EP4731421A1

EP4731421A1 - Mechanical property prediction from manufacturing parameters for additive manufacturing

Info

Publication number: EP4731421A1
Application number: EP23741573.2A
Authority: EP
Inventors: Kaijuan CHEN; Zhi Hui KOH; Kim Quy LE; Jun Zeng; Hejun Du; Kun Zhou
Original assignee: Nanyang Technological University; Peridot Print LLC
Current assignee: Nanyang Technological University; Peridot Print LLC
Priority date: 2023-06-22
Filing date: 2023-06-22
Publication date: 2026-04-29
Also published as: CN121398956A; WO2024263190A1

Abstract

Systems, methods, and computer storage media are provided for predicting material and mechanical properties of an object from manufacturing parameters (108) for additive manufacturing. In examples described herein, a machine learning model (110A) is trained to predict crystallinity of build material based on manufacturing parameters (108) during additive manufacturing. The machine learning model (110A) predicts a crystallinity for at least one location of an object based on corresponding manufacturing parameters (108) for the object. Based on the crystallinity for the location of the object, a mechanical property is computed for the location of the object.

Description

MECHANICAL PROPERTY PREDICTION FROM MANUFACTURING PARAMETERS FOR ADDITIVE MANUFACTURING

BACKGROUND OF THE INVENTION

[0001] Three-dimensional (3D) solid parts may be produced from a digital model using additive manufacturing. Additive manufacturing, such as 3D printing, may be used in rapid prototyping, mold generation, mold master generation, and short-run manufacturing. Additive manufacturing involves the application of successive layers of build material. This is unlike some machining processes that often remove material to create the final part. In some additive manufacturing techniques, the build material may be cured or fused.

BRIEF DESCRIPTION OF THE DRAWING

[0002] Figure 1 is a block diagram of an example apparatus that may be used in material and mechanical property prediction from manufacturing parameters for additive manufacturing.

[0003] Figure 2 is a block diagram illustrating an example computer-readable medium for material and mechanical property prediction from manufacturing parameters for additive manufacturing.

[0004] Figure 3 is a diagram of example material property, mechanical property, and mechanical behavior prediction and simulation from manufacturing parameters for additive manufacturing.

[0005] Figure 4 is a flow diagram showing an example method for crystallinity prediction from manufacturing parameters for additive manufacturing.

[0006] Figure 5 is a flow diagram showing an example method for mechanical property prediction from manufacturing parameters for additive manufacturing.

[0007] Figure 6 is a flow diagram showing an example method for mechanical behavior simulation based on material and mechanical property prediction from manufacturing parameters for additive manufacturing.

[0008] Figure 7 is a flow diagram showing an example method for mechanical property improvement based on mechanical property prediction from manufacturing parameters for additive manufacturing. DET AILED DESCRIPTION OF THE INVENTION

[0009] Additive manufacturing may be used to manufacture three-dimensional (3D) objects. 3D printing is an example of additive manufacturing. Multi-Jet Fusion (MJF) is an example of 3D printing. During the MJF 3D printing process, a 3D printed part is printed layer- by-layer in a build volume. Each layer includes polymer powder spread in a desired orientation and fusing agents and/or detailing agents are selectively deposited on the polymer powder. Generally, detailing agents and fusing agents are applied in a precise manner during the additive manufacturing process. Each layer is then subject to energy from static and/or dynamic energy sources, such as radiation energy from optical energy sources (e.g., lamps) or other types of energy sources, causing the fusing agents to melt the polymer powder. Each layer is then cooled (e.g., by depositing the next layer of cold powder on top of each layer), thereby beginning to crystallize the polymer powder of the layer into solid state, to form the desired pattern for the layer. The process is repeated for each layer until printing of the 3D part is completed and the entire 3D part cools.

[0010] Other additive manufacturing techniques, such as stereolithography (SLA), fused deposition modeling (FDM), Polymer powder bed fusion additive manufacturing (PBF- AM), powder bed fusion, high-speed sintering, selective laser sintering, etc., selectively harden material of each layer to create a part (e.g., an object) layer-by-layer by crystallizing the respective material (e.g., polymers, thermoplastics, etc.) through changes in energy during the additive manufacturing process. Various build materials can be used during additive manufacturing, such as Polyamide (e.g., PA12), ABS (Acrylonitrile Butadiene Styrene), PLA (Polylactic Acid), PET (Polyethylene Terephthalate), latex, TPU (Thermoplastic Polyurethane), etc. As used herein, “build material” refers to the material and/or materials that are used to create the 3D object during additive manufacturing

[0011] As such, in various additive manufacturing techniques, each voxel of each layer of the additively manufactured part is subject to different temperature histories (e.g., thermal histories, temperature gradients, or temperature over time curves during the heating and cooling processes) during the additive manufacturing process, which results in different crystallinity for the material of each voxel. The different crystallinity for the material of each voxel results in different material and mechanical properties for the material of each voxel. A voxel is a representation of a location in a 3D space (e.g., a component of a 3D space). For instance, a voxel may represent a volume that is a subset of the 3D space. In some examples, voxels may be arranged on a 3D grid. For instance, a voxel may be cuboid or rectangular prismatic in shape. In some examples, voxels in the 3D space may be uniformly sized or non-uniformly sized. Examples of a voxel size dimension may include 25.4 millimeters (mm)/150 » 170 microns for 150 dots per inch (dpi), 490 microns for 50 dpi, 2 mm, 4 mm, etc. The term “voxel level” and variations thereof may refer to a resolution, scale, or density corresponding to voxel size.

[0012] Some simulation software is not applicable to additively manufactured (e.g., 3D printed) parts where material properties of the material of each voxel are non-uniform or heterogeneous. In order to test material properties or mechanical behavior of an additively manufactured part where material properties of the material of each voxel are non-uniform when using such software, an end user additively manufactures the part and manually measures the material properties or mechanical behavior of the part. However, the manual testing of the 3D printed part involves a significant amount of testing equipment, experience using the testing equipment, and time to test the part, which is not feasible under most circumstances. Further, in order to address any issues with the material properties or mechanical behavior of the part, the end user guesses as to what changes to the additively manufacturing of the part would affect material properties, mechanical properties, or mechanical behavior of the different portions of the 3D part as there is no way to determine the outcome of specific changes to portions of the 3D part during additive manufacturing without actually manufacturing the part, which takes a significant amount of time. The iterative process of guessing modifications to the 3D part, printing the 3D part, and manually testing the 3D part is financially expensive, time consuming, and computationally expensive. Each time a user manually modifies the 3D part, prints the 3D part, and manually tests the 3D part, it subjects a computing system to a significant number of computing input/output operations for a significant period of time, thereby increasing computing resource utilization, and increases computing network resource utilization when the data is sent over a network, thereby decreasing network throughput and increasing latency.

[0013] As such, examples of the present disclosure are directed to predicting material and mechanical properties of an object from manufacturing parameters for additive manufacturing. In this regard, examples described herein facilitate material and mechanical property prediction from manufacturing parameters in an efficient and effective manner in order to simulate and/or optimize material properties, mechanical properties, and/or mechanical behavior of an object for additive manufacturing. For example, a temperature history during additive manufacturing of an object is determined for each voxel of the object based on manufacturing parameters for additively manufacturing the 3D object. A machine learning model is trained to predict the crystallinity for the material of each voxel based on the simulated temperature history for each voxel. The mechanical properties of the material for each voxel can be determined through a model that provides mechanical properties as a function of the crystallinity of the material. Material properties, mechanical properties, and/or the mechanical behavior can then be simulated for any portion of the object or the entire object based on the mechanical properties determined for each voxel, for example, through finite element analysis. The simulation of the material properties, mechanical properties, and/or the mechanical behavior can be utilized to automatically or manually modify the manufacturing parameters of the object to improve a material property, mechanical property, and/or mechanical behavior of the object.

