Improvements in and relating to ultrasound probes

Abstract

A method and apparatus for providing a weld inspection is disclosed wherein the method comprises: providing a weld inspection apparatus near a weld on a substrate to be inspected; an inspection is performed in which the substrate is provided at an elevated temperature condition above ambient temperature by heating during the inspection, the performing of the inspection comprising: emitting ultrasound waves into the volume of the substrate and the weld; receiving at least a portion of the ultrasonic waves returned from the substrate and the weld, thereby obtaining a plurality of signal sets; then processing one or more of the plurality of signal sets to provide weld inspection data; wherein the treatment comprises correcting a temperature profile within the volume of the substrate and / or weld at an elevated temperature.

Description

[0001] This application is a divisional application of application number 202280090723.8, filed on 22 December 2022, having the title “Improvements in and relating to ultrasonic probes”. TECHNICAL FIELD

[0002] The present invention relates to improvements in and relating to ultrasonic probes, processing of signals thereof and methods of use thereof in non-destructive ultrasonic testing, particularly but not exclusively in relation to high temperature deployment thereof. BACKGROUND

[0003] Ultrasonic testing is used for non-destructive testing of various objects. A transmitting transducer emits ultrasonic waves which enter the object, interact with the object and its sub-features, and then return to a receiving transducer. The temperature of the object has an effect on the speed of sound within the object, and so can have a negative effect on the quality of the testing and imaging.

[0004] When testing a weld, it is desirable to be able to perform ultrasonic testing soon after the weld or individual pass has been completed, in order to minimise the time taken to perform the testing and then make any required corrections to the weld or pass. However, this means performing the ultrasonic testing whilst the object or substrate is at an elevated temperature and there is a temperature variation between parts of the object or substrate. SUMMARY

[0005] It is a potential aim of the present invention to provide ultrasonic probes and processing of signals thereof which take into account the effects of temperature and variation of temperature with location within a substrate. It is a potential aim of the present invention to provide accurate imaging at a wide range of temperatures and to take into account different temperatures at different locations, particularly in high temperature environments.

[0006] According to a first aspect of the present invention, there is provided a method of providing a weld inspection, the method comprising:

[0007] a) providing a weld inspection apparatus in the vicinity of a weld on a substrate to be inspected;

[0008] b) performing an inspection, wherein the substrate is provided in an elevated temperature state above ambient temperature by heating during the inspection, the performance of the inspection comprising:

[0009] a. transmitting ultrasonic waves into a volume of the substrate and the weld;

[0010] b. receiving at least a portion of the ultrasonic waves returned from the substrate and the weld, thereby obtaining a plurality of signal sets;

[0011] c) processing one or more of the plurality of signal sets to provide weld inspection data;

[0012] The process includes temperature correction within the volume of the substrate and / or weld at elevated temperatures.

[0013] The method may include a uniformly elevated temperature throughout the substrate and / or an elevated temperature that includes a temperature distribution (e.g., a temperature gradient) in the substrate and / or weld and / or weld inspection apparatus.

[0014] The process includes correcting the temperature distribution (e.g., temperature gradient) within the volume of the substrate and / or weld at elevated temperatures.

[0015] The method may specify that the correction includes correcting the path of at least a portion of the ultrasonic wave through the weld inspection device and / or the substrate and / or the weld volume to provide a correction path.

[0016] The method can specify that the portion of the ultrasonic wave has path characteristics in one or more elements through which it passes. The method can specify that the portion of the ultrasonic wave in a weld inspection device has path characteristics in one or more dielectric elements. The method can specify that the portion of the ultrasonic wave in a substrate has path characteristics in one or more substrate elements. The method can specify that the portion of the ultrasonic wave in a weld has path characteristics in one or more weld elements. The method can specify that the portion of the ultrasonic wave at the interface between the weld inspection device and the substrate has path characteristics in one or more interface elements.

[0017] Ultrasonic waves can have path characteristics in each of multiple elements (e.g., elements adjacent to preceding and following elements). Ultrasonic waves can have path characteristics in each of one or more dielectric elements. Ultrasonic waves can have path characteristics in each of one or more interface elements. Ultrasonic waves can have path characteristics in each of one or more substrate elements. Ultrasonic waves can have path characteristics in each of one or more weld elements.

[0018] Ultrasonic waves can have path characteristics in each of one or more dielectric elements, and then in each of one or more substrate elements. Ultrasonic waves can have path characteristics in each of one or more interface elements (e.g., after the dielectric element and / or before the substrate element). Ultrasonic waves can have path characteristics in each of one or more weld elements (e.g., after the substrate element).

[0019] For example, in a return path, ultrasound can have path characteristics in each of one or more substrate elements, and then in each of one or more dielectric elements. Ultrasound can have path characteristics in each of one or more weld elements (e.g., before the substrate elements). Ultrasound can have path characteristics in each of one or more interface elements (e.g., after the substrate elements and / or before the dielectric elements).

[0020] The method may specify that, before the portion of the ultrasonic wave enters a first element (e.g., a first substrate element) of the substrate, the portion of the ultrasonic wave has path characteristics in a weld inspection device element (e.g., a dielectric element), and / or in an interface (e.g., an interface element). The method may specify that the first element has a temperature within a temperature distribution (e.g., a temperature gradient) for the volume of the substrate and / or weld, wherein a temperature-corrected path characteristic is determined for the portion of the ultrasonic wave in the first element, the temperature-corrected path characteristic being based on the temperature change between the weld inspection device element (e.g., a dielectric element and / or an interface element) and the first element of the substrate (e.g., the first substrate element).

[0021] The changes in path characteristics can be calculated using Snell's law of refraction.

[0022] The method may specify that, before the portion of the ultrasonic wave enters a second element (e.g., a second substrate element) of the substrate, the portion of the ultrasonic wave has path characteristics in a first element (e.g., a first substrate element). The method may also specify that the first element has a temperature within a temperature distribution (e.g., a temperature gradient) of the volume for the substrate and / or weld, and the second element has a temperature within the same temperature distribution (e.g., a temperature gradient) for the volume for the substrate and / or weld, wherein a temperature-corrected path characteristic is determined for the portion of the ultrasonic wave in the second element (e.g., the second substrate element), the temperature-corrected path characteristic being based on the temperature change between the first and second elements (e.g., between the first and second substrate elements).

[0023] The method may specify that, before the portion of the ultrasonic wave enters a second additional element (e.g., a second additional substrate element) of the substrate, the portion of the ultrasonic wave has path characteristics in a first additional element (e.g., a first additional substrate element). The method may also specify that the first additional element has a temperature within a temperature distribution (e.g., a temperature gradient) of the volume for the substrate and / or weld, and the second additional element has a temperature within the same temperature distribution (e.g., a temperature gradient) for the volume for the substrate and / or weld, wherein a temperature-corrected path characteristic is determined for the portion of the ultrasonic wave in the second additional element (e.g., the second additional substrate element), the temperature-corrected path characteristic being based on the temperature change between the first and second additional elements (e.g., between the first and second additional substrate elements).

[0024] The method may specify that temperature correction path characteristics are determined for each component of the ultrasonic wave that passes through the weld inspection device and / or the substrate and / or the weld (e.g., for each component that passes through the weld inspection device and / or the substrate and / or the weld).

[0025] The method can specify that temperature change is expressed as the change in the speed of sound between one element and the speed of sound in the next element.

[0026] The method may specify that ultrasonic waves are calibrated in multiple different portions of the weld inspection device and / or in the volume of the substrate and / or weld to provide a calibration path, for example, through the weld inspection device and / or the substrate and / or weld.

[0027] The method may specify multiple correction paths, such as continuously through a weld inspection device and / or the substrate and / or the weld and / or back again. The method may specify at least 5 correction paths, possibly at least 15 correction paths, potentially at least 25 correction paths, and optionally at least 40 correction paths, such as 64 correction paths.

[0028] It can provide a correction path for the individual ultrasonic beams emitted by a transducer (e.g., a 64-element phased array).

[0029] The method may specify the selection of a region of interest (ROI) within the volume of the substrate and weld through which the ultrasonic waves have passed, and subdivide this ROI into multiple locations, such as pixels. The method may apply signal correction to, for example, the locations of the pixels. Furthermore, the method may apply signal correction to individual locations within the ROI (e.g., individual pixels), the beam path to that location (e.g., the pixel), and / or the return beam path.

[0030] The method may specify determining signal correction based on, for example, the relationship between pixel position and at least one, preferably at least one pair, correction paths. This relationship may be geometric. It may be a weighted correction based on the relative geometric positions to one or more correction paths.

[0031] The method may specify determining signal correction based on, for example, the relationship between the position of a pixel and at least one, preferably at least one pair of positions on at least one, preferably at least one pair of correction paths. This relationship may be a geometric relationship. It may be a weighted correction based on the relative geometric positions (e.g., distances) between the position and at least one, preferably at least one pair of positions on at least one, preferably at least one pair of correction paths.

[0032] The method may include signal correction of one or more locations (e.g., pixels) traversed by a correction path of the emitted ultrasound beam used for transmission. Signal correction may be primarily or exclusively based on the correction path of the emitted ultrasound beam passing through that location (e.g., pixel). Signal correction may be based on a weighted correction relationship, which is based on the relative geometric location (e.g., distance) between the location and at least one, preferably at least one pair of locations on at least one, preferably at least one pair of correction paths, wherein signal correction may be primarily or exclusively based on the correction path of the emitted ultrasound beam passing through the location (e.g., pixel).

[0033] This method may include signal correction for one or more locations (e.g., pixels) not traversed by the correction path of the transmitted ultrasound beam. The method may also include signal correction for locations (e.g., pixels) not traversed by the correction path of the transmitted beam, based on calculated or observed corrections for locations traversed by the correction path of the transmitted beam. Signal correction may be based on calculated or observed locations for multiple (e.g., four) locations that the transmitted beam has already traversed.

[0034] Signal correction can be based on calculated or observed positions for multiple locations on a first transmit beam and multiple locations on a second transmit beam. The first and second transmit beams can be adjacent beams in a set of beams. The first beam may reach one side of the location requiring signal correction, while the second beam may reach the other side of the location requiring signal correction.

[0035] The signal correction for a location can be a weighted combination of signal corrections from one or more other locations (e.g., one or more other locations the transmitted beam has already passed through). The weighted combination can be based on four locations. Alternatively, it can be based on two locations on one beam and two locations on another beam.

[0036] Signal correction for a given location can be weighted based on a fraction of the distance between the location on the first beam and the location on the second beam. Alternatively, signal correction for a location can be weighted based on the ratio of the distance from the location to the location on the first beam to the distance from the location to the location on the second beam; for example, weighting corrections from these locations. This distance can be considered as an arc equidistant from the transducer, passing through the first beam, the location to be corrected, and the second beam.

[0037] Signal correction for a given location can be weighted based on a fraction of the distance between a first location on the first beam and a second location on the first beam. Signal correction for a location can also be weighted based on the ratio of the distance from the location to the first location on the first beam to the distance from the location to the second location on the first beam, for example, weighting corrections from these locations. This distance can be considered as being along the first beam.

[0038] Distance values ​​for regions of interest or even individual locations (e.g., pixels) within the substrate can be pre-calculated and stored.

[0039] It can calculate the velocity of sound in the region of interest or even at individual locations (e.g., pixels) within the substrate.

[0040] For example, the beam path can be calculated based on Snell's law of refraction, which tells the distance between the individual boundaries of locations (e.g., pixels) in the region of interest or even in the substrate.

[0041] Temperature can be mapped to a region of interest or even individual locations (e.g., pixels) within the substrate. Temperature profiles can be assigned to the region of interest or even the substrate, such as those perpendicular to the substrate surface.

[0042] This method may include determining the travel time through each location, for example, based on distance and speed within the area. The method may establish the net travel time as the sum of the travel times through all locations on the path.

[0043] The method may specify that the result set includes indications of one or more measurements of the geometry of the substrate and / or weld bevel and / or weld, wherein the method further includes a comparison of the measurement indications of the geometry with a modeling indication of the geometry, wherein imaging of the region of interest is accepted if the comparison of the measurement indications of the geometry with the modeling indications of the geometry determines that the measurement indications of the geometry are sufficiently fitted to the modeling indications of the geometry.

[0044] The method may specify that the result set includes indications of one or more measurements of the geometry of the substrate and / or weld bevel and / or weld, wherein the method further includes a comparison of the measurement indications of the geometry with a modeling indication of the geometry, wherein if the comparison of the measurement indications of the geometry with the modeling indications of the geometry determines that the measurement indications of the geometry are not adequately fitted to the modeling indications of the geometry, a temperature distribution for correcting the temperature distribution within the volume of the substrate and / or weld at elevated temperatures is predetermined.

[0045] The method may specify that it includes providing a thermal model and using the thermal model to generate a modeled temperature distribution location of at least a portion of a substrate in an elevated temperature state, the portion of the substrate including a volume. The method may also specify that it includes measuring temperatures at multiple locations while the substrate is in an elevated temperature state to obtain measured temperature distribution locations, and further includes comparing the measured temperature distribution locations with modeled temperature distribution locations.

