System and method for labeling welding monitoring time periods using machine learning techniques

Abstract

Systems and methods for using machine learning techniques to label non-welding time periods are described. In some examples, a welding monitoring system can collect various data from sensors and / or welding equipment in the welding area within a certain time period. This data can be evaluated to divide the time period into welding and non-welding periods. The welding monitoring system can combine the collected data with one or more machine learning models and / or techniques to determine which non-welding activities were performed during the non-welding periods. In some examples, these machine learning models can be continuously trained, updated, and / or improved using feedback from operators and / or other individuals, data from ongoing welding and / or non-welding activities, and data from other welding monitoring systems and / or machine learning models.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to and the benefit of U.S. Provisional Application No. 62 / 899,293, filed on September 12, 2019, entitled “Systems and Methods for Labeling Weld Monitoring Time Periods Using Machine Learning Techniques,” the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure relates generally to welding monitoring systems, and more particularly to systems and methods for marking welding monitoring time periods using machine learning techniques. Background Art

[0004] Welding monitoring systems monitor data related to welding operations and related activities. Conventional welding monitoring systems rely on input from welding operators to understand what is happening when no welding is occurring. Lack of operator input can lead to inefficient monitoring.

[0005] By comparing conventional and traditional systems with the present disclosure set forth in the remainder of this application with reference to the accompanying drawings, the limitations and disadvantages of conventional and traditional methods will become apparent to those skilled in the art. Summary of the invention

[0006] The present disclosure is directed to systems and methods for marking welding monitoring time periods using machine learning techniques, substantially as illustrated by and / or described in connection with at least one of the figures and as more fully set forth in the claims.

[0007] These and other advantages, aspects and novel features of the present disclosure, as well as details of illustrated examples of the disclosure, will be more fully understood from the following description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A welding system in communication with a local monitoring station is shown in accordance with some aspects of the present disclosure.

[0009] Figure 2 is a diagram showing some aspects of the present disclosure Figure 1 A block diagram of a local monitoring station with further details.

[0010] Figure 3a to Figure 3b is a flow chart illustrating an example activity recognition procedure according to some aspects of the present disclosure.

[0011] Figures 4a to 4d Demonstrates the use of some aspects of the present disclosure Figure 3a to Figure 3b Example labeling of a twenty-four hour period by the activity recognition program.

[0012] The drawings are not necessarily drawn to scale. Where appropriate, the same or similar reference numerals are used in the drawings to represent similar or identical elements. For example, a reference numeral using a letter (e.g., welding unit 101a, welding unit 101b) represents an example of the same reference numeral (e.g., welding unit 101) without the letter. DETAILED DESCRIPTION

[0013] Some examples of the present disclosure relate to welding monitoring systems that are configured to use machine learning techniques to mark non-welding time periods. Welding monitoring systems sometimes employ various sensors to monitor welding parameters of the welding system during welding. However, conventional welding monitoring systems rely on operators to tell the monitoring system what is happening when there is no welding. Unfortunately, operators are often preoccupied or forgetful and neglect to describe what is happening when there is no welding.

[0014] Thus, the present disclosure discusses welding monitoring systems that use machine learning techniques to try and understand activities that are occurring when there is no welding. This enables the welding monitoring system to continue its monitoring even if an operator neglects to inform the system of what is occurring. In some examples, the welding monitoring system may use various machine learning models to identify patterns that may indicate that one or more activities are occurring when there is no welding. In some examples, feedback from operators and / or other individuals, data from ongoing welding and / or non-welding activities, and data from other welding monitoring systems and / or machine learning models may be used to continuously train, update, and / or improve these machine learning models.

[0015] Some examples of the present disclosure relate to a welding system, which includes: a welding monitoring system, which is configured to capture one or more characteristic characteristics of a welding-related operation via a user interface or one or more sensors during a first time period; a processing circuit system; and a memory circuit system, which includes one or more machine learning models and computer-readable instructions, which when executed cause the processing circuit system to: identify one or more unmarked non-welding time periods based on the one or more characteristic characteristics; use the one or more machine learning models to determine whether one or more tags are applicable to the one or more unmarked non-welding time periods based on the one or more characteristic characteristics; and in response to determining that a tag among the one or more tags is applicable to a certain unmarked non-welding time period among the one or more unmarked non-welding time periods, associate the tag with the unmarked non-welding time period.

[0016] In some examples, the memory circuit system further includes computer readable instructions that, when executed, cause the processing circuit system to: train the one or more machine learning models using the association between the mark and the unmarked non-welding time period, and train the one or more machine learning models using at least one characteristic feature associated with the unmarked non-welding time period. In some examples, the memory circuit system further includes computer readable instructions that, when executed, cause the processing circuit system to: train the one or more machine learning models using one or more other machine learning models, the one or more other machine learning models being applied to one or more other welding-related operations. In some examples, the memory circuit system further includes computer readable instructions that, when executed, cause the processing circuit system to: determine a confidence level for the mark. In some examples, associating the mark with the unmarked non-welding time period includes: using a first machine learning model to determine a first mark that is applicable to the unmarked non-welding activity time period and a first confidence level for the first mark; using a second machine learning model to determine a second mark that is applicable to the unmarked non-welding activity time period and a second confidence level for the second mark; and associating the mark with the unmarked non-welding activity time period based on the first confidence level or the second confidence level, the mark including the first mark or the second mark.

[0017] In some examples, the memory circuit system further includes a rest model, the rest model includes a model of one or more welding operator rest patterns, and determining whether the one or more tags are applicable to the one or more unmarked non-welding activity time periods further includes: using the rest model, based on the one or more characteristic characteristics, determining whether the one or more rest tags are applicable to the one or more unmarked non-welding time periods. In some examples, the one or more rest tags include one or more of the following: a morning rest period, a lunch rest period, an evening rest period, a restroom rest period, a planned rest period, or a shift change period. In some examples, the memory circuit system further includes a regular activity model, the regular activity model includes a model of one or more regular activity patterns, and determining whether the one or more tags are applicable to the one or more unmarked non-welding time periods further includes: using the regular activity model and cluster analysis, based on the one or more characteristic characteristics, determining whether the one or more regular activity tags are applicable to the one or more unmarked non-welding time periods. In some examples, the cluster analysis forms a classification tree from the one or more unmarked non-welding time periods.

[0018] In some examples, wherein determining whether the one or more tags are applicable to the one or more unmarked non-welding time periods further includes: using the cluster analysis to determine whether an outlier time period in the one or more unmarked non-welding time periods is dissimilar to the other one or more unmarked non-welding time periods or the one or more regular activity patterns to the extent that the outlier time period should be marked as abnormal. In some examples, the memory circuit system further includes computer readable instructions that, when executed, cause the processing circuit system to: associate a critical rating with the outlier time period based on the duration or dissimilarity of the time period. In some examples, the memory circuit system further includes computer readable instructions that, when executed, cause the processing circuit system to: issue an alarm or inhibit operation in response to determining that the outlier time period should be marked as abnormal and is associated with a high critical rating.