[0014] In operation, as described herein, a temperature history for each voxel can be determined based on manufacturing parameters for additively manufacturing the 3D part. For example, the manufacturing parameters for additively manufacturing the 3D part can include a geometry for a part. The manufacturing parameters for additively manufacturing the 3D part can also include the amount and type of materials (e.g., polymer powder, fusing agents and/or detailing agents, etc.) used during additive manufacturing of the part. The manufacturing parameters for additively manufacturing the 3D part can also include the location and orientation of the part within a build volume with respect to energy sources and/or cooling sources. Further, the manufacturing parameters can include printer settings, such as irradiance level of the energy sources (e.g., lamps), powder spread speed, etc.

[0015] A build volume (e.g., a voxel space) is a 3D space for object manufacturing. For example, a build volume may be a cuboid space in which an apparatus (e.g., computer, 3D printer, etc.) may deposit material (e.g., polymer powder, fusing agents, detailing agents, etc.) to manufacture an object or part. In some examples, a build volume can be any other type of 3D volume, such as a cylindrical space, etc. In some examples, an apparatus may progressively fill a build volume layer-by-layer with material and during manufacturing. In some examples, a location within a build volume may be expressed in coordinates. For example, locations in a voxel space may be expressed in three coordinates: x (e.g., width), y (e.g., length), and z (e.g., height). A virtual representation of a build volume can be utilized by computer-aided design (CAD) software and/or build preparation software.

[0016] Energy sources may be any amount or type of energy sources, such as optical energy sources (e.g., lamps, lasers, infrared, etc.). For example, a static energy source (e.g., the energy source does not move) may be located in a certain location above, or adjacent to, a build volume and a dynamic energy source (e.g., the energy source moves during the additive manufacturing process) may move to various locations during the additive manufacturing process, for example, to melt fusing agents at a specific voxel(s).

[0017] In this regard, the manufacturing parameters for additively manufacturing the 3D part can be used to determine the expected temperature history for each voxel based on the expected temperature over time curve (e.g., temperature gradients or thermal history) of the build volume with respect to the energy sources during additive manufacturing of the 3D part and the location of each voxel of the 3D part in the build volume. For example, temperature measurements can be obtained through sensors, such as thermocouples, infrared cameras, etc., at various locations of the build volume at various times during an additive manufacturing process in order to obtain ground truth temperature measurements. An algorithm, such as a neural network, a regression model, and/or the like, can be trained to predict the expected temperature history for the build volume based on the ground truth temperature measurements. In some examples, the algorithm trained to predict the expected temperature history for the build volume can receive temperature measurements during various subsequent additive manufacturing processes in order to continuously train the algorithm to account for variation(s) in printer operation (e.g., environmental variation, printer drift, printer variation, and/or printer functioning). The expected temperature history for each voxel during additive manufacturing of the part can be determined (e.g., interpolated) as a function of boundary conditions of the build volume and/or the 3D part based on a physics-based simulation through thermodynamics laws (e.g., such as how heat flux conducted out of the build volume through a boundary is balanced out by the heat flux taken away by air flow surrounding the boundary by the following equation: -kn* T(x,f)=h*(T _x etc.). In this regard, the expected temperature history of the entire build volume can be predicted based on input manufacturing parameters, such as printing parameters, material properties, printing geometry, etc., in order to predict the expected temperature history of each voxel of the 3D part through a physics-based simulation. [0018] A machine learning model is trained to predict the crystallinity for the material of each voxel based on the simulated temperature history for each voxel. As each voxel is subject to a different temperature history (e.g., temperature gradients, temperature vs. time, etc.), the hardening of the material of each voxel will result in a different crystallinity percentage for the material of each voxel. As an example, in order to train the machine learning model, the temperature for different locations of a training object (e.g., a 3D part used to train the machine learning model) at different times throughout the additive manufacturing process can be determined (e.g., the temperature history as determined above, taking temperature measurements through thermocouples at various locations during the additive manufacturing process, etc.). Also, in order to train the machine learning model, the crystallinity for the different locations can be measured following additive manufacturing of the training object (e.g., by performing differential scanning calorimetry measurement or any other measurement for crystallinity, such as X-ray diffraction, spectroscopy techniques, etc.). The measured crystallinity for the different locations following additive manufacturing of the training object can be compared to the predicted crystallinity for the different locations of the training object by the machine learning model to compute a loss function to backpropagate into the machine learning model.

[0019] The temperature for different locations of the part at different times throughout the additive manufacturing process is the input layer or layers of the machine learning model and the resulting crystallinity for the different locations is the output layer or layers of the machine learning model. For example, in the example shown at block 306 of figure 3, the temperature at each of the timestamps (e.g., T1 through T1415) for each of the locations (e.g., the 294 locations as indicated by the XYZ axes of block 302) are input as an input layer of the machine learning model and the predicted crystallinity of each location is the output layer of the machine learning model. The weights and biases of the machine learning model, including those of a hidden layer, are iteratively optimized by comparing the predicted crystallinity to the measured crystallinity by backpropagating the loss function into the machine learning model for each iteration. Any amount of training objects and data points for the training objects can be utilized to optimize the machine learning model to make better predictions and generalize unseen data. After the machine learning model is trained, the expected temperature history for each voxel can be fed as input features to the machine learning model and the trained hidden layer or layers is used to perform computations to output the predicted crystallinity of the material of each voxel. In other examples, a regression model can be utilized order to predict the crystallinity of material of each voxel based on an expected temperature history for each voxel.

[0020] The mechanical properties of the material for each voxel can be determined through a model that provides mechanical properties as a function of the crystallinity of the material. For example, a viscoelastic-viscoplastic constitutive model as a function of crystallinity can provide mechanical properties. Generally, a viscoelastic-viscoplastic constitutive model includes equations describing the relationship of a material to stress, strain, and time. A viscoelastic-viscoplastic constitutive model mathematically models the deformation behavior of a material under loading by combining viscoelasticity and viscoplasticity in order to represent the behavior of materials as dependent on time and dependent on rate. Viscoelasticity refers to deformation of a material with respect to time, where deformation of the material depends on magnitude of the applied stress and the duration of the applied stress. Viscoplasticity refers to deformation of a material with respect to rate, where deformation of the material depends on loading rate and loading history.

[0021] In this regard, a viscoelastic-viscoplastic constitutive model as a function of crystallinity can provide mechanical properties, such as instantaneous elasticity (e.g., Young’s modulus), viscoelasticity (e.g., non-linear stress function), and viscoplasticity (e.g., hardening modulus), as a function of the crystallinity of the material. Any other material or mechanical properties that are based on the crystallinity of a material can be included in a model that relates the material and/or mechanical properties to the crystallinity of the material.

[0022] Material properties, mechanical properties, and/or the mechanical behavior can then be simulated for any portion of the part or the entire part based on the properties determined for each voxel. For example, the mechanical properties, as determined based on the crystallinity of the material, for each voxel can be input into a finite element analysis (FEA) program (e.g., FEA within computer-aided design (CAD) software and/or build preparation software or any program using finite element method (FEM)) to determine mechanical behavior of the part, such as deformation behavior, failure criteria (e.g., elongation at break), etc., in order to simulate the mechanical behavior of the part or portions of the part.