[0046] The method may specify that it further includes, if the comparison between the measured temperature distribution location and the modeled temperature distribution location determines that the modeled temperature distribution is not a sufficient fit to the measured temperature distribution, then the thermal model and / or the modeled temperature distribution location are corrected, and then the comparison is repeated.

[0047] The method may specify that it further includes calculating the characteristics of the ultrasonic waves emitted during passage of at least a portion of the substrate and / or weld if a comparison between the measured temperature distribution location and the modeled temperature distribution location determines that the modeled temperature distribution is a sufficient fit to the measured temperature distribution.

[0048] The first aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0049] According to a second aspect of the present invention, an apparatus for providing weld inspection is provided, the apparatus comprising:

[0050] a) A weld inspection device, which is placed near the weld on a substrate to be inspected during use, wherein the substrate is provided in an elevated temperature state above ambient temperature by heating during inspection;

[0051] b) Weld inspection devices are suitable for:

[0052] 1) The ultrasonic waves are emitted into the volume of the substrate and the weld;

[0053] 2) Receive at least a portion of the ultrasonic waves returning from the substrate and the weld, thereby acquiring multiple signal sets;

[0054] 3) The weld inspection apparatus includes one or more processors, which include a processor having inputs for one or more of a plurality of signal sets, the processor providing corrections to one or more of the input signal sets to provide corrected weld inspection data, wherein the processor applies temperature corrections to the volume of the substrate and / or weld at elevated temperatures.

[0055] The second aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0056] According to a third aspect of the present invention, a method is provided for determining one or more dimensions of one or more substrates to be welded, the method comprising:

[0057] i. Provide a device for determining the size of emitted ultrasonic waves;

[0058] ii. Introduce one or more substrates to be welded into the dimensional determination device;

[0059] iii. Use ultrasound to determine one or more dimensions of one or more substrates;

[0060] iv. Using one or more dimensions to determine one or more additional actions performed on one or more substrates.

[0061] This method may be part of a welding method and may include the step of providing a welding apparatus. The method may be an arc welding method. The method may include introducing one or more substrates to be welded into the welding apparatus. One or more additional actions may be part of the welding process and / or part of a post-weld process.

[0062] Ultrasound can be provided by one or more beams emitted from the device into one or more substrates. Ultrasound can be provided by multiple beams. Ultrasound can be provided by multiple beams dispersed in an arc. The arc can extend at least 5°, possibly at least 10°, and possibly at least 15°. The arc can extend less than 40°, for example less than 30° or possibly less than 25°.

[0063] One or more dimensions of one or more substrates can be the spacing between a first surface and a second surface, such as thickness. The thickness can be established in the direction of one or more ultrasonic beams. This thickness can be established perpendicular to one or more surfaces of the one or more substrates.

[0064] The dimensional determination device can also be a weld inspection device. Ultrasonic technology can be used in dimensional determination devices and / or weld inspection devices.

[0065] The method may include the step of using heating to raise the temperature of one or more substrates above ambient temperature.

[0066] The step of using heating to raise the temperature of one or more substrates above ambient temperature may be provided before one or more dimensions of the one or more substrates are determined. Alternatively, the step of using heating to raise the temperature of one or more substrates above ambient temperature may be provided after one or more dimensions of the one or more substrates are determined.

[0067] This method may include determining one or more dimensions of one or more substrates at multiple locations on the substrate. The method may include moving (e.g., rolling) the dimension determining device on the substrate between multiple locations. The multiple locations on the substrate may follow a predetermined trajectory for a weld and / or a weld formation trajectory.

[0068] One or more dimensions can be measured at an angle relative to the substrate. The angle relative to the substrate can be considered as the angle relative to the first surface and / or the second surface of the substrate. This angle can be 90°+ / -25°, for example 90°+ / -20°, or possibly 90°+ / -15°.

[0069] One or more dimensions may be measured at an angular range relative to the substrate. The angular range relative to the substrate may be considered as the angular range relative to the first and / or second surfaces of the substrate. The angular range may be 90°+ / - up to 25°, for example 90°+ / - up to 20°, possibly 90°+ / - up to 15°, and particularly 90°+ / - up to 10°.

[0070] One or more dimensions can be measured perpendicular to one or more of the first and / or second surfaces of the substrate.

[0071] The method may include correcting thickness measurements for the effects of temperature and / or temperature gradients within the substrate on ultrasound (e.g., on individual ultrasound beam paths).

[0072] This method may include correcting thickness measurements for the effect of ultrasound refraction on, for example, the individual ultrasound beam paths.

[0073] The method may include providing a weld inspection device.

[0074] This method may include inspecting the generated weld using a weld inspection device. The method may include weld inspection being performed by the inspection device at an inspection location on one or more substrates, wherein the inspection location is at an elevated temperature above ambient temperature.

[0075] One or more dimensions may include the dimensions of the substrate along one or more paths through the substrate, which will be inspected by and / or are being inspected by the weld inspection device.

[0076] The combined dimensional determination device and weld inspection device can perform dimensional determination for the substrate, followed by one or more weld inspections. The combined dimensional determination device and weld inspection device can perform one or more dimensional determinations, followed by one or more weld inspections, then further dimensional determinations, possibly followed by further weld inspections. A cyclical method of dimensional determination, weld inspection, dimensional determination, and weld determination can be employed.

[0077] This method may include using one or more dimensional determinations in one or more additional actions performed on one or more substrates by using one or more dimensional determinations in adjustments and / or corrections for weld inspection. This method may include adjustments to the depth of focus. This method may include adjustments within an image analysis region (e.g., an image analysis region considered to be inspected). This method may include using one or more dimensional determinations in adjustments and / or corrections to positions determined by weld inspection. This method may include using one or more dimensional determinations in adjustments and / or corrections to the depth of penetration into the substrate determined by weld inspection.

[0078] The third aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0079] According to a fourth aspect of the invention, an apparatus is provided for optionally performing dimensional determination during welding, the apparatus comprising:

[0080] i. Optional welding equipment;

[0081] ii. Size determining device;

[0082] The size determination device includes an ultrasonic transmitter and a receiver, the element of the size determination device is provided with a substrate contact surface, and the element of the size determination device provides an output data signal to another element of the size determination device, wherein the other element of the size determination device includes a determiner for determining one or more dimensions of the substrate, and wherein the determiner provides one or more dimensions to another processor, which uses one or more dimensions in one or more additional actions performed with respect to one or more substrates.

[0083] The apparatus may also include a weld inspection device. The apparatus may include a combination of a dimensional determination device and a weld inspection device.

[0084] The determiner and another processor can be part of a common data processing component of the combined size determination device and weld inspection device.

[0085] The fourth aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0086] According to a fifth aspect of the present invention, a method for inspecting welds is provided, the method comprising:

[0087] a) Provide a weld inspection device near the weld on the substrate to be inspected;

[0088] b) Emitting ultrasound into the substrate and weld, and receiving ultrasound reflected from the substrate and weld to acquire multiple signal sets, including a first signal set and a second signal set;

[0089] c) Process at least the first signal set and the second signal set to provide weld inspection data; wherein

[0090] The first signal set includes a first series of first data elements, the first series of data elements spanning a first time period, each first data element including a first variable value for a variable and a first time value for that first variable value, the first time period including a first sub-time period;

[0091] The second signal set includes a second series of second data elements, which span a second time period. Each second data element includes a second variable value for a variable and a second time value for that second variable value. The second time period includes a second sub-time period.

[0092] The first and second sub-time periods cover the same time value;

[0093] The processing of the first and second signal sets determines the modified signal set that contributes to the weld inspection data, wherein the modified signal set includes: a modified sub-time period that covers the same time values ​​as the first and second sub-time periods; if the first and second variable values ​​are within a given relationship with each other, then for each time value, it includes an expression of the first variable value and an expression of the second variable value of the time value; and / or if the first and second variable values ​​are not within a given relationship with each other, then for each time value, it includes either an expression of the first variable value or an expression of the second variable value of the time value.

[0094] The first variable value can be, for example, the amplitude at a given time, particularly an expression of the amplitude of the rectified signal set.

[0095] The second variable value can be, for example, an expression of the amplitude at a given time value, particularly the amplitude of the rectified signal set.

[0096] The first time period, the second time period, and / or one or more additional time periods can be of the same duration. The first time period, the second time period, and / or one or more additional time periods can be of the same duration plus or minus 25%. The first time period, the second time period, and one or more additional time periods can be the duration of the receiver signal used for the ultrasonic beam or wave.

[0097] The first sub-time period and / or the second sub-time period and / or one or more additional sub-time periods can be individual time values. The first sub-time period and / or the second sub-time period and / or one or more additional sub-time periods can be a small set of individual time values.

[0098] The first time period can be considered as multiple distinct or partially overlapping first sub-time periods. The second time period can be considered as multiple distinct or partially overlapping second sub-time periods. Other time periods can be considered as multiple distinct or partially overlapping additional sub-time periods. When multiple sub-time periods are used for the same time value, they can be paired with each other.

[0099] According to this method, the modified signal set can be compiled by considering multiple sub-time periods.

[0100] The method can specify that, in the modified signal set, the expression of the first variable value for that time value and the expression of the second variable value for that time value can be the average of the first variable value and the second variable value. The method can also specify that, in the modified signal set, the expression of the first variable value for that time value and / or the expression of the second variable value for that time value, together with other variable values, can be the average of all variable values ​​within a given relation.

[0101] The method can specify that if the values ​​of the first and second variables are within a given threshold, then the values ​​of the first and second variables are within a given relationship to each other. The nature and / or definition function and / or value of the given threshold can be variable. The threshold can be varied, particularly increased, in the case of increased system gain and / or increased noise in the signal set. The threshold can also be varied, particularly increased, in the case of increased interference levels in the signal set. The nature and / or definition function and / or value of the threshold can be determined in the calibration method.

[0102] The method can specify that if the values ​​of the first and second variables are outside another given threshold, which may be the same threshold, then the values ​​of the first and second variables are not within a given relationship to each other. The nature and / or definition function and / or value of the given threshold can be variable. In cases where system gain increases and / or noise increases in the signal set, the given threshold can be varied, particularly increased. In cases where the interference level in the signal set increases, the given threshold can be varied, particularly increased.

[0103] The method may further specify that, when the first variable value and the second variable value are not within a given relationship to each other, the included expression selected from the first and second expressions is the expression with the lower variable value for that time value. The method may also specify that, when the variable value and the compared variable value are not within a given relationship to each other, the included expression is the expression with the lower variable value for that time value. The lower variable value may be the lower variable value considered as an absolute term.

[0104] A given threshold can be determined relative to the magnitude of the first and second variable values.

[0105] The amplitude can be a predetermined amplitude value. The amplitude can be, for example, a expected amplitude value relative to the expected noise amplitude. The amplitude can be a ratio or multiple of the defined expected noise amplitude.

[0106] The amplitude can be the observed amplitude.

[0107] The magnitude can refer to the maximum magnitude observed in a set of variable values ​​that are considered undisturbed.

[0108] The magnitude can refer to the minimum value observed over a time or period across all considered variable values ​​(e.g., across the first variable value, the second variable value, and one or more additional variable values). The magnitude can be the minimum value plus a factor. This magnitude can be considered as a disturbance-free threshold.

[0109] A threshold can be the difference between an observed value and an analyzed value. Observed values ​​and analyzed values ​​can be magnitudes. An analyzed value can be obtained from the minimum observed over a time period across all considered variable values ​​(e.g., across the values ​​of a first variable, a second variable, and one or more additional variables). An analyzed value can be the minimum value plus a factor.

[0110] The method may also include acquiring one or more additional signal sets as part of a plurality of signal sets.

[0111] The method may also include processing one or more other signal sets to provide weld inspection data.

[0112] The method may also include one or more additional signal sets, which include additional sets of additional data elements, each additional set of additional data elements may span additional time periods, each additional data element includes additional variable values ​​for the variable and additional time values ​​for the additional variable values, each additional time period includes additional sub-time periods.

[0113] The method may also include an additional sub-time period that covers the same time value as at least one of the first sub-time period and / or the second sub-time period.

[0114] The method may also include processing one or more additional signal sets to determine modified signal sets that contribute to weld inspection data.

[0115] The method may also include modifying the signal set across a modified sub-time period, the modified sub-time period covering the same time value as another sub-time period and one or both of the first and second sub-time periods.

[0116] The method may also include modifying the expression of the signal set for a time value to include the expression of the first variable value, the expression of the second variable value, and the expression of another variable value of the time value if the first variable value, the second variable value, and another variable value of the time value are within a given relationship between them.

[0117] The method may alternatively or additionally include, if the first variable value, the second variable value, and the third variable value are not within a given relationship to each other, the modified signal set for the time value includes one of the expression of the first variable value of the time value, the expression of the second variable value of the time value, or the expression of another variable value.

