# US 12,601,713 B2 · Wire bonding defect detection apparatus and operation method thereof > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,601,713 B2 - Document ID: US12601713B2 - Status: Granted - Kind: B2 - Application number: 18/281,718 - Filed: 2022-08-24 - Published: 2024-05-16 - Granted: 2026-04-14 - Earliest claimed: 2021-08-25 - Assignees: LG Energy Solution, Ltd. - Applicants: LG Energy Solution, Ltd. - Inventors: Jee Hoon Choi - Web page: https://patentgrep.com/patent/US12601713B2 ## CPC - G01N 29/069: {Defect imaging, localisation and sizing using, e.g. time of flight diffraction [TOFD], synthetic aperture focusing technique [SAFT], Amplituden-Laufzeit-Ortskurven [ALOK] technique} - H10W 72/0711: {Apparatus therefor} - B23K 20/10: making use of vibrations, e.g. ultrasonic welding - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/048: Activation functions - G06N 3/08: Learning methods - H01M 50/516: by welding, soldering or brazing - G01N 2291/023: Solids - G01N 2291/0289: Internal structure, e.g. defects, grain size, texture - G01N 2291/267: Welds - Y02E 60/10: Energy storage using batteries - G01N: INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES (measuring or testing apparatus or processes other than immunoassay, involving enzymes or microorganisms C12M, C12Q) - H10W: GENERIC PACKAGES, INTERCONNECTIONS, CONNECTORS OR OTHER CONSTRUCTIONAL DETAILS OF DEVICES COVERED BY CLASS H10 - 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 - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - H01M: PROCESSES OR MEANS, e.g. BATTERIES, FOR THE DIRECT CONVERSION OF CHEMICAL ENERGY INTO ELECTRICAL ENERGY - Y02E: REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION ## Abstract An apparatus and method are disclosed for detecting a bonding defect occurring during an ultrasonic wire bonding process between a battery cell and a busbar connected to each other by ultrasonic wire bonding. A bonding parameter collector, coupled to an ultrasonic wire bonding machine, may collect ultrasonic bonding parameters continuously during the ultrasonic wire bonding process, machine learning training may be performed by a convolutional neural network, and a bonding defect may be detected based thereon. ## Claims ### Claim 1 An apparatus for detecting a bonding defect occurring during an ultrasonic wire bonding process between a battery cell and a busbar connected to each other by ultrasonic wire bonding, the apparatus comprising: a bonding parameter collector configured to continuously collect a plurality of ultrasonic bonding parameters; a data generator configured to generate data by applying wavelet transform to the plurality of ultrasonic bonding parameters so as to convert the plurality of ultrasonic bonding parameters into images having a two-dimensional array; a convolutional neural network configured to perform machine-learning training using the data generated by the data generator; and a defect determination component configured to determine a defect due to the ultrasonic wire bonding using the convolutional neural network that is trained, wherein the plurality of ultrasonic bonding parameters comprise a measured value of wire deformation, and wherein the measured value of the wire deformation includes a measured value of a height of a wire changed as the wire is melted. ### Claim 2 (depends on 1) The apparatus according to claim 1, wherein the plurality of ultrasonic bonding parameters comprise an ultrasonic generator current and an ultrasonic generator voltage. ### Claim 3 (depends on 1) The apparatus according to claim 1, wherein the wavelet transform is for converting the plurality of ultrasonic bonding parameters into a value of 0 to 1 through minimum-maximum scaling and generating training data having the two-dimensional array through a scale value of 2 or more using a Mexican Hat waveform transform function. ### Claim 4 (depends on 1) The apparatus according to claim 1, wherein the convolutional neural network comprises four convolution tiers, a gap tier, and a fully-connected tier. ### Claim 5 (depends on 1) The apparatus according to claim 1, wherein the ultrasonic bonding parameters are collected for a time of 10 ms to 1000 ms. ### Claim 6 (depends on 1) The apparatus according to claim 1, wherein the bonding parameter collector is configured to collect the plurality of ultrasonic bonding parameters at intervals of 0.1 ms to 100 ms. ### Claim 7 (depends on 1) A battery pack configured to have a welding defect between the battery cell and the busbar detected using the apparatus according to claim 1. ### Claim 8 (depends on 7) A device comprising the battery pack according to claim 7. ### Claim 9 (depends on 1) The apparatus according to claim 1, further comprising: the battery cell; the busbar; and an ultrasonic wire bonding machine coupled to the bonding parameter collector, wherein the ultrasonic wire bonding machine is configured to perform the ultrasonic wire bonding process on the battery cell and the busbar to have the battery cell and the busbar connected to each other by the ultrasonic wire bonding. ### Claim 10 (depends on 9) The apparatus according to claim 9, wherein the bonding parameter collector or the data generator is configured to sample the plurality of ultrasonic bonding parameters, and wherein the ultrasonic wire bonding machine is configured to provide the plurality of ultrasonic bonding parameters to the bonding parameter collector continuously in real time during the ultrasonic wire bonding process. ### Claim 11 (depends on 1) The apparatus according to claim 1, wherein the bonding parameter collector is configured to collect, continuously in real time during the ultrasonic wire bonding process, the measured value of the height of the wire as the wire is melted. ### Claim 12 A method of detecting a bonding defect occurring during an ultrasonic wire bonding process between a battery cell and a busbar connected to each other by ultrasonic wire bonding, the method comprising: continuously collecting a plurality of ultrasonic bonding parameters; generating a training data by converting the plurality of ultrasonic bonding parameters into images having a two-dimensional array using wavelet transform; training a convolutional neural network with the generated training data; and determining a defect due to the ultrasonic wire bonding using the convolutional neural network, wherein the plurality of ultrasonic bonding parameters comprise a measured value of wire deformation, and wherein the measured value of the wire deformation includes a measured value of a height of a wire changed as the wire is melted. ### Claim 13 (depends on 12) The method according to claim 12, wherein the wavelet transform converts the plurality of ultrasonic bonding parameters into a value of 0 to 1 through minimum-maximum scaling and generates the training data having the two-dimensional array through a scale value of 2 or more using a Mexican Hat waveform transform function. ### Claim 14 (depends on 12) The method according to claim 12, wherein the generating of the training data comprises performing convolution calculation with respect to the images having the two-dimensional array using a 2×2 or 3×3 filter and performing calculation through an activation function and pooling to generate 256 one-dimensional data. ### Claim 15 (depends on 12) The method according to claim 12, wherein the plurality of ultrasonic bonding parameters comprise an ultrasonic generator current and an ultrasonic generator voltage. ### Claim 16 (depends on 7) The method according to claim 7, wherein the method is performed using an apparatus, wherein the apparatus comprises: the battery cell; the busbar; an ultrasonic wire bonding machine; and a bonding parameter collector coupled to the ultrasonic wire bonding machine, wherein the ultrasonic wire bonding machine performs the ultrasonic wire bonding process on the battery cell and the busbar to have the battery cell and the busbar connected to each other by the ultrasonic wire bonding, and wherein the bonding parameter collector performs continuously collecting the plurality of ultrasonic bonding parameters. ### Claim 17 (depends on 16) The method according to claim 16, wherein the apparatus further comprises a data generator, wherein the bonding parameter collector or the data generator samples the plurality of ultrasonic bonding parameters, and wherein the ultrasonic wire bonding machine provides the plurality of ultrasonic bonding parameters to the bonding parameter collector continuously in real time during the ultrasonic wire bonding process.