[0023] As another example, the CAD and/or build preparation software can simulate the distribution of material properties (e.g., physical properties, such as crystallinity, etc.) of the part based on the material properties determined for each voxel of the part. As another example, the CAD and/or build preparation software can simulate the distribution of mechanical properties of the part based on the mechanical properties determined for each voxel of the part. In this regard, the CAD and/or build preparation software can output the final part geometry, the crystallinity distribution at a voxel level, any other material properties distribution at a voxel level, and/or any mechanical property distribution property at a voxel level for display to an end user. Further, the CAD and/or build preparation software can output the simulated final part mechanical behavior, for example, simulating how this part responds to loading, extension, fatigue relevant to this part’s field of use, etc. for display to an end user. If the part is acceptable (e.g., the user determines the mechanical behavior, mechanical properties and/or material properties acceptable), the 3D part can be stored and/or additively manufactured (e.g., 3D printed).

[0024] The simulation of the material properties, mechanical properties, and/or the mechanical behavior can be utilized to automatically or manually modify the manufacturing parameters of the part to improve a material property, mechanical property, and/or mechanical behavior of the part. For example, if a material property, mechanical property, and/or mechanical behavior of the part is unacceptable (e.g., the 3D part or a portion of the 3D part is too weak), the 3D part can be modified using the CAD and/or build preparation software to improve the part manually or in automated fashion.

[0025] As an example, the CAD and/or build preparation software can automatically generate a new geometry for the part to present for approval by an end user. Alternatively, an end user can manually modify the part to generate a new geometry for the part.

[0026] As another example, the CAD and/or build preparation software can automatically determine a different location in the build volume for the part and/or a different orientation of the part in the build volume for additive manufacturing of the part to present for approval by an end user in order to improve a material property, mechanical property, and/or mechanical behavior of the part. In this regard, placing a part at the center of the build volume, as opposed to placing the part on the edge of the build volume, will result in a more gradual cooling curve in the temperature history for each of the voxels of the part, thereby resulting in a different crystallinity (and other material properties), mechanical properties, and mechanical behavior of the part at a voxel level. Alternatively, an end user can manually modify the location in the build volume for the part and/or the orientation of the part in the build volume for additive manufacturing.

[0027] As another example, the CAD and/or build preparation software can automatically determine the amount and type of materials (e.g., how thick of a layer of polymer powder, what contone level of or how much fusing agents and/or detailing agents, etc. will be applied for each voxel of each layer) for each voxel during additive manufacturing of the part to present for approval by an end user in order to improve a material property, mechanical property, and/or mechanical behavior of the part. Alternatively, an end user can manually modify the quantities of material for each voxel.

[0028] As another example, the CAD and/or build preparation software can automatically determine different energy settings for heating and/or cooling the materials during additive manufacturing to present for approval by an end user in order to improve a material property, mechanical property, and/or mechanical behavior of the part. Alternatively, an end user can manually modify energy settings for heating and/or cooling the materials during additive manufacturing.

[0029] In this regard, the 3D part can be modified using the CAD and/or build preparation software to improve the part manually or in automated fashion. As an example, the CAD and/or build preparation software can include a 3D printer application programming interface (API) to interact with the 3D printer to implement any combination of these improvements. An end user can then manually modify the manufacturing parameters of the part in the build preparation and/or CAD software. Alternatively, the build preparation and/or CAD software can automatically provide suggestions to improve the 3D part. As another example, the user can select a certain aspect of the 3D part to improve in the build preparation and/or CAD software and the build preparation and/or CAD software can automatically provide suggestions for approval by the end user. In some examples, the user can also modify certain settings of the additive manufacturing device, such as a 3D printer, from a control panel or control knob on the device. The modification through the control panel or control knob of the device can then be received by the build preparation and/or CAD software to modify the 3D part for further simulation. In examples, if the user determines the mechanical behavior, mechanical properties and/or material properties acceptable, the 3D part (and/or improvements to the 3D part) can be stored and/or additively manufactured.

[0030] Advantageously, material properties, mechanical properties, and/or mechanical behavior of a part for additive manufacturing can be simulated and optimized based on manufacturing parameters using implementations described herein. In particular, the determination of material properties and/or mechanical properties of each voxel based on the predicted crystallinity of material of each voxel allows the simulation of material properties, mechanical properties, and/or mechanical behavior of the part without requiring the user to waste significant time and resources to additively manufacture (e.g., 3D print) the part, purchase and know how to use expensive testing equipment, and subject the manufactured part to expensive testing. Further, the simulation of material properties, mechanical properties, and/or mechanical behavior of the part based on the predicted crystallinity of material of each voxel of the part allows for the optimization (e.g., improvement) of material properties, mechanical properties, and/or mechanical behavior of the part without requiring the iterative, time-consuming process of guessing modifications to the 3D part, additively manufacturing the 3D part, and manually testing the 3D part. In this regard, the simulation of material properties, mechanical properties, and/or mechanical behavior of the part based on the predicted crystallinity of material of each voxel of the part reduces computing input/output operations for significant periods of time, thereby conserving computing resources and computing network resources.

[0031] Figure 1 is a block diagram of an example apparatus 102 that may be used in material and mechanical property prediction from manufacturing parameters for additive manufacturing. The apparatus 102 may be a computing device, such as a personal computer, a server computer, a printer, a 3D printer, a smartphone, a tablet computer, a smart appliance, etc. The apparatus 102 may include and/or may be coupled to a processor 104 and/or a memory 106. The memory 106 may be in electronic communication with the processor 104. For instance, the processor 104 may write to and/or read from the memory 106. In some examples, the apparatus 102 may be in communication with (e.g., coupled to, have a communication link with) an additive manufacturing device (e.g., an additive manufacturing device, such as a 3D printing device). In some examples, the apparatus 102 may be an example of an additive manufacturing device (e.g., a 3D printing device). The apparatus 102 may include additional components (not shown) and/or some of the components described herein may be removed and/or modified without departing from the scope of this disclosure.

[0032] As used herein, the term “processor” refers to a computer processor that uses physical phenomena (e.g., electrical, optical, mechanical, or quantum phenomena) to perform logical or arithmetic operations on physical representations of information. The processor 104 is any type of processor other than the human mind. For example, the processor 104 may be any of a central processing unit (CPU), a semiconductor-based microprocessor, graphics processing unit (GPU), field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), and/or other hardware device suitable for retrieval and execution of instructions stored in the memory 106. In some examples, the processor 104 may perform one, some, or all of the functions, operations, elements, methods, etc., described in relation to one, some, or all of Figures 1-7. As used herein, the term “memory” refers to a device that uses physical phenomena to store physical representations of information. The memory 106 is any type of memory other than the human mind. For example, the memory 106 may be any electronic, magnetic, optical, and/or other physical storage device that contains or stores electronic information (e.g., instructions and/or data), such as Random Access Memory (RAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), a storage device, an optical disc, and/or the like. In some implementations, the memory 106 may be a non-transitory tangible machine-readable storage medium, where the term “non-transitory” does not encompass transitory propagating signals.

[0033] In some examples, the apparatus 102 may include an input/output interface (not shown) through which the processor 104 may communicate with an external device or devices (not shown), for instance, to receive and store the information pertaining to the manufacturing parameters of the part through various input and/or output devices, such as a keyboard, a mouse, a display, another apparatus, electronic device, computing device, computer, removable storage, network device, etc., through which a user may input instructions into and/or receive data from the apparatus 102.

