# US 2022/0227071 A1 · Ultrasonic setting of a connector to an object

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2022/0227071 A1
- Document ID: US20220227071A1
- Status: Published application
- Kind: A1
- Application number: 17/613,308
- Filed: 2020-05-22
- Published: 2022-07-21
- Earliest claimed: 2019-05-24
- Assignees: Multimaterial-Welding AG
- Applicants: Multimaterial-Welding AG
- Inventors: Gregor ECKHARD; Valentin BERSIER
- Web page: https://patentgrep.com/patent/US20220227071A1

## CPC

- B29C 65/08: using ultrasonic vibrations {(non-plastics element to plastics elements B29C65/645)}
- B29C 66/9515: {by measuring their vibration amplitude}
- B23K 20/10: making use of vibrations, e.g. ultrasonic welding
- B29C 65/02: by heating, with or without pressure
- B29C 65/06: using friction, e.g. spin welding {(non-plastics elements to plastic elements B29C65/645)}
- B29C 65/16: Laser beams
- B29C 65/8215: {Tensile tests}
- B29C 66/30326: {in the form of porosity}
- B29C 66/474: {said single elements being substantially non-flat}
- B29C 66/7392: {characterised by the material of at least one of the parts being a thermoplastic}
- B29C 66/9221: {by measuring the pressure, the force or the mechanical power}
- B29C 66/9231: {by measuring the displacement of the joining tools}
- B29C 66/932: {by measuring the speed}
- B29C 66/9511: {by measuring their vibration frequency}
- B29C 66/963: {using stored or historical data sets, e.g. using expert systems}
- B29C 66/964: {involving trial and error}
- B29C 66/965: {using artificial neural networks}
- G05B 13/0265: {the criterion being a learning criterion}
- G05B 13/027: {using neural networks only}
- B29C 66/532: {Joining single elements to the wall of tubular articles, hollow articles or bars}
- B29C 66/721: {Fibre-reinforced materials (B29C66/729 takes precedence)}
- B29C 66/72525: {comprising honeycomb cores}
- B29C 66/727: {being porous, e.g. foam}
- B29C 66/729: {Textile or other fibrous material made from plastics}
- B29C 66/7487: {Wood}
- B29L 2031/3055: {Cars}
- B29L 2031/3076: {Aircrafts (blades, propellers B29L2031/08)}
- B29C: SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING (making preforms B29B11/00; making laminated products by combining previously unconnected layers which become one product whose layers will remain together B32B37/00 - B32B41/00)
- 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
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- B29L: INDEXING SCHEME ASSOCIATED WITH SUBCLASS B29C, RELATING TO PARTICULAR ARTICLES

## Abstract

A computer implemented method comprises the steps of: providing a user interface to a computer terminal; providing a welding machine interface (252) to a welding machine (22; 31) which is equipped with a set of sensors having a power supply sensor (221; 311) configured to sense a power supplied by the welding machine (22; 31) to set a connector to an object in runtime; obtaining a threshold performance metric data signal representing threshold product performance metric predefined via the user interface; obtaining a power supply data signal from the welding machine (22; 31) via the welding machine interface (252), which power supply data signal represents the sensed power supplied by the welding machine (22; 31) to set the connector to the object; applying a machine learning model to the power represented by the obtained power supply data signal such that the machine learning model calculates a model product performance metric, wherein the machine learning model is specifically pre trained with training power sensed by the power supply sensor (221; 311) of the set of sensors of the welding machine (22; 31) and measured product performance metrics; comparing the calculated model product performance metric to the threshold product performance metric represented by the threshold performance metric data signal; and generating a non-consistency data signal when the calculated product performance metric does not comply with the threshold product performance metric.

## Claims

### Claim at position 1

-36. (canceled)

### Claim 37

A computer implemented method, comprising the steps of
providing a user interface to a computer terminal;
providing a welding machine interface to a welding machine which is equipped with a set of sensors having a power supply sensor configured to sense a power supplied by the welding machine to set a connector to an object in runtime;
obtaining a threshold performance metric data signal representing threshold product performance metric predefined via the user interface;
obtaining a power supply data signal from the welding machine via the welding machine interface, which power supply data signal represents the sensed power supplied by the welding machine to set the connector to the object;
applying a machine learning model to the power represented by the obtained power supply data signal such that the machine learning model calculates a model product performance metric, wherein the machine learning model is specifically pre-trained with training power sensed by the power supply sensor of the set of sensors of the welding machine and measured product performance metric;
comparing the calculated model product performance metric to the threshold product performance metric represented by the threshold performance metric data signal; and
generating a non-consistency data signal when the calculated product performance metric does not comply with the threshold product performance metric.