[0118] The method may also alternatively or additionally include, for two variable values ​​within a given relationship with each other, the modified signal set for the time value including two of the following expressions: an expression for a first variable value for that time value; an expression for a second variable value for that time value; and an expression for a further variable value for that time value.

[0119] The method may also alternatively or additionally include a modified signal set that includes only one of the following expressions for the time value: an expression for a first variable value for the time value; an expression for a second variable value for the time value; or an expression for a further variable value for the time value; wherein the variable values ​​are not within a given relationship to each other.

[0120] The method may also specify that the modified signal set is compiled from multiple modified sub-time periods, each modified sub-time period covering the same time value as the first sub-time period and / or the second sub-time period and / or one or more other sub-time periods.

[0121] The fifth aspect of the invention may include any features, options or possibilities set forth elsewhere in this document (including in other aspects of the invention).

[0122] According to a sixth aspect of the present invention, an apparatus for inspecting welds is provided, the apparatus comprising:

[0123] a) Weld inspection device, which includes:

[0124] a. A transmitter for ultrasound entering the substrate and weld during use, and a receiver for ultrasound returning from the substrate and weld during use, the receiver being connected to a processor to provide the processor with multiple signal sets including the acquisition of a first signal set and a second signal;

[0125] b. The processor is adapted to receive multiple acquired signal sets, including a first signal set and a second signal set; wherein the first signal set includes a first series of first data elements, the first series of data elements spanning a first time period, each first data element including a first variable value for a variable and a first time value for the first variable value, and the first time period including a first sub-time period; the second signal set includes a second series of second data elements, the second series of second data elements spanning a second time period, each second data element including a second variable value for a variable and a second time value for the second variable value, and the second time period including a second sub-time period; the first sub-time period and the second sub-time period cover the same area. The processor is adapted to use a first signal set and a second signal set to determine a set of modified signals that contribute to the weld inspection data, wherein the processor is adapted to provide that the modified signal set for a modified sub-time period covering the same time value as the first sub-time period and the second sub-time period includes: if the first variable value and the second variable value are within a given relationship with each other, then for the time value, it includes an expression of the first variable value of the time value and an expression of the second variable value of the time value; and / or if the first variable value and the second variable value are not within a given relationship with each other, then for the time value, it includes either an expression of the first variable value of the time value or an expression of the second variable value of the time value.

[0126] The sixth aspect of the invention may include any features, options or possibilities set forth elsewhere in this document (including in other aspects of the invention).

[0127] According to a seventh aspect of the present invention, a welding method is provided, the welding method comprising:

[0128] a) Provide welding equipment;

[0129] b) Offers multiple sensor types;

[0130] c) Define the first set of welding conditions for the welding method;

[0131] d) Introduce one or more substrates to be welded into the welding apparatus;

[0132] e) To perform welding on one or more substrates;

[0133] f) Acquire data from multiple sensor types during welding;

[0134] g) Provide the correlation between data obtained from at least two selected sensor types out of a plurality of sensor types;

[0135] h) Synchronize one or more data points from data obtained from at least two selected sensor types with one or more data points from data obtained from the other of at least two selected sensor types;

[0136] i) Compare data obtained from multiple sensor types with reference data for one or more of the sensor types;

[0137] j) Based on one or more such comparisons, determine whether the weld has acceptable or unacceptable quality;

[0138] k) Wherein, if the weld has unacceptable quality, the method includes then taking one or more actions.

[0139] The seventh aspect of the invention may include any features, possibilities, or options related to weld inspection that are set forth in the first and / or second aspects of the invention.

[0140] The method can be electric arc welding.

[0141] The method can specify that the correlation is both time correlation and location correlation.

[0142] The method can be specified that the correlation is time-related.

[0143] The method can specify that data from at least two selected sensor types includes the occurrence time of one or more data points within the data. The method can also specify that correlation is the matching occurrence time of data points from at least two selected sensor types.

[0144] The method may specify that data from at least two selected sensor types include multiple times that provide time correlation.

[0145] The method can specify that data from at least two selected sensor types includes the occurrence time of all data points within the data. The method can also specify that data from at least two selected sensor types includes the occurrence time of all data points within various ranges (e.g., bandwidths) into which the data is subdivided.

[0146] The method can specify that, for each of two or more selected sensor types providing correlation, the occurrence time is obtained relative to a timestamp introduced into the data from the selected sensor type. The method can specify, for example, that the occurrence time is obtained directly from the timestamp of the time that coincides with the occurrence time. The method can specify that the occurrence time is obtained by combining the time elapsed since the timestamp with the time of the data point, for example, in the case of a timestamp occurrence, and by calculating the occurrence time based on the time elapsed since the timestamp when the occurrence time is reached.

[0147] Timestamps can be applied by processors, such as those including clock generators and / or processors that provide clock signals.

[0148] The method can be specified that the correlation is a location correlation.

[0149] The method can specify that data from at least two selected sensor types includes multiple locations that provide location correlation.

[0150] The method can specify that data from at least two selected sensor types includes the occurrence positions of one or more data points within the data. The method can also specify that correlation is the matching occurrence positions of data points from at least two selected sensor types.

[0151] The method can specify that data from at least two selected sensor types includes the location of all data points within the data. The method can also specify that data from at least two selected sensor types includes the location of all data points within various ranges (e.g., bandwidths) into which the data is subdivided.

[0152] The method can specify that, for each of two or more selected sensor types providing correlation, the occurrence position is obtained relative to a location stamp introduced into the data from the selected sensor type. The method can specify, for example, that the occurrence position is obtained directly from the location stamp at the same location as the occurrence position. The method can specify that the occurrence position is obtained by combining the time elapsed since the location stamp with the time of the data point, for example, in the case of a location stamp, and calculated based on the time elapsed since the location stamp when the occurrence position is reached and / or based on the distance traveled since the location stamp when the occurrence position is reached.

[0153] Positional dependence can be relative to a position associated with one or more substrates being welded, such as a position relative to one or more substrates. Multiple positional dependence can be provided simultaneously relative to one or more substrates. The position can have a known relationship relative to the location where welding is performed (e.g., relative to the start of the weld and / or relative to the end of the weld).

[0154] Positional correlation can be relative to a position associated with the welding apparatus, such as a position relative to a position on the welding apparatus. Multiple positional correlations can be provided simultaneously relative to the welding apparatus. One or more positions on the welding apparatus may be close to the welding electrode. The position may have a known relationship relative to the position where welding is performed (e.g., relative to the start and / or end of the weld), potentially the known relationship being the separation of the electrode and the substrate.

[0155] Positional correlations can be provided relative to a location associated with one or more substrates to be welded (e.g., a location relative to a location on one or more substrates) and relative to a location associated with the welding apparatus (e.g., a location relative to a location on the welding apparatus).

[0156] The location stamp can be applied by a processor. The processor can receive one or more location signals. The location signals can be emitted by one or more location sensors and / or location indicators. The location sensor can be a vision-based sensor, such as a camera. The location indicator can be a motion detection-based location indicator, such as an incremental encoder, which provides, for example, the occurrence of motion and / or the direction of motion and / or the encoder position. A location indicator based on a tracking signal can be a radio frequency identification (RFID) based system, possibly having one or more RFID tags on the substrate and / or welding apparatus.

[0157] This method may include providing at least one type of correlation between data obtained from at least three, potentially at least four, and possibly at least five selected sensor types. The same type of correlation may be provided between at least two selected sensor types. The same type of correlation may be provided between all selected sensor types.

[0158] This method may include providing at least two different types of correlations between data obtained from at least three, potentially at least four, and possibly at least five selected sensor types from a plurality of sensor types. Different types of correlations may be provided between at least two selected sensor types. Different types of correlations may be provided between all selected sensor types. The different types may include temporal correlations and location correlations.

[0159] The method may include synchronizing one or more data points from data acquired from various selected sensor types. The method may also include using the same one or more correlations to synchronize one or more data points from data acquired from each of at least two, potentially at least three, and possibly at least four selected sensor types.

[0160] The method may include synchronizing at least 10% of data points from data obtained from at least one selected sensor type with at least 10% of data points from data obtained from at least another selected sensor type.

[0161] The method may specify that the method includes at least one of a plurality of sensor types, potentially selected from at least one of the following sensor types: voltage sensor, current sensor, welding arc acoustic emission sensor, weld topology sensor, weld imaging sensor, and ultrasonic imaging sensor.

[0162] A voltage sensor can be one of several sensor types. A voltage sensor can be one of the selected sensor types.

[0163] A current sensor can be one of several sensor types. A current sensor can be one of the selected sensor types.

[0164] The welding arc acoustic emission sensor can be one of several sensor types. It can be a selected sensor type. The welding arc acoustic emission sensor can be an acoustic sensor, such as a microphone. The welding arc acoustic emission sensor can be sensitive to emissions from the electric arc forming the weld and / or the interaction between the arc and the substrate and / or the interaction between the arc and the shielding gas. The welding arc acoustic emission sensor can provide data indicating characteristics related to welding speed and / or weld location sidewall arc initiation and / or weld location sidewall fusion and / or shielding gas flow rate.

[0165] A weld topology sensor can be one of several sensor types. It can be a selected sensor type. It can be a radiation-based sensor. It can be an image-based sensor. It can be a laser sensor. It can be sensitive to radiation applied to and / or reflected from the substrate and / or weld. The applied and / or reflected radiation can be focused by the weld topology sensor. It can inform characteristics related to the geometry of the substrate and / or weld.

[0166] A weld seam imaging sensor can be one of several sensor types. It can be a selected sensor type. It can be a camera-based sensor. The weld seam imaging sensor provides data that informs characteristics related to weld pool size and / or weld pool geometry and / or weld pool temperature and / or the presence and / or pattern of deposited material on the weld seam and / or substrate and / or shape of deposited material on the weld seam and / or substrate and / or the presence of anomalies on the weld seam and / or substrate.

[0167] An ultrasound imaging sensor can be one of several sensor types. An ultrasound imaging sensor can be one of a selected sensor type.

[0168] One or more sensor types can provide raw datasets. Methods can include processing the raw dataset to produce a processed dataset.

[0169] The determination of a single sensor type can take into account both raw and / or processed data. It can also consider the distance of a data point to a known set of data points (e.g., a set of known data points indicating acceptable welding conditions). This known set of data points can be represented by a single element (e.g., the average value of the known set of data points).

[0170] A known set of data points can be represented by a principal component analysis (PCA) model of the sensor type and its signal and / or data values.

[0171] The distance can be Mahalanobis distance.

[0172] Distance can be expressed as outlier scores.

[0173] This determination can be based on the distance of the data point relative to a threshold, such as threshold distance. It can also be based on whether the data point is inside or outside the boundary of a known set of data points.

[0174] The novelty detection model can be used to determine the sensor type by comparing the data points or signal values ​​with a previously developed principal component analysis (PCA) model and its signal and / or data values.

[0175] Processing the original dataset can include applying noise reduction algorithms. The original dataset can be processed to divide it into a series of bandwidths.

[0176] This determination can be made using a model (such as a PCA model with acceptable performance) to determine subsequent data points and / or signals.

[0177] When data points and / or datasets that are determined to be acceptable welds and / or unacceptable welds are added, the model and / or the model's mean can be recalculated.

[0178] One or more sensor types can provide raw datasets, wherein the method includes processing the raw datasets to give processed datasets that utilize the model. The method may include using different variants of the model for different parts of the welding process. Specifically, a variant of the model may be used for one or more weld passes, and a different model may be used for one or more additional weld passes throughout the weld formation process. Different variants of the model can be used for individual weld passes throughout the weld formation process. One or more additional different variants may be used for repair welds within the weld formation process, for example, in cases where a portion of the weld is removed and then re-welded. A separate additional different variant of the model can be used for individual weld passes or weld beads throughout the entire welding process undergoing repair welds.

[0179] The model can be an acoustic signal processing model. The model can be a voltage and / or current and / or power (e.g., output power) processing model.

[0180] A model can include multiple different data sources or sub-models.

[0181] The method may specify that synchronization provides an overall data structure in which data points from at least two selected sensor types are aligned with each other in time and / or location.

[0182] The method may specify that data points from at least two selected sensor types are displayed to the user, wherein the data points are aligned with respect to the time of occurrence and / or the location of occurrence.

[0183] The method may specify that data from two or more selected sensor types and related data from at least two sensor types are displayed to the user and / or stored.

[0184] The method may specify determining whether the weld has acceptable or unacceptable quality, and / or wherein one or more actions are shown and / or stored.

[0185] The method may be specified to include providing correlations between data obtained from at least four selected sensor types out of a plurality of sensor types.

[0186] The method may include inspecting the entire weld. The method may include inspecting multiple welds, for example, before all weld passes are completed, or after each weld pass and before the next weld pass.