[0034] In some examples, the memory 106 may store manufacturing parameters data 108. The manufacturing parameters data 108 may be generated by the apparatus 102 and/or received from another device. Some examples of manufacturing parameters data 108 include 3D model data, such as a 3D manufacturing format (3MF) file or files, a 3D computer-aided design (CAD) image, object shape data, mesh data, geometry data, etc. The manufacturing parameters data 108 may indicate the geometry (e.g., shape) of an object or objects (e.g., a part of parts). In some examples, the manufacturing parameters data 108 may indicate a location of an object with respect to a build volume, or the apparatus 102 may arrange a 3D object model represented by the manufacturing parameters data 108 into a build volume. In some examples, the manufacturing parameters data 108 may indicate the amount and type of materials (e.g., how thick of a layer of polymer powder, what contone level of or how much fusing agents and/or detailing agents, etc. will be applied for each voxel of each layer) used during additive manufacturing of the part. In some examples, the manufacturing parameters data 108 may indicate the location and orientation of the part within a build volume with respect to energy sources and/or cooling sources during additive manufacturing of the part. In some examples, the manufacturing parameters data 108 may indicate the energy emitted by the energy sources and/or removed by the cooling sources during additive manufacturing of the part. In some examples, manufacturing parameters data 108 may include a print mode, which may indicate the contone level of agents to be applied at each location of the build volume (e.g., through a contone map, which is a set of data indicating a location or locations and amount of printing agents), the amount of energy to be applied, etc. during the additive manufacturing process.

[0035] In some examples, the apparatus 102 may utilize the manufacturing parameters data 108 to determine the expected temperature history for each voxel based on the expected temperature over time curve (e.g., temperature gradients or thermal history) of the build volume with respect to the energy sources during additive manufacturing of the 3D part and the location of each voxel of the 3D part in the build volume. For example, apparatus 102 can obtain temperature measurements through sensors, such as thermocouples, at various locations of the build volume at various times during an additive manufacturing process in order to obtain ground truth temperature measurements, which may be stored in the memory 106. An algorithm, such as a neural network, a regression model, and/or the like, which may be stored in the memory 106, can be trained to predict the expected temperature history for the build volume based on the ground truth temperature measurements. In some examples, the algorithm trained to predict the expected temperature history for each voxel can receive temperature measurements during various subsequent additive manufacturing processes in order to continuously train the algorithm to account for variation(s) in printer operation (e.g., environmental variation, printer drift, printer variation, and/or printer functioning). The expected temperature history for each voxel during additive manufacturing of the part can be determined (e.g., interpolated) by apparatus 102 as a function of boundary conditions of the build volume and/or the 3D part based on a physics-based simulation through thermodynamics laws (e.g., such as how heat flux conducted out of the build volume through a boundary is balanced out by the heat flux taken away by air flow surrounding the boundary by the following equation: -kn* VT(x,t)=h*(T «> etc.), which may be stored in memory 106. In this regard, the expected temperature history of the entire build volume can be predicted based on input manufacturing parameters data 108, such as printing parameters, material properties, printing geometry, etc., in order to predict the expected temperature history of each voxel of the 3D part through a physics-based simulation.

[0036] In some examples, the manufacturing parameters data 108 may be utilized to modify a temperature history of each voxel of the part during additive manufacturing of the part in order to modify a material property, a mechanical property and/or mechanical behavior of the part. For example, the apparatus 102 may modify the manufacturing parameters data 108 to modify a material property, a mechanical property and/or mechanical behavior of the part during additive manufacturing of the part, which may be stored in the memory 106.

[0037] In some examples, the memory 106 may store crystallinity prediction instructions 110. The processor 104 may execute the crystallinity prediction instructions 110 to predict a crystallinity of material of each voxel based on the temperature history for each voxel (e.g., the temperature history for each voxel can be determined based on manufacturing parameters 108). In some examples, crystallinity predictions instruction 110 predicts crystallinity, crystal phase, crystal morphologies, crystal size of a material based on the temperature history of the material. In some examples, crystallinity predictions instructions 110 predicts crystallinity of material of each voxel of a 3D part through a machine learning model 110A. In some examples, crystallinity predictions instructions 110 predicts crystallinity of material of each voxel of a 3D part based on the temperature history for each voxel through a regression model, such as linear regression. In some examples, crystallinity predictions instructions 110 predicts crystallinity of material of each voxel of a 3D part based on the temperature history for each voxel through a statistical model, neural network, other similar processes, and/or combinations of processes. The predicted crystallinity from crystallinity prediction instruction 110 can be stored in memory 106.

[0038] Machine learning model 110A can be any type of machine learning model. For example, machine learning model 110A can be a convolutional neural networks (CNNs) (e.g., basic CNN, deconvolutional neural network, inception module, residual neural network, etc.), recurrent neural networks (RNNs) (e.g., basic RNN, multi-layer RNN, bi-directional RNN, fused RNN, clockwork RNN, etc.), graph neural networks (GNNs), etc. As an example, machine learning model 110A can be trained to predict the crystallinity for the material of each voxel based on the simulated temperature history for each voxel as determined from manufacturing parameters 108. In order to train machine learning model 110A, the temperature for different locations of a training object(s) (e.g., a 3D part used to train the machine learning model) at different times throughout the additive manufacturing process can be determined (e.g., the temperature history as determined from manufacturing parameters 108, taking temperature measurements through thermocouples at various locations during the additive manufacturing process, etc.) and stored in memory 106. Also, in order to train the machine learning model 110A, the crystallinity for the different locations can be measured following additive manufacturing of the training object(s) (e.g., by performing differential scanning calorimetry measurement or any other measurement for crystallinity, such as X-ray diffraction, spectroscopy techniques, etc.) and stored in memory 106. The measured crystallinity for the different locations following additive manufacturing of the training object(s) can be compared to the predicted crystallinity for the different locations of the training object(s) by the machine learning model 110A to compute a loss function to backpropagate into the machine learning model 110A. The temperature for different locations of the part at different times throughout the additive manufacturing process is the input layer of the machine learning model 110A and the resulting crystallinity for the different locations is the output layer of the machine learning model 110A.

[0039] The weights and biases of the machine learning model 110A, including those of a hidden layer of machine learning model 110A, are iteratively optimized by comparing the predicted crystallinity to the measured crystallinity by backpropagating the loss function into the machine learning model for each iteration. The weights and biases of the machine learning model 110A can then be stored in memory 106 Any amount of training objects and data points for the training objects can be utilized to optimize the machine learning model 110 A to make better predictions and generalize unseen data. After the machine learning model 110A is trained, the expected temperature history for each voxel can be fed as input features to the machine learning model 110A and the trained hidden layer or layers of machine learning model 110A are used to perform computations to output the predicted crystallinity of the material of each voxel for unseen 3D objects based on manufacturing parameters 108. The predicted crystallinity from machine learning model 110A of crystallinity prediction instructions 110 can be stored in memory 106.