### Claim 38 (depends on 37)

The computer implemented method of claim 37, comprising a step of providing a handling machine interface to a handling machine which is configured to handle the object with the set connector, and transferring the non-consistency data signal to the handling machine via the handling machine interface.

### Claim 39 (depends on 37)

The computer implemented method of claim 37, wherein the set of sensors of the welding machine has a force sensor configured to sense a push force applied to the connector while being set to the object in runtime, the computer implemented method comprising the steps of:
obtaining a push force data signal from the welding machine via the welding machine interface, which push force data signal represents the sensed push force applied to the connector while being set to the object; and
applying the machine learning model to the push force represented by the obtained push force data signal such that the machine learning model involves the push force to calculate the model product performance metric, wherein the machine learning model is specifically pre-trained with training push force sensed by the force sensor of the set of sensors of the welding machine.

### Claim 40 (depends on 37)

The computer implemented method of claim 37, wherein the set of sensors of the welding machine has a distance sensor configured to sense a distance the welding machine forwards the connector to the object in runtime, the computer implemented method comprising the steps of:
obtaining a distance data signal from the welding machine via the welding machine interface, which distance data signal represents the sensed distance the welding machine forwards the connector to the object; and
applying the machine learning model to the distance represented by the obtained distance data signal such that the machine learning model involves the distance to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training distance sensed by the distance sensor of the set of sensors of the welding machine.

### Claim 41 (depends on 37)

The computer implemented method of claim 37, wherein the set of sensors of the welding machine has a velocity sensor configured to sense a velocity by which the welding machine forwards the connector to the object in runtime, the computer implemented method comprising the steps of:
obtaining a velocity data signal from the welding machine via the welding machine interface, which velocity data signal represents the sensed velocity by which the welding machine forwards the connector to the object; and
applying the machine learning model to the velocity represented by the obtained velocity data signal such that the machine learning model involves the velocity to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training velocity sensed by the velocity sensor of the set of sensors of the welding machine.

### Claim 42 (depends on 37)

The computer implemented method of claim 37, comprising the steps of:
determining a signal amplitude of an ultrasonic signal supplied by the welding machine to set the connector to the object and represented by a signal amplitude data signal; and
applying the machine learning model to the determined signal amplitude such that the machine learning model involves the signal amplitude to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with a training signal amplitude determined by the power supply sensor of the set of sensors of the welding machine.

### Claim 43 (depends on 37)

The computer implemented method of claim 37, comprising the steps of:
determining a signal frequency of an ultrasonic signal supplied by the welding machine to set the connector to the object and represented by a frequency data signal; and
applying the machine learning model to the determined signal frequency such that the machine learning model involves the signal frequency to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training signal frequency determined by the power supply sensor of the set of sensors of the welding machine.

### Claim 44 (depends on 37)

The computer implemented method of claim 37, wherein the set of sensors of the welding machine has an acoustic emission sensor configured to sense an acoustic emission generated when the welding machine sets the connector to the object in runtime, the computer implemented method comprising the steps of:
obtaining an acoustic emission data signal from the welding machine via the welding machine interface, which acoustic emission data signal represents the sensed acoustic emission generated when the welding machine sets the connector to the object, and applying the machine learning model to the sensed acoustic emission represented by the obtained acoustic emission data signal such that the machine learning model involves the acoustic emission to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training acoustic emission sensed by the acoustic emission sensor of the set of sensors of the welding machine; and/or
determining an acoustic amplitude of the acoustic emission generated when the welding machine sets the connector to the object and represented by the acoustic emission data signal, and applying the machine learning model to the determined acoustic amplitude such that the machine learning model involves the acoustic amplitude to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training acoustic amplitude determined of the training acoustic emission sensed by the acoustic emission sensor of the set of sensors of the welding machine; and/or
determining an acoustic frequency of the acoustic emission generated when the welding machine sets the connector to the object and represented by the acoustic emission data signal, and/or applying the machine learning model to the determined acoustic frequency such that the machine learning model involves the acoustic frequency to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training acoustic frequency determined of the training acoustic emission sensed by the acoustic emission sensor of the set of sensors of the welding machine.

### Claim 45 (depends on 37)

The computer implemented method of claim 37, wherein the machine learning model comprises an artificial neural, an ensemble machine learning method, a classifier, a regressor, or a combination thereof.