[0019] In some examples, the one or more sensors include one or more of the following: a current sensor, a voltage sensor, a resistance sensor, a wire feed speed sensor, a gas flow sensor, a clamping sensor, an NFC interrogator, an RFID interrogator, a Bluetooth interrogator, a barcode reader, a camera, an optical sensor, an infrared sensor, an acoustic sensor, a sound sensor, a microphone, a position sensor, a global positioning system, an accelerometer, an inertial measurement unit, an X-ray sensor, a radiation detection sensor, a torque sensor, a non-destructive testing sensor, a temperature sensor, or a humidity sensor. In some examples, the one or more characteristic characteristics include one or more operating characteristics, activity-specific characteristics, activity markers, pre-activity stage characteristics, or post-activity stage characteristics. In some examples, the one or more operating characteristics include one or more of the following: a shift start time, a shift end time, a unique operator identification, an operator name, an operator qualification, a filler material characteristic, material preparation information, a material type, a gas type, an operating location, an ambient temperature, or an ambient humidity. In some examples, the one or more activity-specific features include one or more of the following: an activity start time, an activity end time, a previous activity, a previous event, a subsequent activity, a subsequent event, an image of a welding-related operation, or an image of an operating environment. In some examples, the one or more pre-activity phase features or post-activity phase features include a pre-activity phase start time or a post-activity phase start time, a pre-activity phase end time or a post-activity phase end time, a pre-activity phase duration or a post-activity phase duration, a number of completed welds, an arc time, a number of completed parts, a downtime duration, an operating time, an operating position, an ambient temperature, or an ambient humidity. In some examples, all of the one or more unmarked non-welding time periods are within the time period. In some examples, at least one of the one or more unmarked non-welding time periods is outside the time period.

[0020] Some examples of the present disclosure relate to a method for automatically marking non-welding time periods of welding-related operations, the method comprising: capturing one or more characteristic features of the welding-related operation within a time period via a user interface or one or more sensors; identifying one or more unmarked non-welding time periods based on the one or more characteristic features via a processing circuit system; using one or more machine learning models stored in a memory circuit system, determining whether one or more tags are applicable to the one or more unmarked non-welding time periods based on the one or more characteristic features; and in response to determining that a tag among the one or more tags is applicable to a certain unmarked non-welding time period among the one or more unmarked non-welding time periods, associating the tag with the unmarked non-welding time period.

[0021] Figure 1An example welding system 100 and a local monitoring station 200 are shown. As shown, the welding system 100 includes a welding torch 118 and a workpiece fixture 117 coupled to a welding-type power supply 108 within a welding cell 101. As shown, the local monitoring station 200 is electrically coupled (and / or in electrical communication) with the welding-type power supply 108. In some examples, the local monitoring station 200 can also communicate with the welding torch 118 (e.g., via the welding-type power supply 108).

[0022] exist Figure 1 In the example of , an operator 116 is manipulating a welding torch 118 near a welding station 112 within a welding cell 101. In some examples, the welding station 112 can be and / or include a clamping system configured to hold one or more workpieces 110. In some examples, the clamping system can include one or more workpiece clamps 117 (e.g., manual clamps and / or pneumatic clamps). In some examples, the workpiece(s) 110 can be independent of the welding station 112, for example, being free-standing elements such as structural steel elements, pipes, or bridges. Although Figure 1 A human operator 116 is shown in FIG. 1 , but in some examples, the operator 116 may be (and / or control) a robot and / or an automatic welding machine.

[0023] exist Figure 1 In the example of , the welding torch 118 is coupled to the welding power supply 108 via the welding cable 126. The clamp 117 is also coupled to the welding power supply 108 via the clamp cable 115. The welding power supply 108 in turn communicates with the local monitoring station 200, such as via the conduit 130. In some examples, the welding power supply 108 may alternatively or additionally include wireless communication capabilities (e.g., wireless communication circuitry) through which wireless communication can be established with the local monitoring station 200.

[0024] exist Figure 1 In some examples, the welding torch 118 is a welding gun configured for gas metal arc welding (GMAW). In some examples, the welding torch 118 may include an electrode holder (i.e., a stinger) configured for shielded metal arc welding (SMAW). In some examples, the welding torch 118 may include a welding torch and / or welding rod configured for gas tungsten arc welding (GTAW). In some examples, the welding torch 118 may include a welding gun configured for flux cored arc welding (FCAW). In some examples, the welding torch 118 may additionally or alternatively include a welding rod. Figure 1 In the example of , the welding torch 118 includes a trigger 119. In some examples, the trigger 119 can be actuated by the operator 116 to start a welding-type operation (e.g., an arc).

[0025] exist Figure 1 In some examples, the welding-type power supply 108 includes (and / or is coupled to) a wire feeder 140. In some examples, the wire feeder 140 houses a wire spool that is used to provide a wire electrode (e.g., solid wire, flux-cored wire, coated wire) to the welding torch 118. In some examples, the wire feeder 140 further includes motorized rollers that are configured to feed the wire electrode (e.g., from a spool) to the welding torch 118 and / or retract the wire electrode from the welding torch 118 (e.g., to a spool).

[0026] exist Figure 1 In some examples, the welding-type power supply 108 also includes (and / or is coupled to) a gas source 142. In some examples, the gas source 142 supplies a shielding gas and / or a shielding gas mixture to the welding torch 118 (e.g., via the cable 126). As used herein, a shielding gas can refer to any gas (e.g., CO2, argon) or gas mixture that can be provided to the arc and / or weld pool to provide a specific local atmosphere (e.g., to protect the arc, improve arc stability, limit the formation of metal oxides, increase the moisture of the metal surface, change the chemical properties of the weld deposit, etc.).

[0027] exist Figure 1 and Figure 2 In the example of FIG. 1 , the welding-type power supply 108 also includes an operator interface 144. Figure 1 In some examples, the operator interface 144 includes one or more adjustable inputs (e.g., knobs, buttons, switches, keys, etc.) and / or outputs (e.g., display screens, lights, speakers, etc.) on the welding power supply 108. In some examples, the operator interface 144 may include a remote controller and / or a pendant. In some examples, the operator 116 (and / or other users) may use the operator interface 144 to input and / or select one or more welding parameters (e.g., voltage, current, gas type, wire feed speed, workpiece material type, filler type, etc.) of the welding power supply 108 and / or a welding operation of the welding power supply 108. In some examples, the operator interface 144 may further include one or more ports configured to connect to (and / or receive) one or more external memory devices (e.g., floppy disks, compact disks, digital video disks, flash drives, etc.).

[0028] exist Figure 1In an example of the welding-type power supply 108, the welding-type power supply 108 includes a power conversion circuit system 132, which is configured to receive input power (e.g., from a mains, a generator, etc.) and convert the input power into a welding-type output power. In some examples, the power conversion circuit system 132 may include circuit elements (e.g., transformers, rectifiers, capacitors, inductors, diodes, transistors, switches, etc.) capable of converting input power into output power. In some examples, the power conversion circuit system 132 may also include one or more controllable circuit elements. In some examples, the controllable circuit element may include a circuit system configured to change a state (e.g., ignition, on / off, closed / open, etc.) based on one or more control signals. In some examples, the (multiple) states of the controllable circuit element may affect the operation of the power conversion circuit system 132, and / or affect the characteristics of the output power provided by the power conversion circuit system 132 (e.g., current / voltage amplitude, frequency, waveform, etc.). In some examples, the controllable circuit element may include, for example, a switch, a relay, a transistor, etc. In examples where the controllable circuit element includes a transistor, the transistor may include any suitable transistor, such as a MOSFET, a JFET, an IGBT, a BJT, or the like.

[0029] As shown, the welding-type power supply 108 further includes a control circuit system 134 electrically coupled to the power conversion circuit system 132 and configured to control the power conversion circuit system. In some examples, the control circuit system 134 can include a processing circuit system (and / or one or more processors) and analog memory and / or digital memory. In some examples, the control circuit system 134 is configured to control the power conversion circuit system 132 to ensure that the power conversion circuit system 132 generates appropriate welding-type output power to implement the desired welding-type operation.