[0040] In some examples, the memory 106 may store crystallinity to mechanical property instructions 112. The processor 104 may execute the crystallinity to mechanical property instructions 112 to determine mechanical properties of the material of each voxel based on the crystallinity of the material of each voxel. In some examples, crystallinity to mechanical property instructions 112 determines mechanical properties of the material of each voxel based on crystallinity, crystal phase, crystal morphologies, and/or crystal size of the material at each voxel. In some examples, crystallinity to mechanical property instructions 112 determines mechanical properties based on the crystallinity of a material through a constitutive model 112A. For example, constitutive model 112A can be a viscoelastic-viscoplastic constitutive model as a function of crystallinity. In this regard, constitutive model 1 12A can provide mechanical properties, such as instantaneous elasticity (e.g., Young’s modulus), viscoelasticity (e.g., non-linear stress function), and viscoplasticity (e.g., hardening modulus), as a function of the crystallinity of the material. Any other material or mechanical properties (e.g., Poisson’s ratio) that are based on the crystallinity of a material can be included in a model of crystallinity to mechanical property instructions 112 to relate the material and/or mechanical properties to the crystallinity of the material. The mechanical properties of the material of each voxel based on the crystallinity of the material of each voxel as determined by crystallinity to mechanical property instructions 112 can be stored in memory 106. [0041] In some examples, the memory 106 may store simulation instructions 114. The processor 104 may execute the simulation instructions 114 to simulate material properties, mechanical properties, and/or the mechanical behavior for any portion of the part or the entire part based on the mechanical properties determined for each voxel. For example, simulation instructions 114 can input the mechanical properties, as determined based on the crystallinity of the material by crystallinity to mechanical property instructions 112, into a FEA program (e.g., FEA within CAD software and/or build preparation software or any program using FEM) to determine mechanical behavior of the part, such as deformation behavior, failure criteria (e.g., elongation at break), etc., in order to simulate the mechanical behavior of the part or portions of the part. The mechanical behavior can be stored in memory 106.

[0042] As another example, simulation instructions 114 can simulate the distribution of material properties (e.g., physical properties, such as crystallinity, etc.) of the part based on the material properties determined for each voxel of the part. As another example, the CAD and/or build preparation software can simulate the distribution of mechanical properties of the part based on the mechanical properties determined for each voxel of the part. The distribution of mechanical properties and/or mechanical properties of the part can be stored in memory 106. In some examples, simulation instructions 114 can output the final part geometry, the crystallinity distribution at a voxel level, any other material properties distribution at a voxel level, and/or any mechanical property distribution property at a voxel level for display to an end user, which can also be stored in memory 106. In some examples, simulation instructions 114 can output the simulated final part mechanical behavior, for example, simulating how the part responds to loading, extension, fatigue as relevant to the part’s field of use, etc. for display to an end user, which can also be stored in memory 106.

[0043] In some examples, the memory 106 may store manufacturing parameter optimization instructions 1 16. The processor 104 may execute the manufacturing parameter optimization instructions 116 to automatically (or manually allow a user to) modify the manufacturing parameters data 108 of the part to improve a material property, mechanical property, and/or mechanical behavior of the part. The manufacturing parameters data 108 as modified by manufacturing parameter optimization instructions 116 can be stored in memory 106.

[0044] As an example, manufacturing parameter optimization instructions 116 can automatically generate a new geometry for the part to present for approval by an end user. Alternatively, an end user can manually modify the part to generate a new geometry for the part through manufacturing parameter optimization instructions 116.

[0045] As another example, manufacturing parameter optimization instructions 116 can automatically determine a different location in the build volume for the part and/or a different orientation of the part in the build volume for additive manufacturing of the part to present for approval by an end user in order to improve a material property, mechanical property, and/or mechanical behavior of the part. Alternatively, an end user can manually modify the location in the build volume for the part and/or the orientation of the part in the build volume for additive manufacturing through manufacturing parameter optimization instractions 116.

[0046] As another example, manufacturing parameter optimization instructions 116 can automatically determine the amount and type of materials (e.g., how thick of a layer of polymer powder, what contone level of or how much fusing agents and/or detailing agents, etc. will be applied for each voxel of each layer) for each voxel used during additive manufacturing of the part to present for approval by an end user in order to improve a material property, mechanical property, and/or mechanical behavior of the part. Alternatively, an end user can manually modify the quantities of material for each voxel through manufacturing parameter optimization instructions 116.

[0047] As another example, manufacturing parameter optimization instructions 116 can automatically determine different energy settings for heating and/or cooling the materials during additive manufacturing to present for approval by an end user in order to improve a material property, mechanical property, and/or mechanical behavior of the part. Alternatively, an end user can manually modify energy settings for heating and/or cooling the materials during additive manufacturing through manufacturing parameter optimization instructions 116.

[0048] In some examples, the memory 106 may store manufacturing instructions 1 18. The processor 104 may execute the manufacturing instructions 118 to manufacture the 3D part based on manufacturing parameters data 108 through an additive manufacturing device, such as a 3D printer. In some examples, manufacturing instructions 118 interact with an API of the additive manufacturing device. In some examples, if the user determines the mechanical behavior, mechanical properties and/or material properties acceptable, the 3D part (and/or improvements to the 3D part) can be stored and/or additively manufactured through manufacturing instructions 118. [0049] Figure 2 is a block diagram illustrating an example computer-readable medium 202 for material and mechanical property prediction from manufacturing parameters for additive manufacturing. The computer-readable medium 202 may be a non-transitory, tangible computer-readable medium 202. The computer-readable medium 202 may be, for example, RAM, EEPROM, a storage device, an optical disc, and the like. In some examples, the computer-readable medium 202 may be volatile and/or non-volatile memory, such as DRAM, EEPROM, MRAM, PCRAM, memristor, flash memory, and/or the like. In some implementations, the memory 106 described in relation to Figure 1 may be an example of the computer-readable medium 202 described in relation to Figure 2.

[0050] The computer-readable medium 202 may include data (e.g., information, instructions, and/or executable code, etc.). For example, the computer-readable medium 202 may include manufacturing parameters data 208, crystallinity prediction instructions 210, crystallinity to mechanical property instructions 212, simulation instructions 214, manufacturing parameter optimization instructions 216, and/or manufacturing instructions 218. [0051] In some examples, the computer-readable medium 202 may store manufacturing parameters data 208. In some examples, manufacturing parameters data 208 may be as described in relation to any examples described herein, such as in relation to Figures 1-7.

[0052] In some examples, the crystallinity prediction instructions 210 are instructions when executed cause a processor of an electronic device to predict the crystallinity of material of each voxel of a part based on a temperature history of each voxel. In some examples, predicting the crystallinity of voxels of a part may be performed as described in relation to any examples described herein, such as in relation to Figures 1-7.

[0053] In some examples, the crystallinity to mechanical property instructions 212 are instructions when executed cause a processor of an electronic device to determine a mechanical property each voxel of a part based on a crystallinity of material of each voxel. In some examples, determining mechanical properties of voxels of a part may be performed as described in relation to any examples described herein, such as in relation to Figures 1-7.

[0054] In some examples, the simulation instructions 214 are instructions when executed cause a processor of an electronic device to simulate behavior of a part based on mechanical properties of each voxel of the part. In some examples, simulating behavior of a part may be performed as described in relation to any examples described herein, such as in relation to Figures 1-7. [0055] In some examples, the manufacturing parameter optimization instructions 216 are instructions when executed cause a processor of an electronic device to improve a part by optimizing manufacturing parameters of the part to change crystallinity of material of voxels of the part. In some examples, improving a part by optimizing manufacturing parameters of the part may be performed as described in relation to any examples described herein, such as in relation to Figures 1-7.

[0056] In some examples, the manufacturing instructions 218 are instructions when executed cause a processor of an electronic device to cause manufacturing of the part through additive manufacturing, such as 3D printing by a 3D printer. In some examples, causing manufacturing of a part through additive manufacturing may be performed as described in relation to any examples described herein, such as in relation to Figures 1-7.