[0187] The welding apparatus can be mounted on any autonomous, automated deployment system. It can be mounted on a rail system or a column and boom system. It can be mounted on a robotic arm (e.g., a multi-axis arm). The welding apparatus can be an arc welding machine.

[0188] The weld inspection device can be one of several sensor types. The weld inspection device can be one of a selected sensor type. The weld inspection device can use ultrasound, such as phased array ultrasound.

[0189] Weld inspection devices can be mounted on any autonomous, automated deployment system. Welding devices can be mounted on rail systems or column and boom systems. Welding devices can be mounted on robotic arms (e.g., multi-axis arms). Weld inspection devices can be phased array ultrasonic transducers. Weld inspection devices can include ultrasonic transmitters and receivers. Weld inspection devices can include substrate contact surfaces. Weld inspection devices can physically contact the inspection location. The substrate contact surface can physically contact the inspection location.

[0190] Methods may include rolling a weld inspection device (e.g., its substrate contact surface) on the surface of a substrate, for example, rolling it to one side of the weld, but possibly rolling it parallel to the weld.

[0191] The weld inspection apparatus may be equipped with internal cooling. Methods may include supplying coolant to and / or removing coolant from the weld inspection apparatus.

[0192] The method may further specify that at least one of a plurality of sensor types, potentially selected from at least one of the sensor types, is part of a weld inspection apparatus, and the method includes the step of inspecting the weld using the weld inspection apparatus. The method may specify that it includes the step of inspecting the weld using the weld inspection apparatus to determine whether a welding defect exists at the weld location, particularly at a continuous series of locations forming the weld. The method may specify that it includes the step of inspecting the weld using the weld inspection apparatus to determine one or more characteristics of a defect at a location. The method may specify that the characteristics include one or more of the following: dimensions relating to the geometry of the weld and / or dimensions relating to the length of the weld; location; defect type; defect shape; or defect position.

[0193] This method may include determining in real time whether a welding defect exists at the welding location. The method may include determining whether a welding defect exists at the welding location within 100 milliseconds, potentially within 50 milliseconds, and possibly within 20 milliseconds after the welding apparatus moves away from the welding location. The method may be specified to further include comparing one or more characteristics with one or more criteria, and further including determining whether a weld with a defect at the location meets weld criteria.

[0194] The method may include determining in real time whether a defective weld at a location meets weld standards. The method may include determining whether a defective weld at a location meets weld standards within 100 milliseconds, potentially within 50 milliseconds, and possibly within 20 milliseconds after the welding apparatus moves away from the location.

[0195] The method may specify that, if the weld meets weld criteria, a record of the weld may be created and stored for the location of defects relative to the weld geometry and / or relative to the included weld length. The method may also specify that the record includes data from one or more of multiple sensor types. The method may specify that, if the weld does not meet weld criteria, one or more remedial steps are applied to the weld.

[0196] The method may specify that at least one of a plurality of sensor types, potentially selecting at least one of the sensor types, is a weld condition sensor, and the method includes the step of using the weld condition sensor to inspect the weld condition.

[0197] The method may be specified to include steps of inspecting the weld condition to determine one or more parameters of the weld during its formation.

[0198] The method may be specified to include comparing one or more parameters with one or more control parameters, and to further include determining whether the risk level of weld defects exceeds the limit.

[0199] The method may specify that it includes one or more actions, the one or more actions being to change the welding conditions from a first set of welding conditions used for the welding method. The method may specify that the change in welding conditions from the first set of welding conditions includes stopping welding and / or warning the operator. The method may specify that the change in welding conditions from the first set of welding conditions includes changing the welding conditions back to the first set of conditions and / or changing the welding conditions to a second set of conditions.

[0200] The welding method may also include: inspecting the weld formed by the welding method. The welding method may also include:

[0201] i. Provide weld inspection equipment;

[0202] ii. Using heating to raise the temperature of one or more substrates above ambient temperature;

[0203] iii. Welding one or more substrates at elevated temperatures above ambient temperature using a welding apparatus;

[0204] iv. Inspect the generated welds using a weld inspection device;

[0205] v. Wherein, the inspection of the weld is provided by an inspection device at an inspection location on one or more substrates, wherein the inspection location is at an elevated temperature above the ambient temperature.

[0206] The seventh aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0207] According to an eighth aspect of the present invention, an apparatus for monitoring welding is provided, the apparatus comprising:

[0208] a) Multiple inputs for data from multiple sensor types;

[0209] b) One or more processors, wherein the processor is from one or more processors:

[0210] a. Receive input;

[0211] b. Process data obtained from at least two selected sensor types out of a plurality of sensor types to apply the correlation between the data obtained from at least two selected sensor types out of a plurality of sensor types;

[0212] c. Use correlation processing to process data obtained from at least two selected sensor types out of a plurality of sensor types to synchronize one or more data points from data obtained from one of the at least two selected sensor types with one or more data points from data obtained from the other of the at least two selected sensor types;

[0213] c) One or more outputs used to process the data.

[0214] The processor may also provide a comparator for receiving and comparing data from one or more of multiple sensor types (e.g., one or more of selected sensor types) and reference data from one or more sensor types.

[0215] The comparator can output a determination on whether the weld has acceptable or unacceptable quality based on the compared data. If the weld is determined to have unacceptable quality, the device can also provide a control signal to trigger one or more actions via a control unit.

[0216] The device may also include a control unit, for example for receiving a first set of welding conditions for the welding method and / or for receiving control signals to trigger one or more actions via the control unit, such as modification of the first set of welding conditions or termination of welding.

[0217] The apparatus may be specified to include a weld inspection device that determines one or more characteristics of a defect.

[0218] The device can provide input from at least one of a plurality of sensor types, potentially selecting at least one of sensor types, wherein the sensor type and / or selected sensor type is selected from: voltage sensor, current sensor, welding arc acoustic emission sensor, weld topology sensor, weld imaging sensor and ultrasonic imaging sensor.

[0219] A voltage sensor can be one of several sensor types. A voltage sensor can be one of the selected sensor types.

[0220] A current sensor can be one of several sensor types. A current sensor can be one of the selected sensor types.

[0221] The welding arc acoustic emission sensor can be one of several sensor types. It can be a selected sensor type. The welding arc acoustic emission sensor can be an acoustic sensor, such as a microphone. The welding arc acoustic emission sensor can be sensitive to emissions from the electric arc forming the weld and / or the interaction between the arc and the substrate and / or the interaction between the arc and the shielding gas. The welding arc acoustic emission sensor can provide data indicating characteristics related to welding speed and / or weld location sidewall arc initiation and / or weld location sidewall fusion and / or shielding gas flow rate.

[0222] A weld topology sensor can be one of several sensor types. It can be a selected sensor type. It can be a radiation-based sensor. It can be an image-based sensor. It can be a laser sensor. It can be sensitive to radiation applied to and / or reflected from the substrate and / or weld. The applied and / or reflected radiation can be focused by the weld topology sensor. It can inform characteristics related to the geometry of the substrate and / or weld.

[0223] A weld seam imaging sensor can be one of several sensor types. It can be a selected sensor type. It can be a camera-based sensor. The weld seam imaging sensor provides data that informs characteristics related to weld pool size and / or weld pool geometry and / or weld pool temperature and / or the presence and / or pattern of deposited material on the weld seam and / or substrate and / or shape of deposited material on the weld seam and / or substrate and / or the presence of anomalies on the weld seam and / or substrate.

[0224] An ultrasound imaging sensor can be one of several sensor types. An ultrasound imaging sensor can be one of a selected sensor type.

[0225] The apparatus may specify, further specifying, that the comparator includes a first comparator for receiving one or more characteristics and comparing them with one or more criteria, wherein the first comparator outputs a first determination regarding whether a weld with defects meets weld criteria. The apparatus may also specify, further specifying, that the comparator includes a second comparator for receiving one or more parameters of the weld during formation and comparing them with one or more control parameters, wherein the second comparator outputs a second determination regarding whether a risk level for weld defects is exceeded. The apparatus may specify that if the risk level for weld defects is exceeded, a control signal provided by the apparatus is sent to a controller to trigger one or more actions, wherein the one or more actions are changes to welding conditions from a first set of welding conditions for the welding method, for example, to stop welding and / or warn the operator.

[0226] The eighth aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0227] According to a ninth aspect of the present invention, a welding method is provided, the welding method comprising:

[0228] a) Provide welding equipment;

[0229] b) Offers multiple sensor types;

[0230] c) Define the first set of welding conditions for the welding method;

[0231] d) Introduce one or more substrates to be welded into the welding apparatus;

[0232] e) To perform welding on one or more substrates;

[0233] f) Acquire data from multiple sensor types during welding;

[0234] g) Compare data obtained from multiple sensor types with reference data for one or more of the sensor types;

[0235] h) Based on one or more such comparisons, determine whether the weld has acceptable or unacceptable quality;

[0236] i) wherein the acquired data is obtained by processing data from multiple sensor types, a computer model is used to process the acquired data from at least one of the multiple sensor types, and wherein the method includes using different variations of the model for different parts of the welding process.

[0237] Specifically, variations of the model can be used for one or more weld passes, and different models can be used for one or more additional weld passes throughout the weld formation process. Different variations of the model can be used for individual weld passes throughout the weld formation process. One or more additional variations can be used for repair welding during the weld formation process, for example, in cases where a portion of the weld is removed and then re-welded. Individual additional variations of the model can be used for individual weld passes or weld beads throughout the entire welding process undergoing repair welding.

[0238] The model can be an acoustic signal processing model. The model can be a voltage and / or current and / or power (e.g., output power) processing model.

[0239] A model can include multiple different data sources or sub-models.

[0240] The ninth aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0241] According to a tenth aspect of the present invention, an apparatus for monitoring welding is provided, the apparatus comprising:

[0242] a) Multiple inputs for data from multiple sensor types;

[0243] b) One or more processors, wherein the processor is from one or more processors:

[0244] a. Receive input from at least one of multiple sensor types;

[0245] b. Processing data to provide data acquired for at least one of multiple sensor types, the processor having access to a computer model, and wherein the method includes using different variations of the model for different parts of the welding process;

[0246] c) One or more outputs used to process the data.

[0247] The tenth aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0248] According to an eleventh aspect of the present invention, a welding method is provided, the welding method comprising:

[0249] a) Provide welding equipment;

[0250] b) Offers multiple sensor types;

[0251] c) Introduce one or more substrates to be welded into the welding apparatus;

[0252] d) To perform welding on one or more substrates;

[0253] e) Acquire data from multiple sensor types during welding;

[0254] f) Compare data obtained from multiple sensor types with reference data for one or more of the sensor types;

[0255] g) Determine whether the weld quality is acceptable or unacceptable;

[0256] h) Wherein, determining includes considering combined data from at least two of multiple sensor types.

[0257] The method can be defined as forming combined data, and then determining the outcome by considering the combined data.

[0258] The method may specify a first determination, namely, determining whether the weld has acceptable or unacceptable quality based on data obtained from one or more of the multiple sensor types, compared with reference data from one or more of the sensor types.

[0259] The method may specify that determining whether a weld has acceptable or unacceptable quality using combined data is a second determination separate from the first determination. In determining whether a weld has acceptable or unacceptable quality, the result of the second determination may take precedence over the first determination.

[0260] Determining based on combined data may include neural network-based steps. Determining based on combined data may include decision engines that support machine learning. Determining based on combined data may include the use of models.

[0261] Combined data can be used in one or more separate determinations, which are for acceptable and / or unacceptable welding, and in each determination, which uses data from only one sensor type.

[0262] When determinations made using data from only one sensor type in each determination are not possible and / or cannot be made with an acceptable level of associated error and / or different determinations are made for acceptable and / or unacceptable welds, determinations based on combined data can provide a determination about acceptable and / or unacceptable welds.

[0263] Determination based on combined data can be used for data points and / or data series that are close to a threshold, a threshold distance, and / or a distance relating to an indication of acceptable and / or unacceptable welding. Determination based on combined data may not be used when data points and / or a series of data points are far from a threshold, threshold distance, or distance.

[0264] The combined data may come from one or more sensor types related to the welding conditions applied during welding. The combined data may also come from one or more sensor types related to welding observations taken at a location within less than one second of welding (e.g., less than 1 / 10 of a second of welding). The combined data may exclude data from one or more sensor types related to the weld's NDT and / or welding observations taken at a location more than one minute after welding.

[0265] A neural network can be trained by feeding it labeled data. Labeled data can be data points and / or series of data points, along with indications of whether they correspond to acceptable and / or unacceptable welds. Labeled data can also be data points and / or series of data points, along with indications of whether specific characteristics that indicate acceptable or unacceptable welds correspond to acceptable and / or unacceptable welds.

[0266] Neural networks can be trained by providing supervised learning. The neural network can provide supervised learning based on operator input (e.g., labels, such as data points and / or sequences of data points, indicating whether the weld is acceptable and / or unacceptable).

[0267] Training using labeled data and / or supervised learning can be provided during the calibration phase and / or during production welding.