[0030] In some examples, the control circuit system 134 is also electrically coupled to the wire feeder 140 and / or the gas source 142 and / or is configured to control the wire feeder and / or the gas source. In some examples, the control circuit system 134 can control the wire feeder 140 to output the welding wire at a target speed and / or a target direction. For example, the control circuit system 134 can control the motor of the wire feeder 140 to feed the welding wire electrode 250 into the welding torch 118 (and / or retract the welding wire electrode 250 from the welding torch) at a target speed. In some examples, the welding-type power supply 108 can control the gas source 142 to output a target type and / or amount of gas. For example, the control circuit system 134 can control a valve connected to the gas source 142 to adjust the gas delivered to the welding torch 118.

[0031] exist Figure 1In the example of , the welding system 100 further includes several sensors 150. In some examples, one or more of the sensors 150 may include one or more of the following: a current sensor, a voltage sensor, a resistance sensor, a wire feed speed sensor, a gas flow sensor, a clamping sensor, an NFC interrogator, an RFID interrogator, a Bluetooth interrogator, a barcode reader, a camera, an optical sensor, an infrared sensor, an acoustic sensor, a sound sensor, a microphone, a position sensor, a global positioning system, an accelerometer, an inertial measurement unit, an X-ray sensor, a radiation detection sensor, a torque sensor, a non-destructive detection sensor, a temperature sensor and / or a humidity sensor. As shown, the sensor 150 is positioned within, on and / or near the workpiece fixture 117, the welding torch 118, the welding-type power supply 108, the wire feeder 140, the gas source 142 and the power conversion circuit system 132.

[0032] exist Figure 1 In the example of FIG. 1 , the sensor 150 is also shown mounted to and / or suspended from a fixed structure (e.g., a wall, door, ceiling, pillar, curtain, etc.) of the welding cell 101. Although only one sensor 150 is shown mounted to and / or suspended from a fixed structure, in some examples, multiple sensors 150 can be mounted to and / or suspended from a fixed structure. As shown, multiple sensors 150 are also mounted to and / or suspended from an unattended robot vehicle 152 (e.g., a drone). Although in Figure 1 In the examples shown, mechanical vehicle 152 is an aircraft, but in some examples, mechanical vehicle 152 may alternatively be a ground vehicle or a water vehicle.

[0033] In some examples, the sensor 150 can be configured to sense, detect, and / or measure various welding data of the welding system 100. For example, the sensor 150 can sense, detect, and / or measure one or more positions, locations, and / or movements of the operator 116, the welding torch 118, the workpiece 110, and / or other objects within the welding unit 101. As another example, the sensor 150 can sense, detect, and / or measure the air temperature, air quality, electromagnetic characteristics, and / or noise in the welding unit 101. As another example, the sensor 150 can sense, detect, and / or measure the voltage and / or current of the power received by the welding power supply 108, the power conversion circuit system 132, and / or the welding torch, and / or the voltage and / or current of the power output by the welding power supply 108 and / or the power conversion circuit system 132. As another example, the sensor 150 can sense, detect, and / or measure the speed (e.g., rate and / or wire feeding direction) of the wire feeder 140 and / or the type of welding wire fed by the wire feeder 140. As another example, the sensor 150 can sense, detect and / or measure the gas type and / or gas flow rate from the gas source 142 (e.g., through a valve) to the welding torch 118. As another example, the sensor 150 can sense, detect and / or measure a trigger action signal (e.g., actuation, de-actuation, etc.) of the welding torch 118 and / or a clamping signal (e.g., clamping, unclamping, etc.) of the clamp 117.

[0034] In some examples, the sensors 150 may be configured to transmit sensed, detected, and / or measured data to the welding-type power supply 108 and / or the local monitoring station 200. In some examples, the control circuitry 134 may be in communication with some or all of the sensors 150 and / or otherwise configured to receive information from the sensors 150. In some examples, data from the local monitoring station may be in communication with some or all of the sensors 150 and / or otherwise configured to receive information from the sensors 150.

[0035] In some examples, a welding operation (and / or welding process) may begin when an operator 116 actuates a trigger 119 of a welding torch 118 (and / or otherwise activates the welding torch 118). During the welding operation, welding power provided by the welding power supply 108 may be applied to an electrode (e.g., a wire electrode) of the welding torch 118 to generate a welding arc between the electrode and one or more workpieces 110. The heat of the arc may melt a portion of a filler material (e.g., a welding wire) and / or the workpiece 110, thereby generating a molten weld pool. Movement of the welding torch 118 (e.g., by an operator) may move the weld pool, thereby generating one or more welds 111.

[0036] When the welding operation is complete, the operator 116 can release the trigger 119 (and / or otherwise deactivate / deactivate the welding torch 118). In some examples, the control circuit system 134 can detect that the welding operation has been completed. For example, the control circuit system 134 can detect a trigger release signal via the sensor 150. As another example, the control circuit system 134 can receive a torch deactivation command via the operator interface 144 (e.g., in the case where the welding torch 118 is operated by a robot and / or an automatic welding machine).

[0037] In some examples, the sensor 150 can detect certain welding data related to the welding power supply 108, the fixture 117, the worktable 112, and / or the welding torch 118 during the welding process. In some examples, the welding power supply 108 can also detect certain welding data (e.g., welding data input via the operator interface 144, detected by the control circuit system 134, and the like). In some examples, the sensor 150 and / or the welding power supply 108 can be configured to transmit this welding data to the local monitoring station 200 (directly and / or through the welding power supply 108). In some examples, the welding data can be transmitted to the local monitoring station 200 in real time, periodically, and / or after the welding operation during the welding operation.

[0038] Figure 2 is a block diagram showing example components and connections of local monitoring station 200. Figure 2 In the example of , the example local monitoring station 200 is electrically (and / or communicatively) coupled to the sensors 150 and / or welding equipment 151 (e.g., power supply 108, welding torch 118, etc.) of several example welding cells 101. Figure 2 In the example of FIG. 3 , three welding units 101 are shown, but in some examples, there may be more or fewer welding units 101. In some examples, the local monitoring station 200 may receive data from each welding unit 101 continuously and / or periodically.

[0039] exist Figure 2In the example of , the local monitoring station 200 is electrically (and / or communicatively) coupled to a user interface (UI) 202. In some examples, the UI 202 may include one or more input devices (e.g., a touch screen, a mouse, a keyboard, buttons, knobs, microphones, dials, etc.) and / or output devices (e.g., a display screen, speakers, lights, etc.). In some examples, the UI 202 may further include one or more ports configured to connect to (and / or receive external memory devices) one or more external memory devices (e.g., floppy disks, compact disks, digital video disks, flash drives, etc.). In operation, the operator 116 or other user may provide input to the local monitoring station 200 and / or receive output from the local monitoring station via the UI 202. Although in Figure 2 In the example of , UI 202 is shown as a separate component, but in some examples, UI 202 can be part of local monitoring station 200.

[0040] exist Figure 2 In some examples, the local monitoring station 200 communicates with one or more remote monitoring stations 204 and one or more central servers 206 via a network 208 (e.g., the Internet, a wide area network, a local area network, etc.). In some examples, the local monitoring station 200 can communicate directly with the one or more remote monitoring stations 204 and / or the one or more central servers 206 instead of communicating through the network 208. In some examples, the central server(s) 206 can be implemented via the local monitoring station 200 and / or one or more of the remote monitoring stations 204. In some examples, one or more of the remote monitoring station(s) 204 can be a local monitoring station 200 located at a remote location.