[0057] Figure 3 is a diagram of example material property, mechanical property, and mechanical behavior prediction and simulation from manufacturing parameters for additive manufacturing. At block 302, layers of an object or objects are shown at certain XYZ coordinates with respect to a build volume. As can be appreciated, the layers (e.g., the ‘Z’ coordinates) correspond to each layer manufactured during additive manufacturing, such as 3D printing. Each voxel of the 3D object can be identified according to its corresponding XYZ coordinates. For example, voxel “2-4-1” corresponds to a voxel at row ‘2’ of the X-axis, column ‘4’ of the Y-axis, and layer ‘1’ of the Z-axis.

[0058] At block 304, examples of predicted temperature history (e.g., temperature in relation to time) of each certain locations are shown. In the example shown in block 305, the temperature in the range of 0° to 200° Celsius over the time period of 0 to 150,000 seconds is shown. The predicted temperature history of voxels “2-4-1,” “2-4-3,” “2-4-5,” “2-4-7,” “2-4- 9,” “2-4-11,” and “2-4-14” are shown as examples. The temperature history of the voxels in the example shown in block 305 are the based on the manufacturing parameters of the object in the build volume of block 302. The temperature history can be verified based on temperature recording by thermocouples during the additive manufacturing process.

[0059] At block 306, the temperature history for each location of each layer is input into a machine learning model that is trained to predict crystallinity of material of each location. For example, the machine learning model can be a neural network that includes an input layer, a hidden layer, and an output layer. Generally, the input layer receives the features (e.g., temperature history for each location) to be fed into the network, the hidden layer is used to perform computations on the features, and the output layer presents the final prediction result (e.g., the crystallinity of the material at each location). Although the example shown in block 306 is a two-layer neural network, any type of neural network with any amount of layers or other prediction algorithm (e.g., logistic regression) in order to predict the crystallinity of material of each location can be used.

[0060] In the example shown in block 306, a two-layer feedforward Levenberg- Marquardt (LM) backpropagation algorithm can be used to optimize the neural network. Generally, the LM algorithm optimizes weights to minimize the error (e.g., by backpropagating a loss function) between the predicted output of crystallinity of material of each location of the neural network and the actual measured crystallinity of material of each location of an additively manufactured 3D object. In order to optimize the weights of the neural network, the gradients of the weights are computed with respect to the error and the weights are updated through an iterative process. The algorithm of block 306 can use both steepest gradient descent and Newton’s method in order to optimize the iterative process of determining the weights of the neural network. For example, when the prediction is far from the actual solution, the steepest gradient descent method can be used and when the prediction is close to the actual solution, the Newton’s method can be used to increase the efficiency of the convergence process. Any number of locations of any number of objects can be used to train the neural network based on corresponding crystallinity measurements (e.g., through differential scanning calorimetry measurements after the additive manufacturing process) and temperature history (e.g., temperature history as predicted or measured during the additive manufacturing process) taken at the respective locations.

[0061] In the example shown in block 306, after the neural network is trained, the predicted temperature of each location (e.g., each voxel) at 1415 timestamps throughout the additive manufacturing process are fed as input into the neural network. The number of timestamps are shown as an example and any number of predicted temperature at certain times can be fed as input into the neural network.

[0062] Block 308 shows an example comparing crystallinity predicted for each voxel “2-4-1,” through “2-4-14” in comparison to the crystallinity of measured at each voxel “2-4- 1,” through “2-4-14.” As can be appreciated, in the example shown in block 308, the predicted crystallinities of the material at each voxel are close to the experimental crystallinities of the material as measured for each voxel. Further, as can be appreciated in comparison of block 302 to the predicted/measured crystallinity of material of the voxels, a lower crystallinity was found for locations near the boundaries of the build volume and a higher value for locations in the middle of the build volume due to the more gradual cooling in the middle of the build volume. [0063] At block 310, the mechanical properties of the material for each voxel can be determined through a model that provides mechanical properties as a function of the crystallinity of the material. For example, a viscoelastic-viscoplastic constitutive model as a function of crystallinity can provide mechanical properties, such as instantaneous elasticity (e.g., Young’s modulus), viscoelasticity (e.g., non-linear stress function), and viscoplasticity (e.g., hardening modulus), as a function of the crystallinity of the material. Any other material or mechanical properties that are based on the crystallinity of a material can be included in a model to relate the material and/or mechanical properties to the crystallinity of the material. At block 310, the mechanical properties of the material of each voxel as determined from the crystallinity of the material of each voxel can then be used to determine deformation behavior of the part. The deformation behavior of the part can be used to determine failure criteria of the part, such as elongation at break, etc.

[0064] In the example shown in block 310, the relationship between crystallinity of a material to tensile modulus (as shown in 310A) can be determined through testing. As well, in the example shown in block 310, the relationship between crystallinity of a material to elongation at break (as shown in 310B) can be determined through testing. As shown, the relationship between crystallinity of a material and mechanical properties of the material is determined through a method of data fitting the curve to the results of testing the material. Although data fitting is shown to relate crystallinity of a material and mechanical properties of the material, any type of analysis, such as a statistical model, neural network, other similar processes, and/or combinations of processes, may be utilized to relate crystallinity of a material (e.g., crystallinity, crystal phase, crystal morphologies, crystal size, etc.) and mechanical properties of the material.

[0065] As can be appreciated from block 310, for the example material shown (e.g., PA 12), the tensile modulus increases with the increasing crystallinity of the material and the elongation at break decreases with the increasing crystallinity of the material. In this regard, for this specific material, a higher crystallinity results in a higher tensile modulus, which results in an increased rigidity of polymers. The increased rigidity further reduces the mobility of molecular chains, thereby resulting in low fracture strain.

[0066] At block 312, results of simulating the part, such as a distribution of the crystallinity of material of each voxel of the part (as shown in 312A) and the distribution of stress of each voxel of the part (as shown in 312B) can be displayed to an end user. In this regard, material properties, mechanical properties, and/or mechanical behavior of the part can be simulated and displayed to an end user. As an example, the viscoelastic-viscoplastic constitutive model as a function of crystallinity of a material can be implemented into a finite element code as a user-defined material (UMAT) to simulate the deformation of a 3D object. The viscoelastic-viscoplastic constitutive model as a function of crystallinity of a material can include the instantaneous elastic modulus, the relaxation modulus, and the viscoplastic deformation resistance as a power- law relation with the crystallinity for a material. Further, the viscoelastic-viscoplastic constitutive model as a function of crystallinity of a material can include the failure criteria, such as the maximum principal elongation, for the material as a function of the crystallinity of the material. Further, the viscoelastic-viscoplastic constitutive model as a function of crystallinity of a material can include the relationship between the maximum principal elongation and crystallinity of the material. Although the viscoelastic- viscoplastic constitutive model as a function of crystallinity of a material is shown with respect to PA12, any semi-crystalline polymers can utilize a viscoelastic-viscoplastic constitutive model as a function of crystallinity of the material,

[0067] As can be appreciated, block 312 provides an example for evaluating the material properties, mechanical properties, and/or mechanical behavior of a 3D object for additive manufacturing. In the example shown in block 312, the temperature history on each voxel of the part was predicted for additive manufacturing. The temperature history was used to predict the crystallinity of the material of each voxel of the 3D object through a neural network. The crystallinity of material of each voxel of the printed part was mapped to the field variables of each node of finite element model as a distribution of crystallinity of the 3D object in 312A. As can be appreciated from 312 A, the crystallinity of different regions of the 3D object is different, which indicates that the 3D object is nonhomogeneous and anisotropic. A viscoelastic-viscoplastic constitutive model as a function of crystallinity was used to determine the mechanical properties of each voxel based on the crystallinity of each voxel of the 3D object. The mechanical properties of each voxel were simulated to determine mechanical behavior of the 3D object. As shown in 312B, a CAD model can show results of the FEA. The results of the FEA showing a response to maximum stress by the 3D object or deformation of 3D object can be used to evaluate whether the 3D object fails in certain portions of the 3D object. The simulation of the material properties, mechanical properties, and/or the mechanical behavior can be utilized to automatically or manually modify the manufacturing parameters of the part to improve a material property, mechanical property, and/or mechanical behavior of the part. For example, if a material property, mechanical property, and/or mechanical behavior of the part is unacceptable (e.g., the 3D part or a portion of the 3D part is too weak), the 3D part can be modified using the CAD and/or build preparation software to improve the part manually or in automated fashion.