[0268] Neural networks can be trained from existing data, especially libraries of labeled data. These libraries can be fed into neural network classifiers.

[0269] Labeled data and / or supervised learning and / or libraries can provide data from multiple sensor types.

[0270] The second approach (library) can be used as the starting point for the dataset. The first approach (calibration or test runs performed by operator invocation on an actual welding system) can be used as an alternative from the outset. The first approach can be used to add datasets to the second approach.

[0271] Neural networks can be modified over time and / or through learning to refine the definitions used in determining acceptable and / or unacceptable welds.

[0272] Neural networks can be trained using unsupervised learning. Unsupervised learning can be performed during the calibration phase.

[0273] Neural networks can provide clustering or grouping-based processing. They can find patterns in data, such as similarities and / or anomalies.

[0274] Neural networks can be trained using only the first method, or they can be trained using the first method and then the second method, or they can be trained using the first method and the second method in parallel.

[0275] The eleventh aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention).

[0276] According to a twelfth aspect of the present invention, an apparatus for monitoring welding is provided, the apparatus comprising:

[0277] a) Multiple inputs for data from multiple sensor types;

[0278] b) One or more processors, wherein the processor is from one or more processors:

[0279] a. Receive input from at least one of multiple sensor types;

[0280] b. Compare data obtained from multiple sensor types with reference data for one or more of the sensor types;

[0281] c. Determine whether the weld quality is acceptable or unacceptable for output determination;

[0282] The device also includes:

[0283] c) A processor from one or more processors, wherein:

[0284] a. Combined data is generated from data from two or more of multiple sensor types to provide combined data;

[0285] b. Use combined data to determine a second determination, which is whether the weld has acceptable or unacceptable quality;

[0286] d) Used to output one or more determined outputs and a second determined output.

[0287] The twelfth aspect of the invention may include any features, possibilities, or options set forth elsewhere in the literature (including in other aspects of the invention). Attached Figure Description

[0288] Various embodiments of the invention will now be described by way of example only and with reference to the accompanying drawings, in which:

[0289] Figure 1a is a cross-sectional side view of a probe that can be used to implement an embodiment of the present invention;

[0290] Figure 1b is the same view as Figure 1a, but illustrates additional features;

[0291] Figure 2a is a schematic diagram of the effect of temperature difference on the ultrasonic beam in the substrate;

[0292] Figure 2b illustrates the propagation of shear wave ultrasound through a defective substrate with two different uniform temperature distributions and one non-uniform temperature distribution.

[0293] Figure 3a is a graph showing a typical non-uniform temperature distribution along the weld cross-section, as derived from the thermal model and verified by thermocouple measurements.

[0294] Figure 3b is a graph showing the planned focal point and actual geometric location of the ultrasonic beam as it propagates through the heat dissipation boundary, illustrating the planned focal point and actual geometric location of the substrate with a non-uniform temperature distribution.

[0295] Figure 4 It is a schematic diagram illustrating the scan conversion process and how pixel values ​​are derived from the weighted contribution of the nearest bundle from the nearest surrounding time samples;

[0296] Figure 5 It is a schematic diagram of the process flow for the modeling stage, ray tracing stage, and imaging stage;

[0297] Figure 6 The image shows a series of cross-sections through the weld bevel and the location and sequence of weld beads to be added, as well as a representative ultrasonically derived image of the weld bevel in this state;

[0298] Figure 7 This is an illustration of a method for verifying substrate thickness by reconfiguring the beam generated by the same ultrasonic device used for weld imaging;

[0299] Figure 8 The diagram shows ultrasonic echoes observed in the signal trajectory at different times and interference pulse trains in the same signal trajectory.

[0300] Figures 9a, 9b and 9c are illustrations of signal trajectories from three consecutive data acquisitions and indications of interference pulse trains in these trajectories;

[0301] Figure 9d shows the processed signal trajectory based on the average value of the signal trajectories in Figures 9a, 9b, and 9c.

[0302] Figure 10a is a graph of four individual but superimposed signal trajectories. Three of them are rectified signal trajectories from Figures 9a, 9b and 9c. All of these signal trajectories have interference pulse trains, but they occur at different time points. The fourth is the rectified signal obtained after processing to remove interference while retaining the true echo.

[0303] Figure 10b shows the process rectified signal trajectory and four trajectories, which illustrate the contribution selected sample-by-sample from each of the three original rectified trajectories, and the fourth is their average.

[0304] Figure 11 It is a process unrectified signal trajectory from three other trajectories, and also shows the selected contribution of each of these three signal trajectories to the unrectified trajectory;

[0305] Figure 12 This is a diagram of an unrectified waveform according to one embodiment;

[0306] Figure 13This is a schematic flowchart related to the selection process for choosing the interference processing technology to be used;

[0307] Figure 14 It is a diagram of interference on different channels, illustrated by range display, image display, and amplitude + time of flight display;

[0308] Figure 15 The processing pair is shown Figure 14 The impact of multiple channels shown in the image;

[0309] Figure 16 Examples of methods for identifying interference in the trajectory, identifying sub-sections not used in the processed trajectory, and obtaining the processed trajectory are illustrated.

[0310] Figure 17 This is a schematic diagram illustrating the adaptive control capability of the welding process according to the present invention;

[0311] Figure 18a shows an image of the defect detected by the ultrasonic probe;

[0312] Figure 18b is an image of the ultrasonic probe used in the detection process shown in Figure 18a;

[0313] Figure 19 It is a curve of outlier scores obtained from acoustic signal v data points under various welding conditions;

[0314] Figure 20a is a perspective view of the contour sensing device relative to the substrate and the weld.

[0315] Figure 20b is a schematic diagram of the sequence and geometry of a monolithic weld formed by a series of weld passes;

[0316] Figure 21 It is a series of camera images of the welding position during welding;

[0317] Figure 22 These are the curves of arc voltage and Gaussian amplitude x Gaussian center v time during the welding process;

[0318] Figure 23 It is a diagram of the combined data types displayed to the user;

[0319] Figure 24 This is a diagram illustrating the second-level processing applied to data from multiple sensor types. Detailed Implementation

[0320] Ultrasonic testing is used for non-destructive testing of various objects. An emitting transducer emits ultrasonic waves that enter the object, interact with the object and its sub-features, and then return to a receiving transducer. Because the object's temperature affects the speed of sound within it, using an incorrect speed can negatively impact the quality of the test and imaging.

[0321] When testing welds, it is desirable to perform ultrasonic testing shortly after the weld or individual weld beads are formed, in order to minimize the time spent performing the test and subsequently making any necessary corrections or optimizations to the weld or weld beads. However, this means performing ultrasonic testing while the object or substrate is at an elevated temperature and while temperature variations exist between and / or within parts of the object or substrate.

[0322] To accurately detect defects and, more generally, image weld locations, this invention provides a method for establishing a temperature distribution that takes into account the effects of temperature distribution on ultrasound and corrects for location. Therefore, accurate imaging is provided over a wide temperature range, taking into account different temperatures at different locations, particularly in high-temperature environments. This method can be used to allow for early inspection of weld locations without requiring cooling and to reduce the influence of temperature on ultrasound.

[0323] This invention relates in part to generating accurate ultrasonic-based images of welds in high-temperature environments. At the end of this document, full details of a detection device suitable for such environments are provided with reference to Figures 1a and 1b. However, it should be noted that the probe need not be a roller probe, and a wedge-shaped probe placed on a substrate and moved from one location to another can still be used, while still providing the benefits of this invention.

[0324] The basic structure of the detection device includes a rotation axis RR extending through the probe 13, and a second axial element 70 disposed on the rotation axis RR, which is connected to the axial element 64 by a series of releasable fasteners 72. The second axial element 70 provides mounting for the transducer 52, the anti-echo block 74, and the ultrasonic delivery block 76.

[0325] The coupling element 28 between the probe 13 and the substrate [not shown] is a single piece of flexible material, which will be discussed further below. The coupling element 28 has a generally straight cylindrical body portion 54 and inwardly turned edges 56a, 56b at its ends.

[0326] As the probe 13 rolls across the surface of the object, the transducer 52, the anti-echo block 74, and the transport block 76, along with the second axial element 70, the axial element 64, and the first mounting position 20, do not rotate. Therefore, the transducer 52 and associated components maintain the same sensing orientation relative to the object.

[0327] The internal volume 42 of the component to be cooled may be provided with a coolant that can be actively introduced into and removed from the probe to maintain the probe within an acceptable temperature range, reaching a level that may be higher than the maximum vertical range 50 of the transducer 52.

[0328] A mounting element 82 is also mounted on the second axial element 70, which carries a thermistor 84 for sensing the temperature of the coolant at a location 86 near the abutment welding position of the inner surface 78.

[0329] Regarding the transmission of ultrasound, coupling element 28 is made of higher temperature-compatible silicone rubber. The selected material can withstand temperatures exceeding 350°C for extended periods. This material has an attenuation of 0.87 dB / mm at 5 MHz and an acoustic impedance of 1.12 MRayls, thus aligning well with the other materials used.

[0330] Regarding the thickness of the coupling element 28, a balance is struck between increasing the thickness to provide more thermal insulation to the probe contents and increasing the thickness to the detrimental increase in attenuation. Under the operating conditions considered, a thickness between 2 mm and 10 mm (e.g., between 4 mm and 8 mm) is suitable for this material.

[0331] The selected material for coupling element 28 also provides sufficient compliance to conform to the surface of the object under moderate applied force levels. High force levels are undesirable for devices that generate high force levels while still moving the device on the specimen. Given that object surfaces encountered in real-world situations are not highly polished or smooth, compliant materials are needed to achieve good contact for ultrasonic transmission without undue loss.

[0332] Regarding the transmission of ultrasound, the anti-echo block 74 plays a crucial role in preventing ultrasound waves from bouncing back within the probe and causing noise or other negative effects on the probe. Hydrogenated nitrile butadiene rubber (HNBR) was found to be a suitable material, particularly in its N-filler form. This is due to the attenuation of 6.4 dB / mm provided at 5 MHz.

[0333] All these features are used to help the probe successfully couple acoustically to the object through the real-world surface it encounters.

[0334] Regarding coolants, air offers poor heat capacity and conductivity for active cooling. Water is also suboptimal because its acoustic impedance of 1.5 MRayls is poorly matched to other components, it is not well-suited to the temperatures encountered, and it tends to slip interface rather than act as a lubricant. Water-soluble oil-based coolants are available, for example, with an acoustic impedance of 1.1 MRayls, thus better matching the acoustic impedance of the coupling elements [1.1 MRayls or approximately 1.1 MRayls].

[0335] Transducer 52 provides a 5 MHz 64-element phased array and is mounted to generate 55° ultrasonic waves entering the object. A spacing of 0.5 mm and a height of 10 mm can be used. An angled beam facilitates complete inspection of welds from laterally spaced locations. Typically, this laterally spaced location makes good contact between the probe and the object easier than where the weld occurs. For example, in multi-pass welds, significant depressions will exist until the weld is complete, which will hinder good contact and ultrasonic propagation into the object. The tilted inspection angle also better matches typical defect orientations. This is problematic for methods based on 0° or low angles.

[0336] This type of transducer and delivery block configuration can be used to provide a fan-shaped scanning beam defined by an upper beam (angled from the transducer surface) and a lower beam (approaching the transducer surface). Regarding the performance sought in high-temperature applications, this invention provides a probe capable of long-term inspection of objects at approximately 300°C.

[0337] The coupling is dry, yet it still achieves the necessary level of ultrasonic propagation through the interface into and out of the object.

[0338] The high-temperature polymer used in the coupling component is able to withstand prolonged contact with the object at such temperatures and still successfully propagate ultrasound to and from the interface.

[0339] The coolant, and therefore the gaps that fill the coolant, can also effectively propagate ultrasonic waves to and from the conveyor block.

[0340] When the transport block is only exposed to near ambient temperature, it provides the best propagation characteristics for the transport block, so there is no need to select high-temperature resistant materials with low ultrasonic propagation characteristics.

[0341] The velocity of sound in all components along the acoustic path varies with temperature; these components are the ultrasonic delivery block 76 [the wedge between transducer 52 and compliant material 28], compliant material 28 [the tire], and the welded substrate. This, in turn, means that the angle of refraction varies with the material temperature at each interface between them. Furthermore, if a thermal gradient exists within each material, the beam within that material will bend. For example, referring to Figure 2a, the beam wavefront 100 travels more slowly in the hotter portion 102 of the substrate 104 than in the cooler portion 106 of the substrate 104. Therefore, since the beam portion in the cooler portion 106 moves ahead of the beam portion in the hotter portion 102, the beam 100 will deform from a linear beam 100' at the entrance of the substrate 104. Thus, the non-uniform temperature distribution causes the beam 100 to bend, for example, the beam 100... def .