### Claim 46 (depends on 37)

An ultrasonic facility adapted to ultrasonically set a connector to an object, comprising:
a welding machine equipped with a set of sensors having a power supply sensor configured to sense a power supplied to set a connector to an object in runtime, and
a computer configured to execute the computer implemented method according to claim 37,
wherein the computer is in communication with the welding machine via the provided welding machine interface.

### Claim 47 (depends on 46)

The ultrasonic facility of claim 46, comprising a handling machine in communication with the computer via the provided handling machine interface, the handling machine being configured to handle the object with the set connector and to eliminate the object with the set connector when the non-consistency data signal is transferred to the handling machine via the handling machine interface provided by the computer; and/or
wherein the set of sensors of the welding machine comprises a force sensor configured to sense a push force applied to the connector while being set to the object in runtime, a distance sensor configured to sense a distance the welding machine forwards the connector to the object in runtime, a velocity sensor configured to sense a velocity by which the welding machine forwards the connector to the object in runtime, an acoustic emission sensor, or any combination thereof.

### Claim 48

An automated process of setting a connector to an object using a welding machine which is equipped with a set of sensors having a power supply sensor configured to sense a power supplied to set the connector to the object in runtime, the automated process comprising the steps of:
predefining a threshold product performance metric;
obtaining a power supply data signal from the welding machine, which power supply data signal represents the sensed power supplied by the welding machine to set the connector to the object;
applying a machine learning model to the power represented by the obtained power supply data signal such that the machine learning model calculates a model product performance metric, wherein the machine learning model is specifically pre-trained with training power sensed by the power supply sensor of the set of sensors of the welding machine;
comparing the calculated model product performance metric to the threshold product performance metric; and
optionally, physically eliminating the object with the set connector when the calculated product performance metric does not comply with the threshold product performance metric.

### Claim 49 (depends on 48)

The automated process of claim 48, wherein
applying the machine learning model to the power represented by the obtained power supply data signal and comparing the calculated model product performance metric to the threshold product performance metric is automatically performed on a computer; and
the computer triggers a handling machine to physically eliminate the object with the set connector when the calculated product performance metric does not comply with the threshold product performance metric.

### Claim 50 (depends on 48)

The automated process of claim 48, wherein the set of sensors of the welding machine has a force sensor configured to sense a push force applied while being set to the object in runtime, a distance sensor configured to sense a distance the welding machine forwards the connector to the object in runtime, and/or a velocity sensor configured to sense a velocity by which the welding machine forwards the connector to the object in runtime, the automated process comprising the steps of:
obtaining a push force data signal from the welding machine, which push force data signal represents the sensed push force applied to the connector while being set to the object, and applying the machine learning model to the push force represented by the obtained push force data signal such that the machine learning model involves the push force to calculate the model product performance metric, wherein the machine learning model is specifically trained with training push force sensed by the force sensor of the set of sensors of the welding machine; and/or
obtaining a distance data signal from the welding machine, which distance data signal represents the sensed distance the welding machine forwards the connector to the object, and applying the machine learning model to the distance represented by the obtained distance data signal such that the machine learning model involves the distance to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training distance sensed by the distance sensor of the set of sensors of the welding machine;
and/or obtaining a velocity data signal from the welding machine, which velocity data signal represents the sensed velocity by which the welding machine forwards the connector to the object, and applying the machine learning model to the velocity represented by the obtained velocity data signal such that the machine learning model involves the velocity to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training velocity sensed by the velocity sensor of the set of sensors of the welding machine.

### Claim 51 (depends on 48)

The automated process of claim 48, comprising the steps of:
determining a signal amplitude of an ultrasonic signal supplied by the welding machine to set the connector to the object and represented by a signal amplitude data signal, and applying the machine learning model to the determined signal amplitude such that the machine learning model involves the signal amplitude to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with a training signal amplitude determined by the power supply sensor of the set of sensors of the welding machine; and/or
determining a signal frequency of an ultrasonic signal supplied by the welding machine to set the connector to the object and represented by a frequency data signal, and applying the machine learning model to the determined signal frequency such that the machine learning model involves the signal frequency to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training signal frequency determined by the power supply sensor of the set of sensors of the welding machine.

