engineering, vol. 138, 061010: 1-8.
- Huang et al., “Statistical Predictive Modeling and Compensation of Geometric Deviations of Three-Dimensional Printed Products,” Dec. 2014, Journal of Manufacturing Science and Engineering, vol. 136, 061008: 1-10.
- Huang et al., “Predictive Modeling of Geometric Deviations of 3D Printed Products—A Unified Modeling Approach for Cylindrical and Polygon Shapes,” 2014, IEEE International Conference on Automation Science and Engineering.
- Huang et al., “Optimal Offline Compensation of Shape Shrinkage for Three-Dimensional Printing Processes,” 2015, IIE Transactions vol. 47, 431-441.
- Jin et al., “Prescriptive Analytics for Understanding of Out-of-Plane Deformation in Additive Manufacturing,” 2016.
- Jin, et al., “Offline Predictive Control of Out-of-Plane Shape Deformation for Additive Manufacturing,” Dec. 2016, Journal of Manufacturing Science and Engineering, vol. 138, 121005: 1-7.
- Luan et al., “Predictive Modeling of In-Plane Geometric Deviation for 3D Printed Freeform Products,” 2015.
- Luan et al., “Prescriptive Modeling and Compensation of In-Plane Shape Deformation for 3-D Printed Freeform Products,” Jan. 2017, IEEE Transactions on Automation Science and Engineering, vol. 14, No. 1.
- Moroni et al., “Towards Early Estimation of Part Accuracy in Additive Manufacturing,” 2014, Procedia CIRP, 21, 300-305.
- Navangul et al., “Error Minimization in Layered Manufacturing Parts by Stereolithography File Modification Using a Vertex Translation Algorithm,” Jun. 2013, Journal of Manufacturing Science and Engineering, vol. 135, 031006: 1-13.
- Sabbaghi et al., “Predictive Model Building Across Different Process Conditions and Shapes in 3D Printing,” Aug. 21, 2016, 2016 IEEE International Conference on Automation Science and Engineering.
- Sabbaghi et al., “Inference for Deformation and Interference in 3D Printing,” 2014, The Annals of Applied Statistics, vol. 8, No. 3, 1395-1415.
- Sabbaghi et al., “Bayesian Additive Modeling for Quality Control of 3D Printed Products,” Aug. 24, 2015, 2015 IEEE International Conference on Automation Science and Engineering.
- Schmutzlera et al., “Compensating Warpage of 3D Printed Parts Using Free-Form Deformation,” 2016, Procedia CIRP 41, 1017-1022.
- Tong et al., “Parametric Error Modeling and Software Error Compensation for Rapid Prototyping,” 2003, Rapid Prototyping Journal vol. 9, No. 5, 301-313.
- Tong et al., “Error Compensation for Fused Deposition Modeling (FDM) Machine by Correcting Slice Files,” 2008, Rapid Prototyping Journal, 14/1, 4-14.
- Wang et al., “Influence of Process Parameters on Stereolithography Part Shrinkage,” 1996, Materials & Design. vol. 17, No. 4, pp. 205-213.
- Wang et al., “In-Plane Shape-Deviation Modeling and Compensation for Fused Deposition Modeling Processes,” Apr. 2017, IEEE Transactions on Automation Science and Engineering, vol. 14, No. 2.
- Xu et al., “Shape Deviation Modeling for Dimensional Quality Control in Additive Manufacturing,” Nov. 15, 2013, Proceedings of the ASME 2013 International Mechanical Engineering Congress and Exposition.

**Patent History**

**Patent number**: 11150633

**Type:**Grant

**Filed**: May 13, 2019

**Date of Patent**: Oct 19, 2021

**Patent Publication Number**: 20190346830

**Assignees**: Purdue Research Foundation (West Lafayette, IN), University Of Southern California (Los Angeles, CA)

**Inventors**: Raquel de Souza Borges Ferreira (West Lafayette, IN), Arman Sabbaghi (West Lafayette, IN), Qiang Huang (Los Angeles, CA), Kevin Amstutz (West Lafayette, IN)

**Primary Examiner**: Chun Cao

**Assistant Examiner**: Michael Tang

**Application Number**: 16/411,035

**Classifications**

**Current U.S. Class**:

**3-d Product Design (e.g., Solid Modeling) (700/98)**

**International Classification**: G05B 19/4099 (20060101); G05B 13/02 (20060101); B33Y 50/02 (20150101);