[0342] This effect can also be observed in Figure 2b, where: (a) shows the propagation of a shear wave through the plate as a substrate to the defect location, where the plate is at a uniform temperature of 25°C; (b) shows the propagation of different shear waves through the plate as a substrate to the defect location, where the plate is at a uniform temperature of 150°C and the incident angle is adjusted to keep the beam pointing to the same target; and (c) shows the propagation of a shear wave through the plate as a substrate to the defect location with a non-uniform temperature condition.

[0343] To address potential beam bending, it is essential to establish the temperature distribution as far as possible along the entire ultrasound path, particularly in sections with higher temperatures and potential variations. If the temperature distribution is known, beam bending can be predicted, and appropriate corrections can be made in the device configuration and operation (e.g., to deliver the focused beam to the desired location) and / or applied to the signal processing of the returning beam (e.g., to provide more accurate imaging).

[0344] The first phase of the determination is to perform thermal modeling on how the heat input from preheating and the additional heat input from welding itself are distributed as a function of time across the metal volume of the substrate.

[0345] In one method of validating thermal modeling, a large number of distributed thermocouples are arranged on a substrate in the form of a steel plate. Thermocouples are placed on both the upper and lower surfaces of the substrate. The substrate is then preheated to the optimal welding temperature, and various weld beads are then performed to obtain the thermocouple responses. The level of modeling can also be increased by varying the level of preheating, but the preheating level is typically within a consistent and tight temperature range between examples of the same substrate and even between different substrates [thickness, profile, steel type, etc.]. Therefore, small variations due to preheating can be expected.

[0346] After performing this type of thermal modeling, the safe operating temperature of the probe [based on knowing the probe's response over time when exposed to different temperatures] can be compared with the temperature at different locations far from the weld, and thus the modeling makes it possible to establish the closest safe distance at which the probe follows the weld head.

[0347] Important for thermal modeling with substrate depth, these studies identify a rapidly stabilizing temperature distribution occurring in the thermal profile perpendicular to the metal surface after welding. This significantly simplifies the geometry of the temperature distribution and thus simplifies compensation calculations, particularly when compensation for the effects of multiple skips of the substrate is required. A skip is when ultrasound enters the substrate through the interface, reflects off the substrate from within the interface, and returns to the entry interface.

[0348] Regarding thermal modeling of the distribution from the weld location with the highest temperature to the substrate, the temperature distribution is typically, but not always, symmetrical around the weld center, thus the bundles undergo complex S-shaped deformation as they pass through the weld region. This can be seen in the temperature distribution in Figure 3a and the ray tracing analysis in Figure 3b, where location is expressed in mm to either side of the weld location, and the weld location has the highest temperature [1000°C in this case]. In this example, an amplified thermal gradient is used to emphasize the magnitude of the bundle bending.

[0349] Referring to Figure 3b, the deviation caused by temperature distribution in the case of weld inspection is illustrated in more detail.

[0350] A series of plans focus on 110 1 Up to 110 i Extending along plane 112 with a near-plane end, a planned focus 110 is provided at the near-plane end. 1 The near-plane end faces the probe 13. The planned focus 110 is provided at the far-plane end. i This further distances the probe from 13. In the case of a plate, plane 112 is parallel to the upper surface of the substrate, thus representing a constant depth of penetration into the substrate in this case.

[0351] Regarding the focus of the plan 110 1 Up to 110 iThe transducer 52 and the resulting beam are angled relative to the upper and lower interfaces of the substrate. In this respect, an angle of 55° can be used, thus defining the centerline 114 of the scanning arc. The control system of the transducer 52 allows the beam to be diverted to different angles, thus providing a fan-shaped scan. All the elements of the array [64 in this example] are used for each emission of the beam, but each element has a different excitation time to divert and focus the resulting beam to the desired location. In effect, this defines the boundaries 116 and 118 of the scanning arc, which define the minimum and maximum angles at these boundaries, and thus define the two planned focal points 110 at the ends. 1 and 110 i .

[0352] As established in thermal modeling, the temperature profile 120 is typically perpendicular to the surface of the substrate. However, this method can work in conjunction with other profiles or boundary orientations and / or profiles, thus enabling the full and wide range of different temperature gradients in different directions within the substrate to be considered.

[0353] The more thermal boundaries used to delineate the substrate, the more accurate the consideration of temperature distribution can become, and the more accurate the actual geometric location point 122. 1 Up to 122 i The determination can become more accurate. Of course, temperature changes are continuous, but the same effect can be accurately represented using a large number of step changes.

[0354] Each time the beam crosses a thermal boundary, a correction is applied to its path based on the temperature change represented by the boundary. This gives the beam a new path, thus providing a new path length portion before crossing the next thermal boundary. Again, further corrections provide yet another new path, and so on. In Figure 3b, this can be done based on the actual geometric location point 122. 1 Up to 122 i The results show that these actual geometric location points were determined to be the actual geometric locations from which the signal samples originated. Actual geometric location points 122 1 Up to 122 i The profile is caused by the beam bending to a gradually shallower angle and the wavefront decelerating as it encounters higher temperatures toward the weld center plane 124. The angle becomes steeper again, the wavefront accelerates again, and bends back in the opposite direction after passing through the weld center plane 124.

[0355] Given the actual geometric location point 122 of the signal source. 1 Up to 122 i With Plan Focus 110 1 Up to 110 i This difference means the scan conversion algorithm must plot the actual geometric location point, such as 110, in various cases. 1Instead of focusing on the corresponding plan focus, such as 122 1 The signal observed at that location.

[0356] A similar process is performed during reception, where signals from each channel experience different time delays, causing the signals from all channels to be coherently added together to increase the net response from the specified direction and range. A potential drawback here is that the steering angle and focus are fixed at acquisition time, making it impossible to correct for any errors in the thermal distribution later.

[0357] • Figure 4 This provides a further explanation of how the position of the pixel is considered between the first beam i and the next adjacent beam i+1. In this illustration: A ij It is the amplitude of sample j in bundle i;

[0358] •W ij It is a weighting factor for the contribution of sample j from bundle i;

[0359] • α is a fraction of the distance from the hypothetical bundle i to bundle i+1 in which the pixel is located;

[0360] •β is the fraction of the distance from sample j to sample j+1 along the hypothetical bundle where the pixel is located through the sample.

[0361] therefore:

[0362] A pixel = (1-α).(1-β).A ij + (1-α).( β).A ij +1 + (α).(1-β).A i+1j + (α).( β ).A i+1j+1

[0363] A pixel = W ij A ij + W ij+1 A ij+1 + W i+1j A i+1j + W i+1j+1 A i+1j+1

[0364] The configuration remains unchanged between each acquisition; therefore, to accelerate image generation, i+α, j+β, and W are pre-calculated for each pixel in the display area. ij W ij+1 W i+1j and W i+1j+1The values ​​of i+α and j+β are stored in a lookup table. During the imaging operation, the addresses of the bundle and sample data contributing to each pixel (i, i+1, j, and j+1) can be derived by simply truncating the i+α and j+β values ​​in the lookup table. When the bundle is straight, generating a cover for, for example, the bundle markers for the selected A-type scan data is a straightforward operation, but it is more complex for curved bundles (where thermal gradients exist). For this, a reverse lookup process from pixel to bundle and sample is required, which means that it is beneficial to maintain the i+α and j+β values ​​in the lookup table.

[0365] Figure 5 The complete workflow, including the modeling phase, ray tracing phase, and imaging phase, is shown.

[0366] During the modeling phase, a thermal model is generated, which informs the temperature of each data point. The initial thermal model can be based on information from previous actual welding operations and / or previously stored information from previous thermal models.

[0367] Next, real thermal data of the welding environment is collected. This can be obtained through inputs [e.g., provided preheating temperature and / or welding conditions], and / or from temperature sensors in the welding environment [e.g., thermistors associated with the probe, other temperature sensors, infrared sensors, etc.], which can inform the temperature and location used for this data. The data is combined to form the actual thermal dataset.

[0368] Then, the actual thermal dataset and the thermal model are compared to assess how well they correspond to the temperature at a given location. If the difference in correspondence is too large, the thermal model is updated. A new comparison is performed, and this cycle is repeated until the actual thermal dataset and the thermal model are considered to match. The ray tracing phase can then begin.

[0369] In the ray tracing stage, the thermal model received in the modeling stage is converted into a sound velocity map. The substrate is known, and its temperature-dependent sound velocity can be established through calibration tests. This calibration data is stored and accessible for generating the sound velocity map. The temperature at the data points is converted into velocity. The velocity profile intervals can then be set, and the sound velocity map is subjected to the application of profiles through data points with the same velocity. In this way, the substrate is divided into different velocity regions.

[0370] The starting point and initial path of each ray [beam component] are known. Each time the projection of this path crosses a velocity boundary, a refractive representation is applied. The applied refractive representation varies in degree according to the velocity difference represented by this profile compared to the previous profile, and in direction according to whether the velocity profile increases or decreases compared to the previous profile.

[0371] The speed and path length between contours give the travel time through that speed region.

[0372] This process is repeated for each profile that intersects within the probe during its entry into the substrate. The process can also be repeated for echoes on each profile that intersects during its exit from the substrate. After completing the ray tracing step for each ray, the process can proceed to the imaging stage.

[0373] In the imaging stage, the initial step is to generate the region of interest from which the desired result is to be displayed. This can be affected by factors such as the shape of the substrate, the shape or construction of the weld, and the manner in which the weld is provided.

[0374] The pixel values ​​are generated essentially through linear interpolation between the four closest real samples. To avoid time penalties when computing linear interpolation at runtime, the addresses of the four samples and their respective weights are pre-calculated and stored in a lookup table. Then, each newly acquired dataset is scanned and transformed into a linear display image by processing each pixel sequentially, accessing the four samples and multiplying each sample by the relevant weighting function before summing to form the final pixel value.

[0375] Pre-existing information about the geometry exists regarding the weld groove location, shape, and orientation, which can be checked against the geometry observed in the results. If these are relevant, the process can proceed. If they are not relevant, the compensation has not been applied correctly, and the temperature data supplied to the compensation needs to be checked.

[0376] If the detected temperature does not match the thermal model, it needs to be corrected and the sound velocity map readjusted for the correction. The corrected sound velocity map can then be used in the above process steps.

[0377] exist Figure 6 The image shows some example signal results, in this case, of a weld bevel gradually filled through multiple weld beads. In the first sub-image, the bevel has no weld, and the edge of the bevel is detected in the ultrasonic image. As the weld beads gradually fill the bevel from the bottom in a numbered sequence, the extent of the exposed bevel edge decreases, so the range of the echo in each image is smaller than the range of the echo in the previous image, and eventually disappears.

[0378] If a defect is found, it will appear in the image facing to the right, compared to the rest of the weld groove wall.

[0379] The above implementation is based on using thermal profiles to derive the travel time and thus correct for the temperature distribution. Beam reflection is fully considered, and the beam path is determined.

[0380] In alternative implementations, for the same purpose, Full Matrix Capture (FMC) can be combined with the Full Focus Method (TFM), but for larger and more complete datasets.

[0381] Full-focusing method (FMC) is a technique similar to holography in which raw data signals containing amplitude and phase are collected from each combination of transmit (Tx) and receive (Rx) elements. An image is reconstructed from this FMC dataset by creating a virtual path from each element to each pixel and performing differential delay as a post-processing operation. This means that each pixel can be ideally focused on both the transmit and receive sides, and for this reason, the method is called the Total Focusing Method (TFM). The same technique described above (where the gradient is approximated by multiple individual thermal regions) is used to determine the travel time for thermal gradient compensation from individual elements to individual pixels required by the TFM reconstruction algorithm.

[0382] Verification of substrate thickness

[0383] As mentioned above, temperature distribution within the substrate is a significant issue that needs to be addressed. The applicant also discovered that even when the manufacturer specifies a given thickness, such as the thickness of the substrate (e.g., a plate), material variations exist. This also needs to be considered if imaging optimization is desired.

[0384] The bundle path length is affected by thickness, and the observed variation, compared to the thickness value associated with typical material thickness, gives a variation of 0 to 5 mm in path length. It should be remembered that the path length is multiplied by the number of half-jumps performed. To accurately determine the defect location, the actual distances of the path length to and from the defect location need to be established, rather than an assumed path length based on the stated substrate value.

[0385] As a pre-weld inspection step, a probe can be moved along the same path on the substrate during the weld inspection. The ultrasonic beam can be used to determine the thickness of the substrate. This method... Figure 7 Example in.

[0386] exist Figure 7 In this diagram, only a portion of the aforementioned roller probe 13 is shown with respect to coupling element 28 and conveyor block 76. Transducer 52 is not shown, but it will be positioned near the angled surface of conveyor block 76. Unlike the angled shear wave fan-shaped scan of welds used for the aforementioned inspection purposes, the NDT beam 800 and thickness measurement beam 802 operate at 0° (i.e., perpendicular to the metal surface) to the substrate 804 or object. Control electronics are used to steer the beams away from the NDT beam 800 position to determine the position of the thickness measurement beam 802. To ensure that the 0° beam is included, a fan-shaped scan centered at 0° and including multiple longitudinal beams is used.

