Predicting process control parameters for fabricating an object using deposition

Abstract

Process control parameters are predicted to fabricate an object using deposition. An input design geometry is provided for the object. A training data set includes past post-build physical inspection data for a plurality of objects that comprise at least one object that is different from the object to be physically fabricated; and training data generated through a repetitive process of randomly choosing values for each of multiple process control parameters and scoring adjustments to the multiple process control parameters as leading to either undesirable or desirable outcomes, the outcomes based respectively on the presence or absence of defects detected in a fabricated object arising from the process control parameter adjustments. A machine learning algorithm is trained using the provided training data set and a predicted optimal set of the multiple process control parameters is generated for initiating and performing the deposition process to fabricate the object.

What is claimed is: 
     
         1 . A method for generating a predicted optimal set of deposition process control parameters to fabricate an object having an input design geometry, the method comprising:
 providing an input design geometry for an object to be physically fabricated using the deposition process;   providing a training data set that comprises:   past post-build physical inspection data for a plurality of objects that comprise at least one object that is different from the object to be physically fabricated; and   training data generated through a repetitive process of randomly choosing values for each of multiple process control parameters and scoring adjustments to the multiple process control parameters as leading to either undesirable or desirable outcomes, the outcomes based respectively on the presence or absence of defects detected in a fabricated object arising from the process control parameter adjustments;   training a machine learning algorithm using the provided training data set; and   generating a predicted optimal set of the multiple process control parameters for initiating and performing the deposition process to fabricate the object, wherein the predicted optimal set of the multiple process control parameters are derived using the trained machine learning algorithm.   
     
     
         2 . The method of  claim 1 , wherein training the machine learning algorithm further comprises randomly choosing values within a specified range of a process window, generating process simulation data with a tool to simulate a fabrication process, using the randomly chosen values and incorporating the generated process simulation data, into the training data set to improve a learned model that maps process control parameter values to process outcomes. 
     
     
         3 . The method of  claim 1 , wherein providing the training data set further comprises providing process simulation data, process characterization data, or in-process inspection data for a plurality of design geometries or portions thereof. 
     
     
         4 . The method of  claim 1 , wherein providing the training data set further comprises providing process characterization data, in-process inspection data, or past post-build physical inspection data that is generated by a skilled operator while manually adjusting one or more of the multiple the process control parameters. 
     
     
         5 . The method of  claim 1 , further comprising:
 receiving data for multiple object properties from each of multiple sensors as the object is being physically fabricated,   wherein providing the training data set further comprises providing the data, and   wherein generating a predicted optimal set of the multiple process control parameters comprises adjusting one or more of the multiple process control parameters as the object is being physically fabricated.   
     
     
         6 . The method of  claim 5 , further comprising removing noise from the data prior to providing the data. 
     
     
         7 . The method of  claim 5 , wherein the data comprises acoustic energy or mechanical energy that is reflected, scattered, absorbed, transmitted, or emitted by the object. 
     
     
         8 . The method of  claim 1 , wherein the machine learning algorithm comprises an artificial neural network. 
     
     
         9 . The method of  claim 1 , wherein the defects detected in the fabricated object are detected as differences between object property data and a reference data set that are larger than a specified threshold, and are classified using a one-class support vector machine and a training data set that comprises object property data for defective and defect-free objects. 
     
     
         10 . The method of  claim 1 , further comprising providing instructions to an apparatus to perform a free form deposition process to fabricate the object, wherein generating a predicted optimal set of the multiple process control parameters further comprises adjusting the one or more process control parameters while the apparatus is physically performing the free form deposition process. 
     
     
         11 . A machine-readable storage medium having instructions therein which when executed by the machine cause the machine to perform operations comprising:
 receiving an input design geometry for an object to be physically fabricated by a deposition process using a predicted optimal set of deposition process control parameters;   receiving a training data set that comprises:   past post-build physical inspection data for a plurality of objects that comprise at least one object that is different from the object to be physically fabricated; and   training data generated through a repetitive process of randomly choosing values for each of multiple process control parameters and scoring adjustments to the multiple process control parameters as leading to either undesirable or desirable outcomes, the outcomes based respectively on the presence or absence of defects detected in a fabricated object arising from the process control parameter adjustments;   training a machine learning algorithm using the provided training data set; and   generating a predicted optimal set of the multiple process control parameters for initiating and performing the deposition process to fabricate the object, wherein the predicted optimal set of the multiple process control parameters are derived using the trained machine learning algorithm.   
     
     
         12 . The machine-readable storage medium of  claim 11 , wherein training the machine learning algorithm further comprises randomly choosing values within a specified range of a process window, generating process simulation data to simulate a fabrication process, using the randomly chosen values and incorporating the generated process simulation data, into the training data set to improve a learned model that maps process control parameter values to process outcomes. 
     