[0041] exist Figure 2In an example of the present invention, the local monitoring station 200 includes a communication circuit system 210, a processing circuit system 212, and a memory circuit system 214 interconnected to each other via the same electrical bus. In some examples, the processing circuit system 212 may include one or more processors. In some examples, the communication circuit system 210 may include one or more wireless adapters, wireless network cards, cable adapters, line adapters, dongles, radio frequency (RF) devices, wireless communication devices, Bluetooth devices, devices compliant with IEEE802.11, WiFi devices, cellular devices, GPS devices, Ethernet ports, network ports, lightning cable ports, cable ports, etc. In some examples, the communication circuit system 210 may be configured to facilitate communication via one or more wired media and / or protocols (e.g., (multiple) Ethernet cables, (multiple) universal serial bus cables, etc.) and / or wireless media and / or protocols (e.g., near field communication (NFC), ultra-high frequency radio waves (commonly known as Bluetooth), IEEE 802.11x, Zigbee, HART, LTE, Z-Wave, Wireless HD, WiGig, etc.). In some examples, local monitoring station 200 may be implemented by a desktop computer or a local server computer.

[0042] exist Figure 2 In the example of FIG. 1 , the memory circuitry 214 stores an activity identification program 300, discussed further below, certain planned non-welding activities 216, and a database 218. Although in FIG. Figure 2 In the example of FIG. 3 , some or all of activity identification program 300 may be stored in the memory circuitry of central server(s) 206 and / or executed by the processing circuitry of central server(s) 206. Although in FIG. Figure 2 200, but in some examples, some or all of database 218 may be stored in memory circuitry of central server(s) 206 and / or one or more remote monitoring stations 204. In some examples, database 218 may actually include multiple databases.

[0043] In some examples, the activity recognition program 300 can analyze data collected by the sensors 150 and / or welding equipment 151 of each welding unit 101, as well as data collected via the UI 202 (e.g., operator information). In some examples, the activity recognition program 300 can determine certain characteristic features based on the analysis of the data. These characteristic features can be used to define time periods of welding activity and / or non-welding activity. In some examples, the characteristic features can also include one or more activity markers attributed to one or more non-welding time periods.

[0044] In some examples, planned non-welding activities 216 can be used to associate activity tags with some or all non-welding time periods that have not been marked. In some examples, planned non-welding activities 216 may include one or more non-welding activities that may occur in one or more time periods of a given day, week, month, etc. For example, planned non-welding activities 216 may include one or more non-welding activities scheduled for a specific operator 116, welding equipment 151, welding unit 101, work place and / or welding operation and (multiple) date / times at which the one or more activities are expected to occur. For example, planned non-welding activities 216 may include data representing one or more planned shift start periods, shift end periods, maintenance periods, breakfast periods, lunch periods, dinner periods, and / or rest periods for operators 116 working with specific welding equipment 151 in a specific welding unit 101, at a specific work place, and / or for a specific welding operation. In some examples, the planned non-welding activities 216 may be input via the UI 202 , transmitted from the remote monitoring station(s) 204 , the central server(s) 206 , and / or other devices, and / or determined programmatically.

[0045] exist Figure 2 In the example of , the activity recognition program 300 includes several machine learning models 220. Figure 2 In the example of , the machine learning model 220 is shown as part of the activity recognition program 300, but in some examples, the machine learning model 220 can be separate from the activity recognition program 300. In some examples, one or more machine learning models 220 can be used to determine one or more activity markers that belong to non-welding time periods that remain unmarked after considering planned non-welding activities 216. In some examples, one or more machine learning models 220 can be used to further analyze feature characteristics to determine the one or more activity markers. In some examples, one or more machine learning models 220 can be associated with one or more welding units 101, welding equipment 151, operators 116, work locations, and / or welding operations.

[0046] In some examples, one or more machine learning models 220 can be continuously trained and / or updated using data from other machine learning models 220 associated with other welding cells 101, welding equipment 151, operators 116, work locations, and / or welding operations. In some examples, one or more machine learning models 220 of the local monitoring station 200 can be continuously trained and / or updated using data from other machine learning models 220 of other remote monitoring stations 204. In some examples, one or more machine learning models 220 of the local monitoring station 200 can be continuously trained and / or updated using planned non-welding activities 216 from the local monitoring station 200 and / or other remote monitoring stations 204.

[0047] Figure 3a to Figure 3b is a flow chart illustrating an example activity recognition program 300. In some examples, the activity recognition program 300 can be implemented using machine-readable (and / or processor-executable) instructions stored in the memory circuitry 214 and / or executed by the processing circuitry 212. Although in the following description, the activity recognition program 300 is described for a single welding unit 101, in some examples, multiple instances of the activity recognition program 300 can be executed simultaneously (e.g., each welding unit 101 has an instance of the activity recognition program 300).

[0048] exist Figure 3a In the example of FIG. 3 , the activity recognition program 300 begins at block 302. At block 302, the sensor 150 and / or welding device 151 of the welding unit 101 collects data and transmits the data to the local monitoring station 200. In some examples, the data may also be collected and / or transmitted by the UI 202. In some examples, such data collection may occur continuously and / or periodically. In some examples, the collected data may be stored in the database 218. Although for ease of understanding in FIG. 3 , the data may be collected and / or transmitted by the UI 202. Figure 3a , however, in some examples, such data collection and / or transmission may occur outside of the process of activity identification program 300 .

[0049] exist Figure 3aIn the example of , after box 302, the activity recognition program 300 proceeds to box 304. At box 304, the activity recognition program 300 analyzes the data collected by the sensor 150, the welding device 151, and / or the UI 202 at box 302 to determine one or more characteristic characteristics. For example, the activity recognition program 300 can determine when to start an arc welding operation and / or the duration of an arc welding operation based on one or more trigger action signals, voltage data, current data, wire feed data, and / or gas source data detected and / or measured by the sensor 150. As another example, the activity recognition program 300 can determine a gas type and / or a welding wire material type based on data detected and / or measured by the sensor 150 and / or data input by the operator 116 via the operator interface 144 and / or the UI 202. As yet another example, activity recognition program 300 may determine information about operator 116 based on credentials supplied by operator 116 via operator interface 144 and / or UI 202 and / or based on operator data detected by sensors 150 .

[0050] In some examples, one or more characteristic features may include one or more operational features, activity-specific features, activity markers, pre-activity phase features, and / or post-activity phase features. In some examples, the operational features may include one or more of the following: shift start time, shift end time, unique operator identification, operator name, operator qualifications, filler material properties, material preparation information, material type, gas type, operating location, ambient temperature, ambient humidity, operation type, or job type. In some examples, the activity-specific features may include one or more of the following: activity start time, activity end time, previous activity, previous event, subsequent activity, subsequent event, image of welding-related operation, and / or image of operating environment. In some examples, the pre-activity phase features and / or post-activity phase features may include pre-activity phase start time and / or post-activity phase start time, pre-activity phase end time and / or post-activity phase end time, pre-activity phase duration and / or post-activity phase duration, number of completed welds, arc time, number of completed parts, downtime duration, operation time, operating location, ambient temperature, and / or ambient humidity. In some examples, activity recognition program 300 can store these characteristic features in database 218 .

[0051] exist Figure 3aIn the example of , after box 304, the activity recognition program 300 proceeds to box 306. At box 306, the activity recognition program 300 identifies one or more welding time periods 402 and / or non-welding time periods 404 based on the characteristic characteristics determined at box 304. For example, the activity recognition program 300 can determine characteristic characteristics related to the start, end, and / or duration of the arc welding operation at box 304, and these characteristic characteristics can be used to determine one or more welding time periods and / or non-welding time periods. In some examples, one or more welding time periods and / or non-welding time periods can be identified based on characteristic characteristics related to one or more earlier, later, and / or overlapping time periods. In some examples, the activity recognition program 300 can store these time periods in the database 218. In some examples, the activity recognition program 300 can associate the welding activity marker with one or more welding time periods 402, and / or store the corresponding association(s) in the database 218.