[0387] Although the 0° beam has been used to check the amplitude of the back wall echo from the substrate 804 as a way to monitor the coupling quality between the probe 13 and the substrate 804, in this case, the beam is also used to measure the thickness rather than just to ensure the presence of a given echo intensity.

[0388] The strongest echo received at the detector in probe 13 corresponds to the first interface 806 where probe 13 abuts against substrate 804. The second strongest echo corresponds to the interface 808 on the opposite side (internal reflection) of the substrate. This method uses the echoes and their repetitions to measure local thickness in order to compensate for the localization of the region of interest.

[0389] By using a narrow fan-shaped beam instead of a single beam as the thickness measurement beam 802, this method is able to determine which angle produces the strongest repetition, and this can provide information about any tapered compression of the coupling element 28 (in this example, a tire). Tapering can cause beam refraction, and is therefore another issue that may need to be corrected when imaging the weld with the correct geometry.

[0390] As described above, the temperature-corrected detection location can be used to establish the actual distance along the path length. Temperature correction is still required in the transport block 76 to achieve a 0° measurement direction; that is, perpendicular to the substrate and the minimum distance or thickness through the substrate. This can be noted so that it can be used at the same location along the path on the substrate during weld inspection to correct the imaging and thus the location of any defects found.

[0391] Alternatively, when the probe advances along the weld path during actual weld inspection to determine the thickness, the probe can quickly switch between weld inspection mode and thickness determination mode and repeat the cycle.

[0392] As a result of either approach, the actual thickness is known at any location along the weld path compared to the theoretical value, and this makes the thickness usable to correct the image relative to the location of geometric features and / or defects.

[0393] RF interference handling

[0394] Significant interference bursts from robot motors used for welding arms, probe arms, and other tasks were observed in the signal trajectory used for ultrasonic testing. These bursts were found to be several times larger than the ultrasonic echoes (even those from planar interfaces). Other bursts were less strong and therefore more difficult to distinguish. Planar interface echoes are typically much stronger than defect echoes. Other sources of interference exist and can be addressed using the techniques presented in this paper. For example, electrical interference is generated by a variety of motors, such as those used in pump cooling systems encountered in welding applications and their measurements. Other sources may include drive motors used for lifts, power supplies, etc. Therefore, solutions to address this interference are desirable for achieving effective imaging.

[0395] The broadband nature of pulse trains means that bandpass filtering techniques have limited effectiveness in reducing the amplitude of interference.

[0396] The smaller the returned signal, the bigger the problem. In high-temperature probe environments, the distance from the transducer to the compliant material, the dry interface to the substrate, and the same factors on the return all contribute to signal attenuation and make noise more problematic.

[0397] In addition, the timing of the interference pulse train is irregular in both occurrence and duration, depending on the motor tasks required of each arm.

[0398] Figure 8 It is a diagram of an ultrasonic echo with a defect pointing towards the midpoint of the signal trajectory and an interference pulse train pointing towards the rear part of the trajectory.

[0399] However, the intermittent nature of the disturbance can be helpful in finding a solution.

[0400] In a first embodiment designed to address interference bursts, a transducer is used to emit a series of beams through a substrate and to detect the return signal as previously described. Three such detection signal trajectories are shown in the examples illustrated in Figures 9a, 9b, and 9c.

[0401] Interference from the robot motors is essentially a high-frequency pulse train separated by gaps significantly longer than the repetition period of an ultrasonic pulse. This time difference means that if a pulse train from the robot motor appears on one beam, it will not appear on subsequent beams, thus allowing interference-free data from this second signal to replace contaminated samples in the first beam. If a second, independent robot is present, it will act as a second source of interference, with its pulse train asynchronous with that from the first robot. If the interference from this second robot happens to occur in the desired interference-free sample region of the second signal, then the sample from the third signal will be clean and can be used to replace the contaminated sample.

[0402] In all cases, since the defect is applied to the same location in all three tracks, the defect simultaneously causes echoes in the tracks. The start time of the signal tracks does not affect the echoes because the three tracks are generated by the same beam emitted along virtually the same path into and out of the substrate [because the probe does not move a meaningful amount along that path within that time frame] and encounter the same substrate.

[0403] Interference bursts are present in the latter part of the first signal track in Figure 9a. Potentially, similar interference bursts are also present in the earlier part of the second signal track in Figure 9b. Additional interference bursts are present in Figure 9b, but not in 9a. Two more potential interference bursts are present in Figure 9c.

[0404] Figure 9d shows the signal track, which is a sample-wise average of the three signal tracks in Figures 9a, 9b, and 9c, used to improve the signal-to-noise ratio (SNR). As a result, the echo portion of the signal track remains significant, but these interference bursts are reduced because only one signal track contributes to the average of the interference bursts at any point in the signal track. This averaging method can, in a basic way and in some cases, indicate which are echoes and which are interference bursts. The technique does not attempt to identify the presence or timing of any interference bursts; any such burst contributes to the average signal track in Figure 9d. However, the amplitude of any burst is significantly reduced due to averaging multiple bundles; the more bundles averaged, the greater the reduction.

[0405] However, in some cases, when the interference pulse train is similar in amplitude to or greater than the echo and / or the signal is noisier or more complex, simply reducing the amplitude may not be sufficient. If the pulse train gives the impression of an echo in the results, this will lead to false negatives on the weld and unnecessary further consideration and / or remedial measures.

[0406] An improvement over the above technique can be achieved by correctly identifying the interference bursts and ensuring that any echoes appearing in the same part of the signal trajectory are detected in some way.

[0407] In Figure 10a, a more sophisticated method is used to remove interference from the signal traces. Figure 10a shows three separate rectified signal traces 0, 1, and 2, which represent the same bundle on three separate transmissions, plotted against each other.

[0408] The mechanism used to determine whether a sample is likely from a interference burst is done by comparing the amplitudes of rectified (envelope detection) samples from each of signals 0, 1, and 2. The signal with the largest amplitude sample is the candidate due to interference, and therefore the minimum value across the signal is selected for use in the final processed waveform. If no signal sample is from the interference burst, the amplitude difference across the signal will be very small due to sensor or instrument random noise. Therefore, comparing this amplitude range to a threshold is a mechanism to indicate the absence of interference and thus use the average of samples across all signals to improve SNR. The threshold can be chosen as a value related to the expected noise amplitude or as a defined proportion of the signal amplitude.

[0409] Signal trajectory 0 includes a high-amplitude portion towards the end of the signal trajectory that is not present in 1 or 2, and this is therefore considered to be a burst of pulses from RF in flow 0. Therefore, when forming the portion of flow 0 that contributes to the average signal trajectory, this portion is omitted from the contribution. This can be seen in the RF0 flow at the bottom of Figure 10b, where this portion of trajectory 300 is omitted. Other portions 302 are also omitted from the RF0 contribution, where those spikes are not visible in other flows.

[0410] Averaging the contributions from those trajectories that are allowed to contribute to the average signal trajectory means that averaging the uncontaminated trajectories improves the signal-to-noise ratio.

[0411] Similarly, returning to Figure 10a, 1 includes two high-amplitude portions towards the beginning of the signal trajectory that are not present in 0 or 2, and these are therefore considered to be the pulse trains from RF in stream 1. Therefore, when forming the portion of stream 1 that contributes to the average signal trajectory, that portion is omitted from that contribution. This can be seen in the RF1 stream at the bottom of Figure 10b, where these portions of trajectory 304 are omitted.

[0412] A similar method is used to omit the two sets of two signal trajectory spikes in the latter half of the trajectory and the partial omission of the contribution from flow 2 in part 306.

[0413] Excluding these portions, the portion that contributes to the averaging trajectory is shown at the very bottom of Figure 10b, and this gives the plotted signal trajectory in Figure 10b, which shows the expected echo spike 310 clearly observed in Figure 10a, as well as two other echo spikes 312 in Figure 10a. The remaining signal portion of the lowest trajectory in Figure 10b is informative and provides a substantial amount of time during which the defect signal will be detected.

[0414] In this stage of processing, there are portions of the averaging trajectory, and these portions represent samples that have been determined to have no interference with any of 0, 1, and 2 and can therefore be replaced by the average of 0, 1, and 2; see the lowest trajectory among the four trajectories in Figure 10b. This averaging provides the benefit of reduced signal-to-noise ratio. Where these portions are absent, one or more of 0, 1, or 2 have been identified as originating from interfering bursts, and therefore only the average of the clean signal is used. This provides some improvement in signal-to-noise ratio through averaging, but the amount of signal contributing to the average is reduced by excluding certain signals.

[0415] Using this minimum or threshold method across all potential tracks is a fast and highly efficient approach that combines exclusion of potential interference with partial track replacement. However, this technique fails to identify bursts, instead assuming that high-amplitude portions of the matched values, substantially greater than those from other tracks that do not imply interference, originate from bursts. A more explicit technique for identifying bursts would be useful, as if the interference were identified only on one bundle; averaging across the remaining bundles would provide better signal fidelity and improved SNR compared to the minimum method across all streams.

[0416] In another method, interference detection and trajectory partial substitution are handled in separate steps. The following section combines... Figure 16 Describe this alternative method, and this is a sample-by-sample approach.

[0417] By providing a general method that allows for the identification of burst regions on each track, the uncontaminated portions of all tracks can be appropriately identified. The criteria used to determine the identification of bursts and the alternative methods for replacing portions containing bursts can each be controlled individually, or their criteria can be changed as needed.

[0418] In this general approach, the first step is to identify suspicious interference components on the flow or trajectory of each beam. This is achieved through the following steps:

[0419] 1.1. If the input data trajectory / stream is a full RF waveform, envelope detection is performed on each trajectory / stream waveform to generate the rectified trajectory / stream of each bundle.

[0420] 1.2. For each bundle, find the minimum value of each rectified sample across all trajectories / flows, as this will be a good approximation of the undisturbed value of the sample at that time.

[0421] 1.3. Derive the difference between the sample of each trajectory / flow and this minimum value, and mark it as interference if it exceeds a threshold. Note that this threshold can be constant or adaptive, such as the minimum value multiplied by a defined factor.

[0422] In the second step, the method provides interference removal through the following steps:

[0423] 2.1. Use interference tags (derived in 1.3) to reject any input trajectory / flow samples for the bundle.

[0424] 2.2. Average value over the remaining samples across the trajectory / flow.

[0425] 2.3. Use this average value as a time sample of the processed bundle.

[0426] refer to Figure 16 In the rectified data, the three tracks at the top illustrate... Figure 16 The top, middle, and bottom waveforms in the upper part represent the cases of single-beam flows 0, 1, and 2. The highlighted sections on these interference detection diagrams—one A towards the middle of flow 0; one B towards the end of flow 1; and two C and D towards the beginning of flow 2—represent all areas where interference exists. Peak e is the echo.

[0427] The algorithm described above is used to detect this interference, and it is marked in the interference processing diagram. Figure 16 The bottom three trajectories, 0", 1", and 2", are excluded from further consideration to arrive at the processed trajectory. Individual sub-sections within the interference pulse train segments A, B, C, and D [represented by individual bar peaks] are highlighted for discarding. Then, from... Figure 16 These identified, highlighted sub-sections are excluded during the formation of the intermediate processing trajectory (trajectory P). This processing trajectory P shows a processed sample of the bundle, illustrating the removal of interfering pulse trains and the reduction of noise in the remainder of the waveform.

[0428] Figure 11 and Figure 12 A further modification to the method so far is shown, which can be applied to unrectified signals. This is also a sample-by-sample selection, so the general approach of the previous embodiments can also be used here, accompanied by flexibility in the criteria used and an improvement in the signal-to-noise ratio.

[0429] Here, the method of identifying samples to be removed from the interference burst by selecting among signals with the smallest amplitude will not work. Instead, the signal is first rectified, and then the original method is applied to these modified waveforms to determine whether the samples in the final waveform come from signals 0, 1, 2, or the average value. This selector mechanism is then applied to the original unrectified signal, producing an unrectified processed signal.

[0430] exist Figure 11In the presence of flows 0, 1, and 2, a situation similar to that in Figure 10a exists. Flow 0 has an RF burst in the later part of its signal trajectory, flow 1 has a burst in the early part of its signal trajectory, and flow 2 has two bursts in the latter half of its signal trajectory. If these contributions are excluded to give... Figure 11 The average trajectory at the bottom shows the same type of missing parts.

[0431] However, with further processing, these missing portions can be reconstructed using portions obtained from one or both of the other signal trajectories. Therefore, for portion 330 excluded due to RF pulse train 332 in stream 1, portion 338 in the average trajectory can be filled using portion 334 from stream 0 or portion 336 from stream 2. If both portions 334 and 336 are used for filling, their average can be used.