### Claim 52 (depends on 48)

The automated process of claim 48, wherein the set of sensors of the welding machine has an acoustic emission sensor configured to sense an acoustic emission generated when the welding machine sets the connector to the object in runtime, the automated process comprising the steps of:
obtaining an acoustic emission data signal from the welding machine, which acoustic emission data signal represents the sensed acoustic emission generated when the welding machine sets the connector to the object, and applying the machine learning model to the acoustic emission represented by the obtained acoustic emission data signal such that the machine learning model involves the acoustic emission to calculate the model product performance metric data signal, wherein the machine learning model is specifically pre-trained with training acoustic emission sensed by the acoustic emission sensor of the set of sensors of the welding machine; and/or
determining an acoustic amplitude of the acoustic emission generated when the welding machine sets the connector to the object and represented by the acoustic emission data signal, and applying the machine learning model to the determined acoustic amplitude such that the machine learning model involves the acoustic amplitude to calculate the model product performance metric data signal, wherein the machine learning model is specifically trained with training acoustic amplitude determined of the training acoustic emission sensed by the acoustic emission sensor of the set of sensors of the welding machine; and/or
determining an acoustic frequency of the acoustic emission generated when the welding machine sets the connector to the object and represented by the acoustic emission data signal, and applying the machine learning model to the determined acoustic frequency such that the machine learning model involves the acoustic frequency to calculate the model product performance metric data signal, wherein the machine learning model is specifically trained with training acoustic frequency determined of the training acoustic emission sensed by the acoustic emission sensor of the set of sensors of the welding machine.

### Claim 53 (depends on 37)

A manufacturing method of manufacturing a computer program having code means configured to implement the computer implemented method of claim 37, wherein when being executed on a computer, comprising specifically training the machine learning model by
setting a plurality of test connectors to objects by means of a welding machine, having a power supply sensor configured to sense a training power supplied to set the test connectors to the objects in runtime;
obtaining power supply data signals from the welding machine, which power supply data signals represent the sensed training power supplied by the welding machine to set the plurality of test connectors to the objects;
feeding the machine learning model with the training power represented by the obtained power supply data signals such that the machine learning model calculates a training model product performance metric for each of the plurality of test connectors;
measuring a product performance metric for each of the plurality of test connectors, which product performance metric preferably is dimensioned to pull the respective test connector out of the associated object; and
validating the machine learning model by comparing the training model product performance metrics calculated by the machine learning model to respective product performance metrics measured for the plurality of test connectors.

### Claim 54 (depends on 53)

The manufacturing method of claim 53, wherein the welding machine has a force sensor configured to sense a training push force applied to the test connectors while being set to the objects, a distance sensor configured to sense a training distance the welding machine forwards the test connectors to the objects, and/or a velocity sensor configured to sense a training velocity by which the welding machine forwards the test connectors to the objects, wherein specifically training the machine learning model comprises:
obtaining push force data signals from the welding machine, which push force data signals represent the sensed training push force applied to the test connectors while being set to the objects, and feeding the machine learning model with the training push force represented by the obtained push force data signals; and/or
obtaining distance data signals from the welding machine, which distance data signals represent the training distance the welding machine forwards the test connectors to the objects, and feeding the machine learning model with the training distance represented by the obtained distance data signals; and/or
obtaining velocity data signals from the welding machine, which velocity data signals represent the training velocity by which the welding machine forwards the test connectors to the objects, and feeding the machine learning model with the training velocity represented by the obtained velocity data signals.

### Claim 55 (depends on 53)

The manufacturing method of claim 53, wherein specifically training the machine learning model comprises
determining training signal amplitudes of ultrasonic signals supplied by the welding machine to set the test connectors to the objects and represented by signal amplitude data signals, and feeding the machine learning model with the training signal amplitudes; and/or
determining training signal frequencies of ultrasonic signals supplied by the welding machine to set the test connectors to the objects and represented by the signal frequency data signals, and feeding the machine learning model with the training signal frequencies.

### Claim 56 (depends on 53)

The manufacturing method of claim 53, wherein the welding machine has an acoustic emission sensor configured to sense a training acoustic emission generated when the welding machine sets the test connectors to the objects, wherein specifically training the machine learning model comprises:
obtaining acoustic emission data signals from the welding machine when the welding machine sets the test connectors to the objects, and feeding the machine learning model with the training acoustic emission represented by the obtained velocity data signals; and
determining training acoustic amplitudes of the acoustic emission generated when the welding machine sets the connectors to the objects, and feeding the machine learning model with the training force amplitudes; and/or
determining training acoustic frequencies of the acoustic emission generated when the welding machine sets the test connectors to the objects, and feeding the machine learning model with the determined training acoustic frequencies.

### Claim 57 (depends on 53)

The manufacturing method of claim 53, wherein the plurality of test connectors and objects comprises at least 50 test connectors and objects, or at least 100 test connectors and objects.

### Claim 58 (depends on 37)

A computer program product comprising computer code means configured to control a processor of a computer to implement the computer implemented method of claim 37 when being executed on the computer.