[0052] Figures 4a to 4d An example twenty-four hour time period 400 of an example welding cell 101 is shown to help illustrate the activity identification procedure 300. Figure 4a In the example of FIG. 4 , the time period 400 has been divided into several welding time periods 402 and non-welding time periods 404 by block 306 of the activity identification procedure 300. Figure 4a In the example of , all non-welding time periods 404 are unmarked, and all welding time periods 402 are marked (eg, marked as normal welding activity).

[0053] exist Figure 3a In the example of , after box 306, the activity recognition program 300 proceeds to box 308. At box 308, the activity recognition program 300 assigns one or more identified non-welding time periods to one or more tags and / or associates one or more tags with one or more identified non-welding time periods based on the planned non-welding activity 216 and / or one or more user inputs. For example, the planned non-welding activity 216 can mark one or more identified non-welding time periods as belonging to the planned non-welding activity (e.g., shift change, maintenance operation, etc.). As another example, the user can manually mark one or more identified non-welding time periods as belonging to some other non-welding activity (e.g., restroom break, resupply, meeting, etc.). In some examples, the (multiple) manually entered tags can be considered as part of the data collected at box 302 and / or the characteristic characteristics determined at box 304. In some examples, the activity recognition program 300 can associate (multiple) non-welding tags with (multiple) appropriate non-welding time periods 404, and / or store the (multiple) corresponding associations in the database 218.

[0054] exist Figure 4b In the example of , non-welding time periods 404a, non-welding time periods 404c, and non-welding time periods 404e have been marked by box 308 of the activity identification program 300. For example, non-welding time periods 404a and non-welding time periods 404b may have been marked as the end of the shift activity at box 308 based on the planned non-welding activity 216. As another example, non-welding time period 404c may have been manually marked as a lunch break activity by the operator 116 (e.g., via UI 202) at boxes 302 or 304 of the activity identification program 300. As shown, the duration of non-welding time period 404a has also been shortened, and a new non-welding time period 404f has been identified in the remaining time. For example, this may occur because the planned non-welding activity 216 indicates that the end of the shift activity is scheduled to end at 9 a.m. Therefore, the activity identification program 300 now needs to try and understand what is happening during the non-welding time period 404f after 9 a.m. Figure 4b In the example of FIG. 4 , the non-welding time period 404 b , the non-welding time period 404 d , and the non-welding time period 404 f are still not marked.

[0055] exist Figure 3a In the example of , after box 308, the activity recognition program 300 proceeds to box 310. At box 310, the activity recognition program 300 uses the (multiple) labeled non-welding time periods 404 of box 308 to train one or more machine learning models 220. In some examples, the activity recognition program 300 can also use the collected data and / or feature characteristics associated with the (multiple) labeled non-welding time periods (and / or one or more previous time periods and / or successive time periods) to train the (multiple) machine learning model(s) 220. In this way, the (multiple) machine learning model(s) 220 can learn to recognize future non-welding time periods 404 with similar characteristics and automatically associate the (multiple) correct labels. Although in Figure 3a In the example of FIG. 3 , block 310 is shown immediately following block 308 , but in some examples, block 310 may occur later toward the end of activity identification procedure 300 .

[0056] exist Figure 3aIn the example of, after box 310, the activity recognition program 300 proceeds to box 350. At box 350, the activity recognition program 300 uses one or more machine learning models 220 and / or one or more (e.g., unsupervised) machine learning techniques to try and mark the (multiple) remaining unmarked non-welding time periods 404. In some examples, multiple machine learning models 220 can be used. In some examples, the (multiple) machine learning models 220 used can depend on the welding unit 101, the welding equipment 151, the operator 116, and / or other factors. In some examples, the activity recognition program 300 can associate any successfully determined (multiple) labels with (multiple) appropriate non-welding time periods 404, and / or store the (multiple) corresponding associations in the database 218. The following about Figure 3b Block 350 is discussed in greater detail.

[0057] exist Figure 3a In the example of , after box 350, the activity recognition program 300 proceeds to box 312. At box 312, the activity recognition program 300 performs one or more system actions. In some examples, the system action may include a reminder, notification, communication, and / or prohibition. For example, the activity recognition program 300 may (e.g., via UI 202) remind the user that the activity recognition program 300 has identified one or more abnormal non-welding time periods that require closer attention. As another example, the activity recognition program 300 may send and / or output a certain (certain) communication (e.g., to a supervisor) to declare that one or more time periods have been associated with the marked mark. As another example, the activity recognition program 300 may determine that one or more required non-welding time periods 404 (e.g., for preheating and / or polishing) do not occur before and / or after welding. In response, the activity recognition program 300 may send one or more signals to one or more pieces of welding equipment 151, the one or more signals representing a command for disabling the welding equipment 151 and / or prohibiting one or more functions of the welding equipment 151.

[0058] exist Figure 3aIn the example of FIG. 314, after block 312, the activity recognition program 300 proceeds to block 314. At block 314, one or more machine learning models 220 are updated and / or trained based on the tags and / or associated collected data and / or feature characteristics applied to the non-welding time period 404 at block 350. After block 314, the activity recognition program 300 proceeds to block 316, where one or more machine learning models 220 are updated and / or trained based on other machine learning models 220 and / or tags (and / or associated collected data and / or feature characteristics) applied to the non-welding time period by other machine learning models 220. For example, the machine learning model(s) 220 for one welding cell 101a can learn to identify the non-welding time period 404 and / or tag the non-welding time period 404 with certain feature characteristics. The machine learning model(s) 220 for the different welding cells 101b may be trained to perform the same identification and / or labeling using data from the machine learning model(s) 220 for the welding cell 101a. In some examples, the other machine learning models 220 and / or associated data may be machine learning models and / or data of the local monitoring station 200 and / or one or more remote monitoring stations 204.

[0059] exist Figure 3a In the example of, after box 316, the activity recognition program 300 proceeds to box 318. At box 318, the activity recognition program 300 determines whether there is any applicable manually entered feedback. In some examples, the manually entered feedback may include one or more inputs (e.g., input via UI 202) for correcting or confirming the mark and / or time period. For example, the activity recognition program 300 may mark one or more non-welding time periods 404 as being related to Class A non-welding activities. Thereafter, the user may (e.g., via UI 202) manually indicate that the content that has been marked as Class A (due to lack of a better mark) by the activity recognition program 300 can be more accurately marked as a cleaning activity. As another example, the user may (e.g., via UI 202) manually indicate that the content that has been marked as abnormal by the activity recognition program 300 is actually a restroom break, and / or the content that has been marked as a maintenance activity by the activity recognition program 300 is correctly marked as a maintenance activity. As yet another example, a user may manually indicate (eg, via UI 202 ) that what activity recognition program 300 identified as one long unmarked non-welding time period is actually several shorter non-welding time periods 404 grouped together.

[0060] exist Figure 3aIn the example of , if there is no manual feedback to be considered, the activity recognition program 300 ends. However, if there is manual feedback to be considered, the activity recognition program 300 proceeds to box 320, where the database 218 is updated based on the feedback, and / or one or more machine learning models are updated and / or trained based on the feedback. Thus, over time, the (multiple) machine learning model 220 can be continuously updated and / or improved based on applied tags, other models, and / or feedback from users, thereby becoming more accurate and / or more comprehensive.