     
         13 . The machine-readable storage medium of  claim 12 , wherein generating process simulation data comprises generating process simulation data using a finite element analysis. 
     
     
         14 . A system for controlling a deposition process, the system comprising:
 a deposition apparatus configured to fabricate an object using a deposition process and based on an input design geometry;   one or more sensors configured to characterize the deposition process of the deposition apparatus, wherein the one or more sensors provide data for one or more process parameters or object properties; and   a processor programmed to:   provide a predicted optimal set of one or more input process control parameters to control parameters of the deposition process during fabrication of the object, wherein the predicted optimal set of the one or more input process control parameters are derived using a machine learning algorithm that has been trained using a training data set;   receive the data from the one or more sensors as input to the machine learning algorithm;   detect defects in the object during fabrication of the object using the data from the one or more sensors;   classify the detected object defects in real time using the machine learning algorithm;   provide the classification of detected object defects as input to the machine learning algorithm;   adjust one or more input process control parameters based on the classification of the detected object defects, wherein the adjustments are derived using the machine learning algorithm; and   provide instructions to the deposition apparatus to adjust the one or more input process control parameters during fabrication of the object.   
     
     
         15 . The system of  claim 14 , wherein the processor is further programmed to perform iteratively and incorporate the detected object defect classifications for each iteration into the training data set. 
     
     
         16 . The system of  claim 14 , implemented as a distributed, modular system comprising a first deposition apparatus, a first sensor, and a first processor, wherein the first deposition apparatus, the first sensor, and the first processor are configured to share training data and process characterization data with a second processor via a network. 
     
     
         17 . The system of  claim 15 , wherein the one or more sensors comprise a laser interferometer. 
     
     
         18 . The system of  claim 15 , wherein the object defects are detected as differences between object property data and a reference data set that are larger than a specified threshold, and are classified using an autoencoder algorithm. 
     
     
         19 . The system of  claim 15 , wherein the training data set further comprises process characterization data, in-process inspection data, or post-build inspection data that is generated by an operator while manually adjusting the one or more process control parameters. 
     
     
         20 . A method comprising:
 providing a predicted optimal set of one or more input process control parameters to control parameters of a deposition fabrication process during fabrication of an object, wherein the predicted optimal set of the one or more input process control parameters are derived using a machine learning algorithm that has been trained using a training data set;   receiving data characterizing the deposition process using properties of the object being fabricated from one or more sensors as input to the machine learning algorithm;   detecting defects in the object during fabrication of the object using the data from the one or more sensors;   classifying the detected object defects in real time using the machine learning algorithm;   providing the classification of detected object defects as input to the machine learning algorithm;   adjusting one or more input process control parameters based on the real-time classification of the detected object defects, wherein the adjustments are derived using the machine learning algorithm; and   providing instructions to adjust the one or more input process control parameters during fabrication of the object.   
     
     
         21 . The method of  claim 20 , further comprising:
 providing a training data set that comprises:   past post-build physical inspection data for a plurality of objects that comprise at least one object that is different from the object to be physically fabricated; and   training data generated through a repetitive process of randomly choosing values for each of multiple process control parameters and scoring adjustments to the multiple process control parameters as leading to either undesirable or desirable outcomes, the outcomes based respectively on the presence or absence of defects detected in a fabricated object arising from the process control parameter adjustments; and   training the machine learning algorithm using the provided training data set.   
     
     
         22 . A method for adaptive control of a free form deposition process, the method comprising:
 providing an input design geometry for an object to be physically fabricated using the free form deposition process;   providing a training data set that comprises:   process characterization data for a plurality of objects that comprise at least one object that is different from the object to be physically fabricated; and   training data generated through a repetitive process of randomly choosing values for each of multiple process control parameters and scoring adjustments to the multiple process control parameters as leading to either undesirable or desirable outcomes, the outcomes based respectively on the presence or absence of defects detected in a fabricated object arising from the process control parameter adjustments;   providing a predicted optimal set of the multiple process control parameters for initiating the free form deposition process, wherein the predicted optimal set of the multiple process control parameters are derived using a machine learning algorithm that has been trained using the training data set;   providing a real-time classification of detected object defects using the machine learning algorithm that has been trained using the training data, wherein real-time data from multiple sensors is provided as input to the machine learning algorithm, and wherein the real-time classification of detected object defects is output from the machine learning algorithm; and   providing instructions to perform the free form deposition process to fabricate the object, wherein the machine learning algorithm adjusts the multiple process control parameters while physically performing the free form deposition process.