# US 12,118,475 B2 · Computer modeling for detection of discontinuities and remedial actions in fastening systems > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,118,475 B2 - Document ID: US12118475B2 - Status: Granted - Kind: B2 - Application number: 17/734,891 - Filed: 2022-05-02 - Published: 2022-08-18 - Granted: 2024-10-15 - Earliest claimed: 2021-04-12 - Assignees: Newfrey LLC - Applicants: Newfrey LLC - Inventors: Amir Reza Kashani Pour; Maxim Dalton - Web page: https://patentgrep.com/patent/US12118475B2 ## CPC - G06F 7/544: for evaluating functions by calculation {(G06F7/4824 takes precedence)} - G06N 5/022: Knowledge engineering; Knowledge acquisition - B23K 9/0953: {using computing means} - B23K 9/201: {of the extremity of a small piece on a large basis} - B23K 31/006: {relating to using of neural networks} - B23K 31/125: {Weld quality monitoring} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - B23K: SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM ## Abstract Disclosed herein are systems and methods for identifying welding anomalies and discontinuities in stud welding using AI models. Instead of conventional welding accuracy methods (e.g. destructive and/or image generation methods) a processor may communicate with one or more sensors associated with a joining machine to retrieve joining data and attributes. The processor may then execute an AI model that is trained based on previously performed stud welding, their corresponding welding attributes, and their corresponding discontinuities and/or anomalies. The processor may execute the AI model using data retrieved from the sensors and may calculate a likelihood of a discontinuity and discontinuity attributes, such as, location, depth, and the like. The processor may also execute a second AI model to identify an appropriate course of action to remedy the identified/predicted discontinuity. ## Claims ### Claim 1 A method of training a neural network to predict a likelihood of at least one discontinuity or anomaly being present within a joining created in a welding system, comprising: receiving, by a processor, a plurality of attributes corresponding to the joining; generating, by the processor, a training dataset based on the plurality of attributes corresponding to the joining and a plurality of attributes corresponding to previous joinings having at least one discontinuity or anomaly, the plurality of attributes corresponding to previous joinings comprising at least a voltage value, a current value, and a lift value for each previous joining, the lift value indicating a distance a weld stud is raised above a base material during the previous joining; training, by the processor, the neural network using the training dataset, the neural network trained to predict the likelihood of the at least one discontinuity or anomaly and respective likelihoods of potential causes of the at least one discontinuity or anomaly; executing, by the processor, the neural network using a plurality of second attributes corresponding to a second joining as input, the neural network generating one or more attributes associated with at least one second discontinuity or anomaly of the second joining, each of the one or more attributes identifying a respective likelihood of a respective physical condition of the welding system causing the at least one second discontinuity or anomaly of the second joining; and transmitting, by the processor, to a joining machine, an instruction to modify at least one joining parameter of its welding based on the one or more attributes generated by the neural network. ### Claim 2 (depends on 1) The method of claim 1, wherein the plurality of attributes corresponding to previous joinings further comprise at least one of metal sheet compositional attributes, metal sheet geometric attributes, stud compositional attributes, or stud geometric attributes. ### Claim 3 (depends on 1) The method of claim 1, wherein the neural network calculates a depth and a type associated with each discontinuity or anomaly. ### Claim 4 (depends on 1) The method of claim 1, wherein the joining is a stud welding. ### Claim 5 (depends on 1) The method of claim 1, wherein the joining is performed via a rivet welding method. ### Claim 6 (depends on 1) The method of claim 1, further comprising: executing, by the processor, a second neural network that receives, as input, the one or more attributes associated with the at least one discontinuity or anomaly generated by the neural network and generates a remedial action associated with the at least one discontinuity or anomaly, wherein the second neural network was trained based on one or more attributes associated with at least one previous discontinuity or anomaly corresponding to one or more previous joinings and one or more corresponding remedial actions; and presenting, by the processor, for display, the remedial action generated by the second neural network. ### Claim 7 (depends on 6) The method of claim 6, wherein the remedial action generated by the second neural network is a corrective or preventative action. ### Claim 8 (depends on 1) The method of claim 1, further comprising presenting, by the processor, at a user interface, a table displaying, in a column corresponding to the second joining, the respective likelihood of the respective physical condition of the welding system causing the at least one second discontinuity or anomaly for each of the one or more attributes generated by the neural network. ### Claim 9 (depends on 1) The method of claim 1, wherein the at least one joining parameter comprises at least one of a voltage, current, or lift of the joining machine. ### Claim 10 (depends on 1) The method of claim 1, wherein the instruction causes the joining machine to cease welding. ### Claim 11 (depends on 1) The method of claim 1, further comprising: presenting, by the processor, at a user interface, the