[0061] Figure 3b It further demonstrates Figure 3a FIG. 3 is a flow chart of a machine learning block 350 of an activity recognition program 300. As shown, the machine learning block 350 begins at block 352, where the activity recognition program 300 applies one or more rest models (e.g., from the machine learning model 220) to one or more remaining unmarked non-welding time periods 404 (and / or characteristic features associated with the time period and / or one or more preceding and / or following time periods). In some examples, the rest model(s) may model rest patterns observed to occur for certain operators 116. In some examples, there may be several different rest models, each with its own different patterns.

[0062] In some examples, the activity recognition program 300 can be configured to identify characteristic features that are consistent with the pattern(s) of the rest model(s). For example, the rest model for a given welding operator 116 can model the rest patterns observed for that operator 116. One of these rest patterns can be the observed pattern of the operator 116 taking a twenty-minute restroom break most mornings (e.g., after having morning coffee). As another example, the rest model for a more novice welding operator 116 can model the observed pattern of the more novice operator 116 spending half an hour in the morning and / or afternoon reviewing training and / or instructional materials before starting welding. In some examples, the activity recognition program 300 can analyze characteristic features associated with one or more remaining unmarked non-welding time periods and determine whether these characteristic features are consistent with any pattern of the rest model(s). If the activity recognition program 300 determines that the characteristic characteristics associated with the non-welding time period 404 are consistent with the model pattern(s) (e.g., similar time, similar operator or operator type), the activity recognition program 300 can accordingly mark the non-welding time period 404. In some examples, the activity recognition program 300 can preliminarily associate one or more tags with the non-welding time period(s) 404 that appear to be consistent with the model pattern(s).

[0063] exist Figure 3b In the example of FIG. 3 , after block 352, the activity recognition program 300 proceeds to block 354. At block 354, the activity recognition program 300 determines one or more confidence levels for the one or more preliminary tag associations determined at block 352. The confidence levels can be used to resolve conflicts, such as a situation where two or more different rest models associate two or more different tags to the same untagged non-welding time period.

[0064] exist Figure 3b In the example of FIG. 354, after block 354, the activity recognition program 300 proceeds to block 356. At block 356, the activity recognition program 300 resolves any label conflicts based on the various confidence levels determined at block 354. For example, in the case where two or more different rest models associate two or more different labels to the same unlabeled non-welding time period, the activity recognition program 300 can select the label associated with the highest confidence level. In some examples, blocks 354 and / or 356 can be skipped (e.g., in the case where only one rest model is applied at block 352, and / or in the case where there are no conflicts).

[0065] Figure 4c The non-welding time period 404b is shown as having been marked by the blocks 352 to 356 of the activity recognition program 300. For example, the activity recognition program 300 can identify the non-welding time period 404b (and / or related characteristic features) as matching one or more coffee break patterns of the break model(s) and mark the non-welding time period 404b as a coffee break activity. As shown, the non-welding time period 404d and the non-welding time period 404f remain unmarked.

[0066] exist Figure 3b In the example of FIG. 35 , after block 356, the activity recognition program 300 proceeds to block 358. At block 358, the activity recognition program 300 performs a cluster analysis on one or more remaining unmarked non-welding time periods 404 (and / or characteristic features associated with the time period and / or one or more previous and / or subsequent time periods). In some examples, a cluster analysis may be performed on one or more marked and / or unmarked non-welding time periods 404 (and / or associated characteristic features).

[0067] In some examples, the cluster analysis can generate a classification tree that groups (and / or clusters) one or more time periods together into one or more classifications. In some examples, the classification tree can include a hierarchy of classifications. In some examples, the activity recognition program 300 can use the classification tree and / or one or more classifications to assign one or more tags to the remaining unlabeled non-welding time periods. For example, in the case where the remaining unlabeled non-welding time periods are clustered into a classification with one or more similarly (or identically) labeled non-welding time periods (e.g., based on similar feature characteristics), the activity recognition program 300 can associate the same (or similar) tags with the unlabeled non-welding time periods.

[0068] exist Figure 3b In the example of FIG. 350 , after block 358 , the activity recognition program 300 proceeds to block 360 . At block 360 , the activity recognition program 300 applies one or more conventional activity models (e.g. , in the machine learning model 220 ) to one or more remaining unmarked non-welding time periods (and / or characteristic features associated with these non-welding time periods). In some examples, the conventional activity model(s) may model periodic non-welding activities observed to occur for certain welding cells 101 , welding operations, welding positions, etc. For example, a conventional activity model for a welding cell 101 in Alaska may model an observed pattern of prolonged preheating prior to welding (e.g. , to heat and / or soften the workpiece in the cold Alaska environment). In some examples, there may be several different rest models, each with its own different pattern (e.g. , based on the welding cell 101 , welding operation, welding position, etc.).

[0069] In some examples, the activity recognition program 300 can analyze the characteristic features associated with one or more remaining unmarked non-welding time periods and determine whether the characteristic features are consistent with any pattern of the conventional activity model(s). If so, one or more markers can be preliminarily associated with the time period(s) that appear to be consistent with the marker(s) based on the pattern of the conventional activity model(s). The preliminary associations can remain preliminary until the confidence level of each marker can be evaluated.

[0070] exist Figure 3b In the example of FIG. 3 , after block 360, the activity recognition program 300 proceeds to block 362. At block 362, the activity recognition program 300 determines one or more confidence levels for the one or more preliminary tag associations determined at block 360. The confidence levels can be used to resolve conflicts, such as situations where two or more different conventional activity models associate two or more different tags to the same unlabeled non-welding time period.

[0071] exist Figure 3b In the example of FIG. 36A , after block 362, the activity recognition program 300 proceeds to block 364. At block 364, the activity recognition program 300 resolves any label conflicts based on the various confidence levels determined at block 362. For example, where two or more different conventional activity models associate two or more different labels to the same unlabeled non-welding time period, the activity recognition program 300 may select the label associated with the highest confidence level. In some examples, blocks 362 and / or 364 may be skipped (e.g., where only one conventional activity model is applied at block 360, and / or where no conflicts exist).

[0072] exist Figure 4d In the example of FIG. 3 , non-welding time period 404f has been marked by blocks 360 to 364 of activity recognition program 300. For example, activity recognition program 300 may identify non-welding time period 404f (and / or associated characteristic features) as matching one or more pre-heating modes of (multiple) conventional activity models and mark non-welding time period 404b as a pre-heating activity. As shown, non-welding time period 404d remains unmarked.

[0073] exist Figure 3b In the example of, after box 364, the activity identification program 300 proceeds to box 366. At box 366, the activity identification program 300 marks each of the remaining unmarked non-welding time periods as an unknown downtime activity, an abnormal activity, and / or an outlier time period. In some examples, the activity identification program 300 may associate a criticality rating or a rating with each unknown downtime activity, an abnormal activity, and / or an outlier time period. In some examples, the criticality rating may be based on the duration and / or degree of dissimilarity of the unknown downtime activity, the abnormal activity, and / or the outlier time period. For example, the activity identification program 300 may associate a low criticality rating with an unknown downtime non-welding time period that is relatively short in duration (e.g., 20 minutes) and / or at least slightly similar to one or more other marked time periods (e.g., closer than a certain threshold distance in the classification tree). On the other hand, the activity identification program 300 may associate a high criticality rating with an unknown downtime non-welding time period that is relatively long in duration (e.g., 3 hours) and / or very dissimilar to one or more other marked time periods (e.g., farther than a certain threshold distance in the classification tree).