[0068] Figure 4 is a flow diagram showing an example method 400 for crystallinity prediction from manufacturing parameters for additive manufacturing. The method 400 and/or an element or elements of the method 400 may be performed by an apparatus (e.g., electronic device). For example, the method 400 may be performed by the apparatus 102 described in relation to Figure 1.

[0069] In block 402, manufacturing parameters for a 3D part for additively manufacturing the 3D part can be obtained. For example, the manufacturing parameters for additively manufacturing the 3D part can include a geometry for a part. The manufacturing parameters for additively manufacturing the 3D part may indicate the amount and type of materials (e.g., how thick of a layer of polymer powder, what contone level of or how much fusing agents and/or detailing agents, etc. will be applied for each voxel of each layer) used during additive manufacturing of the part. The manufacturing parameters for additively manufacturing the 3D part can also include the location and orientation of the part within a build volume with respect to energy sources and/or cooling sources. The manufacturing parameters for additively manufacturing the 3D part can also include the energy emitted by the energy sources and/or removed by the cooling sources during additive manufacturing of the part.

[0070] In block 404, temperature history for each voxel of the 3D part is determined (e.g., predicted) based on the manufacturing parameters for additively manufacturing the 3D part as each voxel of each layer of the 3D part will have a different heating and/or cooling history based on the location of each voxel with respect to the energy sources and/or cooling sources within the build volume.

[0071] In block 406, the crystallinity is predicted for material of each voxel of the 3D part. In some examples, the crystallinity is predicted by a machine learning model trained to predict the crystallinity for material of each voxel based on the corresponding temperature history for each voxel. In order to train the machine learning model, the temperature for different locations of a training object (or training objects) at different times throughout the additive manufacturing process can be determined and the crystallinity for the different locations can be measured following additive manufacturing of the training object(s). The temperature for different locations of the part at different times throughout the additive manufacturing process is the input layer of the machine learning model and the resulting crystallinity for the different locations is the output layer of the machine learning model. The weights and biases of the machine learning model, including those of a hidden layer, are iteratively optimized by comparing the predicted crystallinity to the measured crystallinity to determine a loss function and backpropagating the loss function into the machine learning model for each iteration.

[0072] In block 408, the distribution of crystallinity of material of the 3D part can be displayed. For example, CAD and/or build preparation software can display the simulated 3D part with corresponding crystallinity for each voxel of the 3D part as a distribution of crystallinity in the simulated 3D part. The user can then choose to modify the manufacturing parameters to the part or manufacture the product through additive manufacturing.

[0073] Figure 5 is a flow diagram showing an example method 500 for mechanical property prediction from manufacturing parameters for additive manufacturing. The method 500 and/or an element or elements of the method 500 may be performed by an apparatus (e.g., electronic device). For example, the method 500 may be performed by the apparatus 102 described in relation to Figure 1. In block 502, manufacturing parameters for additively manufacturing a 3D part can be obtained. In block 504, temperature history for each voxel of the 3D part is determined (e.g., predicted) based on the manufacturing parameters for additively manufacturing the 3D part. In block 506, the crystallinity is predicted for material of each voxel of the 3D part.

[0074] In block 508, mechanical properties for each voxel can be determined. For example, the mechanical properties of the material for each voxel can be determined through a model that provides mechanical properties as a function of the crystallinity of the material, such as a viscoelastic-viscoplastic constitutive model as a function of crystallinity. In this regard, a viscoelastic-viscoplastic constitutive model as a function of crystallinity can provide mechanical properties, such as instantaneous elasticity (e.g., Young’s modulus), viscoelasticity (e.g., non-linear stress function), and viscoplasticity (e.g., hardening modulus), as a function of the crystallinity of the material. Any other material or mechanical properties that are based on the crystallinity of a material can be included in a model to relate the material and/or mechanical properties to the crystallinity of the material. [0075] In block 510, the distribution of mechanical properties of the 3D part can be displayed. For example, CAD and/or build preparation software can display the simulated 3D part with a corresponding mechanical property for each voxel of the 3D part as a distribution of the mechanical property in the simulated 3D part. The user can then choose to modify the manufacturing parameters to the part or manufacture the product through additive manufacturing.

[0076] Figure 6 is a flow diagram showing an example method 600 for mechanical behavior simulation based on material and mechanical property prediction from manufacturing parameters for additive manufacturing. The method 600 and/or an element or elements of the method 600 may be performed by an apparatus (e.g., electronic device). For example, the method 600 may be performed by the apparatus 102 described in relation to Figure 1. In block 602, manufacturing parameters for additively manufacturing a 3D part can be obtained. In block 604, temperature history for each voxel of the 3D part is determined (e.g., predicted) based on the manufacturing parameters for additively manufacturing the 3D part. In block 606, the crystallinity is predicted for material of each voxel of the 3D part. In block 608, mechanical properties for each voxel can be determined (e.g., through a viscoelastic-viscoplastic constitutive model as a function of crystallinity of the material).

[0077] In block 610, the mechanical behavior of the part can be simulated. For example, the mechanical properties, as determined based on the crystallinity of the material, for each voxel can be input into an FEA program (e.g., within a CAD and/or build preparation software) to determine mechanical behavior of the part, such as deformation behavior, failure criteria (e.g., elongation at break), loading, extension, fatigue, etc., in order to simulate the mechanical behavior of the part or portions of the part. The user can then choose to modify the manufacturing parameters to the part or manufacture the product through additive manufacturing.

[0078] Figure 7 is a flow diagram showing an example method 700 for mechanical property improvement based on mechanical property prediction from manufacturing parameters for additive manufacturing. The method 700 and/or an element or elements of the method 700 may be performed by an apparatus (e.g., electronic device). For example, the method 700 may be performed by the apparatus 102 described in relation to Figure 1. In block 702, manufacturing parameters for additively manufacturing a 3D part can be obtained. In block 704, temperature history for each voxel of the 3D part is determined (e.g., predicted) based on the manufacturing parameters for additively manufacturing the 3D part. In block 706, the crystallinity is predicted for material of each voxel of the 3D part. In block 708, mechanical properties for each voxel can be determined (e.g., through a viscoelastic-viscoplastic constitutive model as a function of crystallinity). In block 710, the mechanical behavior of the part can be simulated (e.g., through a FEA application).

[0079] In block 712, manufacturing parameters can be modified (e.g., manually by an end user or in an automated fashion) to improve material properties (e.g., crystallinity), mechanical properties, and/or mechanical behavior of the part based on a corresponding change to crystallinity of material of the 3D part. For example, if a material property, mechanical property, and/or mechanical behavior of the part is unacceptable to an end user (e.g., the 3D part or a portion of the 3D part is too weak), the 3D part can be modified using CAD and/or build preparation software to improve the part manually or in automated fashion.