one or more attributes associated with at least one discontinuity or anomaly of the second joining; and training, by the processor, the neural network based on the one or more attributes and input to the user interface. ### Claim 12 A system for training neural network to predict a likelihood of at least one discontinuity or anomaly being present within a joining created in a welding system, comprising: a server comprising a processor and a non-transitory computer-readable medium containing instructions that when executed by the processor causes the processor to perform operations comprising: receiving a plurality of attributes corresponding to the joining; generating a training dataset based on the plurality of attributes corresponding to the joining and a plurality of attributes corresponding to previous joinings having at least one discontinuity or anomaly, the plurality of attributes corresponding to previous joinings comprising at least a voltage value, a current value, and a lift value for each previous joining, the lift value indicating a distance a weld stud is raised above a base material during the previous joining; training the neural network using the training dataset, the neural network trained to predict the likelihood of the at least one discontinuity or anomaly and respective likelihoods of potential causes of the at least one discontinuity or anomaly; executing the neural network using a plurality of second attributes corresponding to a second joining as input, the neural network generating one or more attributes associated with at least one second discontinuity or anomaly of the second joining, each of the one or more attributes identifying a respective likelihood of a respective physical condition of the welding system causing the at least one second discontinuity or anomaly of the second joining; and transmitting, to a joining machine, an instruction to modify at least one joining parameter of its welding based on the one or more attributes generated by the neural network. ### Claim 13 (depends on 12) The system of claim 12, wherein the plurality of attributes corresponding to previous joinings further comprise at least one of metal sheet compositional attributes, metal sheet geometric attributes, stud compositional attributes, or stud geometric attributes. ### Claim 14 (depends on 12) The system of claim 12, wherein the neural network calculates a depth and a type associated with each discontinuity or anomaly. ### Claim 15 (depends on 12) The system of claim 12, wherein the joining is a stud welding. ### Claim 16 (depends on 12) The system of claim 12, wherein the joining is performed via a rivet welding method. ### Claim 17 (depends on 12) The system of claim 12, wherein the instructions further cause the processor to: execute a second neural network that receives, as input, the one or more attributes associated with the at least one discontinuity or anomaly generated by the neural network and generates a remedial action associated with the at least one discontinuity or anomaly, wherein the second neural network was trained based on one or more attributes associated with at least one previous discontinuity or anomaly corresponding to one or more previous joinings and one or more corresponding remedial actions; and present, for display, the remedial action generated by the second neural network. ### Claim 18 (depends on 17) The system of claim 17, wherein the remedial action generated by the second neural network is a corrective or preventative action. ### Claim 19 A system for training neural network to predict a likelihood of at least one discontinuity or anomaly being present within a joining created in a welding system, comprising: a joining machine; and a processor in communication with the joining machine, the processor configured to: receive a plurality of attributes corresponding to the joining; generate a training dataset based on the plurality of attributes corresponding to the joining and a plurality of attributes corresponding to previous joinings having at least one discontinuity or anomaly, the plurality of attributes corresponding to previous joinings comprising at least a voltage value, a current value, and a lift value for each previous joining, the lift value indicating a distance a weld stud is raised above a base material during the previous joining; train the neural network using the training dataset, the neural network trained to predict the likelihood of the at least one discontinuity or anomaly and respective likelihoods of potential causes of the at least one discontinuity or anomaly; execute the neural network using a plurality of second attributes corresponding to a second joining as input, the neural network generating one or more attributes associated with at least one second discontinuity or anomaly of the second joining, each of the one or more attributes identifying a respective likelihood of a respective physical condition of the welding system causing the at least one second discontinuity or anomaly of the second joining; transmit, to the joining machine, an instruction to modify at least one joining parameter of its welding based on the one or more attributes generated by the neural network. ### Claim 20 (depends on 19) The system of claim 19, wherein the processor is further configured to: execute a second neural network that receives, as input, the one or more attributes associated with the at least one discontinuity or anomaly generated by the neural network and generates a remedial action associated with the at least one discontinuity or anomaly, wherein the second neural network was trained based on one or more attributes associated with at least one previous discontinuity or anomaly corresponding to one or more previous joinings and one or more corresponding remedial actions; and present, for display, the remedial action generated by the second neural network.