[0074] exist Figure 4d In the example of FIG. 4 , the non-welding time period 404d is still not marked, but its duration is also relatively short. Figure 4dIn the example of , the activity identification program 300 can mark the non-welding time period 404d as an unknown downtime activity, an abnormal activity, and / or an outlier time period. However, since the duration of the non-welding time period 404d is relatively short, the activity identification program 300 can associate a low criticality rating.

[0075] exist Figure 3a and Figure 3b In the example of , after block 366, the activity identification program 300 returns to block 312. As discussed above, at block 312, the activity identification program 300 can perform one or more system actions. In some examples, the activity identification program 300 can perform one or more high priority system actions (e.g., activating one or more alarms, notifications, communications, and / or inhibitions) at block 312 in response to unknown downtime activities, abnormal activities, and / or outlier time periods having a high criticality rating. In some examples, the activity identification program 300 can perform a low priority system action (e.g., simply setting a replay flag) in response to unknown downtime activities, abnormal activities, and / or outlier time periods having a low criticality rating.

[0076] The present disclosure contemplates the use of machine learning techniques to try and understand non-welding activities that are occurring when the operator 116 neglects to tell the system what is occurring. In some examples, the welding monitoring system can use various machine learning models and / or techniques to identify data patterns that indicate one or more similar and / or identical non-welding activities. In some examples, feedback from operators and / or other individuals, data from ongoing welding and / or non-welding activities, and data from other welding monitoring systems and / or machine learning models can be used to continuously train, update, and / or improve these machine learning models.

[0077] The method and / or system can be implemented with hardware, software, or a combination of hardware and software. The method and / or system can be implemented in a centralized manner in at least one computing system, or in a distributed manner with different elements spread over several interconnected computing systems or cloud systems. Any type of computing system or other device suitable for executing the method described herein is suitable. A typical combination of hardware and software can be a general computing system with a program or other code, which controls the computing system when loaded and executed so that the computing system executes the method described herein. Another typical implementation may include a dedicated integrated circuit or chip. Some implementations may include a non-temporary machine-readable (e.g., computer-readable) medium (e.g., a flash drive, an optical disk, a magnetic storage disk, etc.), which stores one or more lines of code executable by a machine, so that the machine executes the process described herein.

[0078] Although the present method and / or system has been described with reference to certain embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the scope of the present method and / or system. In addition, many modifications may be made to adapt specific circumstances or materials to the teachings of the present disclosure without departing from the scope of the present disclosure. Therefore, the present method and / or system is not intended to be limited to the specific embodiments disclosed, but the present method and / or system will include all embodiments falling within the scope of the appended claims.

[0079] As used herein, "and / or" refers to any one or more of the multiple items in a list connected by "and / or". For example, "x and / or y" refers to any element in the three-element set {(x), (y), (x, y)}. In other words, "x and / or y" means "one or both of x and y". As another example, "x, y and / or z" refers to any element in the seven-element set {(x), (y), (z), (x, y), (x, z), (y, z), (x, y, z)}. In other words, "x, y and / or z" means "one or more of x, y and z".

[0080] As used herein, the terms "eg," and "for example," introduce a list of one or more non-limiting examples, instances, or illustrations.

[0081] As used herein, the terms "coupled," "coupled to," and "coupled with" refer to structural and / or electrical connections, whether attached, attached, connected, joined, fastened, associated, and / or otherwise secured, respectively. As used herein, the term "attach" refers to attaching, coupling, connecting, joining, fastening, associated, and / or otherwise securing. As used herein, the term "connect" refers to attaching, attaching, coupling, joining, fastening, associated, and / or otherwise securing.

[0082] As used herein, the terms "circuit" and "circuitry" refer to physical electronic components (i.e., hardware) and any software and / or firmware ("code") that may configure, be executed by, and / or otherwise be associated with the hardware. As used herein, for example, a particular processor and memory may constitute a first "circuit" when executing a first line or more lines of code, and constitute a second "circuit" when executing a second line or more lines of code. As used herein, a circuitry is "operable" and / or "configured" to perform a function when the circuitry includes the necessary hardware and / or code (if necessary) to perform the function, regardless of whether performance of the function is disabled or enabled (e.g., by a user-configurable setting, a factory adjustment, etc.).

[0083] As used herein, control circuitry may include digital circuit systems and / or analog circuit systems, discrete circuit systems and / or integrated circuit systems, microprocessors, DSPs, etc., located on one or more circuit boards forming part or all of a controller and / or software, hardware and / or firmware for controlling welding processes and / or devices such as power supplies or wire feeders.

[0084] As used herein, the term "processor" refers to a processing device, apparatus, program, circuit, component, system and subsystem, whether implemented in hardware, software in tangible form, or both hardware and software, and whether or not it is programmable. As used herein, the term "processor" includes, but is not limited to, one or more computing devices, hard-wired circuits, signal modification devices and systems, devices and machines for controlling systems, central processing units, programmable devices and systems, field programmable gate arrays, application-specific integrated circuits, systems on chips, systems including discrete components and / or circuits, state machines, virtual machines, data processors, processing facilities, and any combination of the foregoing. The processor may be, for example, any type of general-purpose microprocessor or microcontroller, a digital signal processing (DSP) processor, an application-specific integrated circuit (ASIC), a graphics processing unit (GPU), a reduced instruction set computer (RISC) processor with an advanced RISC machine (ARM) core, etc. The processor may be coupled to a memory device and / or integrated with the memory device.

[0085] As used herein, the terms "memory" and / or "memory device" refer to computer hardware or circuitry for storing information for use by a processor and / or other digital device. The memory and / or memory device may be any suitable type of computer memory or any other type of electronic storage medium, such as read-only memory (ROM), random access memory (RAM), cache memory, compact disk read-only memory (CDROM), electro-optical memory, magneto-optical memory, programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), computer-readable media, etc. The memory may include, for example, non-transitory memory, non-transitory processor-readable medium, non-transitory computer-readable medium, non-volatile memory, dynamic RAM (DRAM), volatile memory, ferroelectric RAM (FRAM), first-in-first-out (FIFO) memory, last-in-first-out (LIFO) memory, stack memory, non-volatile RAM (NVRAM), static RAM (SRAM), cache, buffer, semiconductor memory, magnetic memory, optical memory, flash memory, flash card, compact flash card, memory card, secure digital memory card, micro card, mini card, expansion card, smart card, memory stick, multimedia card, picture card, flash memory device, subscriber identity module (SIM) card, hard disk drive (HDD), solid state drive (SSD), etc. The memory may be configured to store code, instructions, applications, software, firmware and / or data, and may be external to the processor, internal to the processor, or both internal and external to the processor.

[0086] For convenience, the term "power" is used throughout this specification, but "power" also includes related measurements such as energy, current, voltage, and enthalpy. For example, controlling "power" may involve controlling voltage, current, energy, and / or enthalpy, and / or controlling based on "power" may involve controlling based on voltage, current, energy, and / or enthalpy.

[0087] As used herein, welding-type electricity refers to electricity suitable for the following: welding, cladding, brazing, plasma cutting, induction heating, carbon arc cutting and / or hot wire welding / preheating (including laser welding and laser cladding), carbon arc cutting or scraping, and / or resistive preheating.

[0088] As used herein, a welding-type power supply and / or power source refers to any device that is capable of providing power for welding, cladding, brazing, plasma cutting, induction heating, laser processing (including laser welding, laser hybrid welding, and laser cladding), carbon arc cutting or gouging, and / or resistive preheating when power is applied thereto, including but not limited to transformer-rectifiers, inverters, converters, resonant power supplies, quasi-resonant power supplies, switch-mode power supplies, etc., as well as control circuit systems and other auxiliary circuit systems associated therewith.