[0080] As an example, a new geometry can be automatically generated for the part and presented for approval by an end user or an end user can manually modify the part to generate a new geometry for the part in order to improve a material property, mechanical property, and/or mechanical behavior of the part. As another example, a different location and/or a different orientation of the part in the build volume for additive manufacturing of the part can automatically be determined and presented for approval by an end user or an end user can manually modify the location and/or the orientation of the part in the build volume in order to improve a material property, mechanical property, and/or mechanical behavior of the part. As another example, different quantities of material (e.g., how thick of a layer of polymer powder, what contone level of or how much fusing agents and/or detailing agents, etc. will be applied for each voxel of each layer) can automatically be determined for each voxel during additive manufacturing of the part to present for approval by an end user or an end user can manually modify the quantities of material for each voxel in order to improve a material property, mechanical property, and/or mechanical behavior of the part. As another example, different energy settings for heating and/or cooling the materials during additive manufacturing can automatically be determined to present for approval by an end user or an end user can manually modify energy settings for heating and/or cooling in order to improve a material property, mechanical property, and/or mechanical behavior of the part.

[0081] In block 714, the mechanical behavior of the part can be re-simulated and the results can be displayed to the user. In this regard, the 3D part can be modified to improve the part manually or in automated fashion. The user can then choose to further modify the manufacturing parameters to the part or manufacture the product through additive manufacturing. For example, if the user determines the mechanical behavior, mechanical properties and/or material properties of the 3D part to be acceptable, the 3D part (and/or improvements to the 3D part) can be stored and/or additively manufactured.

[0082] While various examples of techniques are described herein, the techniques are not limited to the examples. Variations of the examples described herein may be implemented within the scope of the disclosure. For example, operations, functions, aspects, or elements of the examples described herein may be omitted or combined.

Claims

1. A non-transitory computer readable medium comprising instructions that, when executed by a processor, cause the processor to execute a machine learning model trained to predict crystallinity of build material based on manufacturing parameters during additive manufacturing, and further comprising instructions that, when executed by a processor, cause the processor to: predict, by the machine learning model, a crystallinity for at least one location of a plurality of locations of an object based on corresponding manufacturing parameters for the object; and compute a mechanical property for the at least one location of the object based on the crystallinity for the at least one location of the plurality of locations of the object.

2. The non-transitory computer readable medium of claim 1 , wherein the corresponding manufacturing parameters for the object comprise at least one of a build volume with energy sources, energy source settings, a location of the object in the build volume, an orientation of the object in the build volume, a geometry of the object, an amount of fusing agents for the object, and an amount of detailing agents for the object.

3. The non-transitory computer readable medium of claim 1, further comprising instructions that, when executed by a processor, cause the processor to: predict a temperature history for the at least one location of the plurality of locations of the object based on the corresponding manufacturing parameters for the object during additive manufacturing of the object; and predict, by the machine learning model, the crystallinity for the at least one location of the plurality of locations of the object based on the temperature history for the at least one location of the plurality of locations of the object.

4. The non-transitory computer readable medium of claim 1, wherein the machine learning model is trained to predict crystallinity based on temperature history of the build material during additive manufacturing and further comprising instructions that, when executed by a processor, cause the processor to: train the machine learning model to predict crystallinity of build material based on temperature history of the build material during additive manufacturing by: determining a temperature history for each location of a plurality of locations of a manufactured object; measuring a crystallinity for each location of the plurality of locations of the manufactured object; and training a hidden layer of the machine learning model with the temperature history for each location of the plurality of locations of the manufactured object as input and the crystallinity for each location of the plurality of locations of the manufactured object as output.

5. The non-transitory computer readable medium of claim 1, further comprising instructions that, when executed by a processor, cause the processor to: use a viscoelastic-viscoplastic constitutive model as a function of crystallinity of the build material to compute the mechanical property for the at least one location of the object.

6. The non-transitory computer readable medium of claim 5, wherein the viscoelastic-viscoplastic constitutive model comprises instantaneous elastic modulus, relaxation modulus, and viscoplastic deformation resistance as a function of crystallinity of the build material.

7. A computer-implemented method comprising: receiving manufacturing parameters for additive manufacturing of an object; predicting a crystallinity for each location of a plurality of locations of the object based on the manufacturing parameters for the object; and simulating a mechanical property of the object based on the crystallinity predicted for each location of the plurality of locations of the object.

8. The computer-implemented method of claim 7, wherein the manufacturing parameters for additive manufacturing of the object comprise at least one of a build volume with energy sources, energy source settings, a location of the object in the build volume, an orientation of the object in the build volume, a geometry of the object, an amount of fusing agents for the object, and an amount of detailing agents for the object.

9. The computer-implemented method of claim 7, further comprising: predicting a temperature history for each location of the plurality of locations of the object based on the manufacturing parameters for additive manufacturing of the object; and predicting, by a machine learning model, the crystallinity for each location of the plurality of locations of the object based on the temperature history for each location of the plurality of locations of the object.

10. The computer-implemented method of claim 7, wherein simulating the mechanical property of the object based on the crystallinity predicted for each location of the plurality of locations of the object further comprises: computing the mechanical property for each location of the plurality of locations of the object based on the crystallinity for each location of the plurality of locations of the object, wherein the mechanical property is at least one of strain, stress, and Young’s modulus; and computing a distribution of the material property based on the mechanical property computed for each location of the plurality of locations of the object.

1 1. The computer-implemented method of claim 7, further comprising: simulating mechanical behavior of the object based on the mechanical property of the object by: computing the mechanical property for each location of the plurality of locations of the object based on the crystallinity for each location of the plurality of locations of the object; and computing mechanical behavior of the object based on the mechanical property computed for each location of the plurality of locations of the object, wherein the mechanical behavior of the object is at least one of loading, extension, and fatigue.

12. The computer-implemented method of claim 7, further comprising: modifying the manufacturing parameters for the object to improve the object by: determining a change in crystallinity for at least one location of the plurality of locations of the object to improve the object; and modifying the manufacturing parameters based on the change in crystallinity for the at least one location of the plurality of locations of the object, wherein the modification to the manufacturing parameters includes at least one of modifying geometry of the object, modifying a location of the object in a build volume, modifying an orientation of the object in the build volume, modifying quantities of fusing agents for additive manufacturing of the object, modifying quantities detailing agents for additive manufacturing of the object, and modifying settings for energy sources for additive manufacturing of the object.

13. A computing system comprising: a processor; and a non-transitory computer-readable medium having stored thereon instructions that when executed by the processor, cause the processor to perform operations including: predicting a crystallinity for at least one location of a plurality of locations of an object based on manufacturing parameters for the object; computing a mechanical property for the at least one location of the plurality of locations of the object based on its crystallinity; and simulating, based on the mechanical property for the at least one location of the plurality of locations of the object, mechanical behavior of the object.

14. The system of claim 13, wherein the instructions that when executed by the processor, cause the processor to perform operations further including: predicting a temperature history for each location of the plurality of locations of the object based on the manufacturing parameters for additive manufacturing of the object; and predicting, by a machine learning model, the crystallinity for each location of the plurality of locations of the object based on the temperature history for each location of the plurality of locations of the object.

15. The system of claim 13, wherein the instructions that when executed by the processor, cause the processor to perform operations further including: causing at least one of displaying of results of simulating the mechanical behavior of the object and 3D printing of the object.