[0089] Disabling of circuit systems, actuators, and / or other hardware can be accomplished via hardware, software (including firmware), or a combination of hardware and software, and can include physical disconnection, power failure, and / or software control that limits the implementation of commands to actuate circuit systems, actuators, and / or other hardware. Similarly, enabling of circuit systems, actuators, and / or other hardware can be accomplished via hardware, software (including firmware), or a combination of hardware and software using the same mechanism as disabling.

Claims

1. A welding system, comprising: a welding monitoring system configured to capture, via a user interface or one or more sensors, one or more characteristic features of a welding-related operation during a first time period; Processing circuit system; as well as memory circuitry comprising one or more machine learning models and computer readable instructions that, when executed, cause the processing circuitry to: identifying one or more unmarked non-welding time periods based on the one or more characteristic features, determining, using the one or more machine learning models, whether one or more markings are applicable to the one or more unmarked non-welding time periods based on the one or more characteristic features, and In response to determining that a tag of the one or more tags is applicable to an unmarked non-welding time period of the one or more unmarked non-welding time periods, the tag is associated with the unmarked non-welding time period.

2. The system of claim 1, wherein: The memory circuit system further includes computer readable instructions that, when executed, cause the processing circuit system to train the one or more machine learning models using the association between the markings and the unmarked non-welding time periods and at least one characteristic feature associated with the unmarked non-welding time periods.

3. The system of claim 1, wherein: The memory circuit system further includes computer-readable instructions, which, when executed, cause the processing circuit system to: train the one or more machine learning models using one or more other machine learning models, and the one or more other machine learning models are applied to one or more other welding-related operations.

4. The system of claim 1, wherein: The memory circuitry further includes computer readable instructions that, when executed, cause the processing circuitry to: determine a confidence level for the signature.

5. The system of claim 1, wherein: Associating the marking with the unmarked non-welding time period comprises: determining, using a first machine learning model, a first marking applicable to the unmarked non-welding activity time period and a first confidence level for the first marking; determining, using a second machine learning model, a second marking applicable to the unmarked non-welding activity time period and a second confidence level for the second marking; and The marking is associated with the unmarked non-welding activity time period based on the first confidence level or the second confidence level, the marking including the first marking or the second marking.

6. The system of claim 1, wherein: The memory circuitry further includes a rest model, the rest model including a model of one or more welding operator rest patterns, and determining whether the one or more tags are applicable to the one or more untagged non-welding activity time periods further includes: Using the rest model, it is determined whether one or more rest markings are applicable to the one or more unmarked non-welding time periods based on the one or more characteristic features.

7. The system of claim 6, wherein: The one or more break markers include one or more of: a morning break period, a lunch break period, an evening break period, a restroom break period, a planned break period, or a shift change period.

8. The system of claim 1, wherein: The memory circuitry further includes a normal activity model including a model of one or more normal activity patterns, and determining whether the one or more markings are applicable to the one or more unmarked non-welding time periods further includes: Using the conventional activity model and cluster analysis, it is determined whether one or more conventional activity markers are applicable to the one or more unmarked non-welding time periods based on the one or more characteristic features.

9. The system of claim 8, wherein: The cluster analysis forms a classification tree using the one or more unlabeled non-welding time periods.

10. The system of claim 8, wherein: Determining whether the one or more markings are applicable to the one or more unmarked non-welding time periods further comprises: The cluster analysis is used to determine whether an outlier time period among the one or more unmarked non-welding time periods is dissimilar to the other one or more unmarked non-welding time periods or the one or more regular activity patterns to such an extent that the outlier time period should be marked as abnormal.

11. The system of claim 10, wherein: The memory circuitry further includes computer readable instructions that, when executed, cause the processing circuitry to: associate a criticality rating with the outlier time period based on a duration or a degree of dissimilarity of the time period.

12. The system of claim 11, wherein: The memory circuitry further includes computer readable instructions that, when executed, cause the processing circuitry to: issue an alarm or inhibit operation in response to determining that the outlier time period should be marked as an anomaly and associated with a high criticality rating.

13. The system of claim 1, wherein: The one or more sensors include one or more of the following: current sensor, voltage sensor, resistance sensor, wire feed speed sensor, gas flow sensor, clamping sensor, NFC interrogator, RFID interrogator, Bluetooth interrogator, barcode reader, camera, optical sensor, infrared sensor, acoustic sensor, sound sensor, microphone, position sensor, global positioning system, accelerometer, inertial measurement unit, X-ray sensor, radiation sensor, torque sensor, non-destructive testing sensor, temperature sensor or humidity sensor.

14. The system of claim 1, wherein: The one or more feature characteristics include one or more operational features, activity specific features, activity markers, pre-activity phase features, or post-activity phase features.

15. The system of claim 14, wherein: The one or more operating characteristics include one or more of: shift start time, shift end time, unique operator identification, operator name, operator qualifications, fill material characteristics, material preparation information, material type, gas type, operating location, ambient temperature, or ambient humidity.

16. The system of claim 14, wherein: The one or more activity-specific features include one or more of: an activity start time, an activity end time, a previous activity, a previous event, a subsequent activity, a subsequent event, an image of a welding-related operation, or an image of an operating environment.

17. The system of claim 14, wherein: The one or more pre-activity phase characteristics or post-activity phase characteristics include the pre-activity phase start time or the post-activity phase start time, the pre-activity phase end time or the post-activity phase end time, the pre-activity phase duration or the post-activity phase duration, the number of completed welds, the arc time, the number of completed parts, the downtime duration, the operating time, the operating position, the ambient temperature or the ambient humidity.

18. The system of claim 1, wherein: All of the one or more unmarked non-welding time periods are within the first time period.

19. The system of claim 1, wherein: At least one of the one or more unmarked non-welding time periods is outside the first time period.

20. The system of claim 1, wherein: The memory circuitry further includes computer readable instructions that, when executed, cause the processing circuitry to: associate the one or more tags with one or more of the identified non-welding time periods based on the planned non-welding activity, determining, using the one or more machine learning models, whether one or more markings are applicable to one or more remaining unmarked non-welding time periods based on the one or more characteristic features, and In response to determining that a certain one of the one or more markings is applicable to a certain one of the one or more remaining unmarked non-welding time periods, the marking is associated with the unmarked non-welding time period.

21. A method for automatically marking a non-welding time period of a welding-related operation, the method comprising: capturing one or more characteristic characteristics of the welding-related operation via a user interface or one or more sensors over a period of time; identifying, via the processing circuitry, one or more unmarked non-welding time periods based on the one or more characteristic features; determining, using one or more machine learning models stored in the memory circuitry, whether one or more markings are applicable to the one or more unmarked non-welding time periods based on the one or more characteristic features; as well as In response to determining that a tag of the one or more tags is applicable to an unmarked non-welding time period of the one or more unmarked non-welding time periods, the tag is associated with the unmarked non-welding time period.

22. The method of claim 21, further comprising: associating the one or more indicia with one or more of the identified non-welding time periods based on the planned non-welding activity; determining, using the one or more machine learning models stored in the memory circuitry, whether one or more markings are applicable to one or more remaining unmarked non-welding time periods based on the one or more characteristic features; and In response to determining that a certain one of the one or more markings is applicable to a certain one of the one or more remaining unmarked non-welding time periods, the marking is associated with the unmarked non-welding time period.

Images