[0432] The result of this filling is Figure 12 As can be seen, a significant portion of the entire trajectory comes from the average of three trajectories, some from the average of two trajectories, but other portions are filled using one of the other trajectories. The missing portions from stream 0 are filled using stream 1 and / or stream 2, etc. Figure 12 It provides a continuous trajectory with no lost echo results and fully considers interference, without lost defects or echoes.

[0433] Although the robot motor interference is significant, the welding arc interference frequency is independent of the frequency of the ultrasonic signal of interest, and therefore no special processing is required.

[0434] The method used for unrectified signals is also directly applicable to full-matrix acquisition signals. Here, the combination of interference burst removal and SNR improvement through averaging in the absence of bursts is particularly beneficial due to the inherently low SNR of each received signal, as it is caused by a small excitation energy.

[0435] Minimizing the computation time for processing FMC signals is particularly important because there are many more signals to process in the FMC dataset. If all received signals transmitted by a given element are acquired in parallel, the nature of electromagnetic interference means they will all have bursts in the same sample. Therefore, it is necessary to analyze the multi-pulse set from only one receiver channel to derive a sample selector that can be used across all parallel receive channels, thereby reducing computation time.

[0436] Since the aforementioned techniques require varying degrees of computation, it is beneficial to choose only the more computationally intensive technique when necessary, and otherwise use alternative techniques. In this regard, Figure 13This indicates consideration of the "sample average" method [described above with reference to Figures 9a to 9d]; or the "minimum sample value" method [described above with reference to Figures 10a and 10b]. Figure 11 and Figure 12 [Description]; or, in the case of full matrix capture, a possible choice of the "single-channel rectification" method. Rectifying a single channel as described above is used for more general rectification.

[0437] Therefore, refer to Figure 13 The system is configured to scan from multiple transmitters and acquire multiple signal trajectories 1300. The repetition rate used is adjusted to avoid aliasing patterns 1302, and then data [signal trajectories] can be acquired 1304.

[0438] Regarding the acquired data 1304, the processing takes into account whether there is interference 1306 across all parallel receive (Rx) channels, as is the case in the full matrix capture method.

[0439] If there is no interference in all channels 1308, then the signal trajectory 1310 of each of the multiple transmit Tx is considered sequentially for each bundle. Then, rectification is performed using the method described above 1312. The minimum and maximum values ​​of the signal trajectory 1314 are obtained and then compared with a predetermined threshold.

[0440] If the value is below the threshold 1318, the method is performed by averaging the samples from each of the multiple sending Tx streams to obtain the result 1320. This is the method described above with reference to Figures 9a to 9c.

[0441] If this value is higher than the threshold 1322, a more discriminative analysis method is required, and this method constructs the rectified flow 1324 by selecting the minimum value of each sample. The same selector is used to construct the rectified trajectory 1326 from the original signal trajectory, thus obtaining the result 1320. This is in reference to Figures 10a and 10b above. Figure 11 and Figure 12 The method described.

[0442] Returning to an earlier stage of the process, where the problem of interference in all channels is considered (1306), if it is (1328), an alternative process method is used. In this case, only one channel needs to be analyzed (1330), and this is done by re-rectifying (1332). Therefore, the minimum and maximum values ​​are found (1334), and a sample selector can be used (1336) to select the appropriate Tx stream for all parallel Rx channels. The result (1320) is reached again.

[0443] exist Figure 14The example illustrates a method where interference is observed on all channels, where, prior to processing, substantial interference is present in another scan on the range display 1400, image 1402, amplitude trajectory 1404, and time-of-flight diagram 1406.

[0444] The processed location is Figure 15 As shown in the figure, there are significant improvements in each scan, namely, the processed range display 1500, the processed image 1502 and the amplitude trajectory 1504, and the time-of-flight diagram 1506 are also present in another scan.

[0445] By using multiple transmissions and selecting the minimum rectified value on a sample-by-sample basis, a good solution to any interference is provided.

[0446] The processing result is compared with the acquired data stream and signal trajectory to generate a selector waveform, which indicates which sample comes from which stream. The selector can be used with the raw RF stream to create a processed RF stream, thereby allowing FMC data acquisition and processing.

[0447] In the absence of detected interference (where the range of values ​​across the flow is small), the selected processing method has low computational strength when using only the average sample. However, when interference is detected, the processing method can choose an alternative approach to provide the minimum sample value and thus successfully handle the potential impact of the interference.

[0448] Increasing the number of transmissions considered allows the method to address and handle a larger number of asynchronous interference sources.

[0449] The range of methods provided by these embodiments is useful when interference prevention or suppression is insufficient.

[0450] Other publicly available information

[0451] The additional disclosure relates to welding based on the use of multiple sensor types and combinations of data from these sensor types, quality control in welding, and improvements to the methods of using these methods, as well as related improvements.

[0452] During welding, a significant number of variables influence weld formation and its quality. Attempts have been made to monitor these variables during welding to detect deviations and the resulting potential for compromised weld quality and / or defect formation. The weld exhibits various characteristics during and immediately after formation. Attempts have been made to determine these characteristics during welding to detect deviations and the resulting potential for compromised weld quality and / or defect formation.

[0453] One potential objective of this invention is to provide a method for welding and quality verification in which increased benefits are gained by mitigating welding defects or other quality problems through greater use of variable monitoring and / or characteristic determination. Another potential objective is to provide a method for welding and verifying welding conditions that seeks to synchronize and fuse data from multiple sensor types and / or sources to obtain more information than the sum of the parts. Additionally, potential benefits can be derived from attempting to analyze various data types in a manner that provides more information.

[0454] Referring to the seventh to twelfth aspects of the present invention described above.

[0455] The following descriptions are provided in relation to these aspects.

[0456] It is known that a nondestructive test (NDT) is performed on the weld after it has formed and cooled to check for any defects. Any defects found can then be remedied. However, in the prior art, this NDT is performed a long time after weld formation, thus providing information about defects too late to prevent or mitigate further defect formation.

[0457] During welding, there are a significant number of variables that affect the formation and quality of the weld. These variables may include: wire feed rate, voltage, current, welding speed, distance between welding equipment and welding position, shielding gas, shielding gas flow rate, substrate shape / profile / structure / size, substrate preheating temperature, weld groove shape / profile / structure / size, and weld or weld bead shape / profile / structure / size.

[0458] Attempts have been made to monitor individual variables during welding to detect deviations in these variables and the resulting potential for compromised weld quality and / or defect generation.

[0459] When a weld is formed, and immediately after its formation, the weld exhibits a variety of characteristics. These characteristics may include: acoustic emissions from the welding location and / or the arc, the topology of the formed weld, and the visual characteristics of the weld during and / or after its formation.

[0460] Attempts have been made to identify individual characteristics during welding in order to detect deviations in these characteristics and the resulting potential for poor weld quality and / or defect generation.

[0461] Overall process

[0462] This invention seeks to mitigate other quality problems arising from or contributing to welding defects by making greater use of variable monitoring and / or characteristic determination, thereby gaining increased benefits in welding. This invention also seeks to synchronize and fuse data from multiple sources to obtain more information than the sum of the parts. Furthermore, this invention attempts to analyze various data types in a manner that provides more information.

[0463] This invention also reduces the risk of defect formation by improving the quality of the initial weld. This is achieved through an increased range of welding-related data types collected, analyzed, and employed as welding progresses.

[0464] This invention utilizes data derived from multiple sensors and / or sensor types in real time, where machine learning / artificial intelligence algorithms are used to derive near real-time decisions for weld process control. Therefore, the decision benefits from the involvement of multiple different sensor types.

[0465] In the example invention, four different sensor types are used to collect data from their combined use, providing adaptive control capabilities for welding and thus reducing the chance of defect formation. However, this method is suitable for handling a greater number of sensor types or data types, and is also suitable for use with sensor types different from those illustrated.

[0466] refer to Figure 17 It provides a schematic diagram of adaptive control capabilities.

[0467] The welding process is the process of defining and controlling a series of variables that affect and control the welding process. These variables may include: wire feed rate, voltage, current, welding speed, distance between welding equipment and welding position, shielding gas, shielding gas flow rate, substrate shape / profile / structure / size, substrate preheating temperature, weld groove shape / profile / structure / size, and weld bead shape / profile / structure / size.

[0468] During the welding process, Figure 17 On the left side, a high-temperature compatible roller ultrasonic probe [described in further detail below] is used to perform the NDT step [post-weld thermal NDT], and this step provides ultrasonic data [NDT indication], which is supplied to provide an overall assessment of the weld qualification at the end of the process.

[0469] In addition to monitoring the welds produced and taking any remedial action, the welding process also leads to a series of parallel steps that begin with process monitoring, which attempt to prevent or minimize the extent of defects.

[0470] In the process monitoring step, data on the applied voltage and current of the welding equipment over time are measured and collected. In this embodiment, voltage and current sensing is used to consider issues related to proper weld formation or proper weld formation.

[0471] Data from five other sensor types is then fed into the data stream to form sensor data. These data streams are laser, visual, and acoustic data streams from appropriate sensors of this type, described in more detail below. In this embodiment, the use of laser scanning to sense the weld contour, visual evaluation of the weld, and acoustic sensing of sound originating from the weld are also used to consider issues related to proper weld formation or proper weld formation.

[0472] The processing sequence includes collecting sensor data using laser, visual, and acoustic data streams. Additionally, a process monitoring step collects data on the applied voltage and current of the welding equipment. Post-weld NDT (Normally Detonated Weld) also provides NDT indication; a fifth data type. These form the complete data stream under consideration. Now consider each of these sensor types and the data types they provide. The different sensor types are:

[0473] • Phased array ultrasonic sensor type - for measuring weld quality;

[0474] • Acoustic sensor type - Derives weld process performance from the sound of the welding arc;

[0475] • Laser sensor type - monitoring weld topology;

[0476] • Visual monitoring type - replacing the operator's vision in the visualization of the weld pool;

[0477] • Voltage and current monitoring type - measures power to maintain the welding process at its optimal state.

[0478] Ultrasonic sensor types

[0479] This type of sensor deals with the results from the resulting weld, while other sensor types focus more on the weld itself during its formation. Therefore, this type of sensor provides information about the remedial measures needed, while the later sensor types described below can be used primarily to mitigate defect formation.

[0480] Ultrasonic sensor types include transducers providing a 5 MHz 64-element phased array, mounted to generate 55° ultrasonic waves entering the substrate. A spacing of 0.5 mm and a height of 10 mm can be used. Angled beams facilitate complete inspection of welds from laterally spaced locations. Typically, this laterally spaced location makes good contact between the probe and the object easier than where the weld occurs. For example, in multi-pass welds, significant indentations will exist until the weld is complete, hindering good contact and ultrasonic propagation into the object. This is a problem with methods based on 0° or low angles.

[0481] This type of transducer and associated ultrasonic delivery block configuration can be used to provide a fan-shaped scanning beam defined by an upper beam (angled to the vertical line away from the transducer surface) and a lower beam (vertical line close to the transducer surface).

[0482] Regarding the performance sought by the probe in terms of high-temperature performance, the probe is able to inspect substrates at approximately 300°C for extended periods.

[0483] The coupling is dry, yet it still achieves the necessary level of ultrasonic propagation through the interface into and out of the object.

[0484] The high-temperature polymer used in the coupling component is able to withstand prolonged contact with the object at such temperatures and still successfully propagate ultrasound to and from the interface.

[0485] The gaps between the coolant and therefore the sensor-type components are also able to effectively propagate ultrasonic waves to and from the transport block.

[0486] When the transport block is only exposed to near ambient temperature, it provides the best propagation characteristics for the transport block, so there is no need to select high-temperature resistant materials with low ultrasonic propagation characteristics.

[0487] Sensor types also include integrated surface temperature measurement and coolant temperature measurement sensor types.

[0488] Welding position conditions and effects

[0489] Arc welding uses a power source to create a sufficient voltage difference between the electrodes of the welding equipment and the substrate to be welded to generate an electric arc. Current is generated. The arc heats the substrate to a molten state [potentially accompanied by electrode wear]. Upon cooling, the molten metal solidifies to bond the two substrates together.

[0490] The speed at which welding equipment moves relative to the substrate affects weld quality by influencing the degree of melting and the shape of the weld pool.

[0491] Welding sites are typically protected by shielding gases to prevent atmospheric oxygen, water, or water vapor from reaching the welding area. Inert or semi-inert gases are commonly used as shielding gases; examples include argon and helium. The flow rate of the shielding gas affects its effectiveness.

[0492] To achieve high weld quality, numerous operational variables need to be carefully controlled during the welding process. These variables can be considered indirectly, as exemplified in the following sections. Although not directly sensed in the example embodiments, other sensor types can measure the speed of the welding equipment movement, the shielding gas flow rate, and the shielding gas velocity, and these sensor types and their datasets are added to the processing.

[0493] Acoustic sensor types

[0494] Acoustic sensors collect high-frequency audio signals generated during welding. These signals originate from the electric arc forming the weld, the interaction between the arc and the substrate, and the interaction between the arc and the shielding gas. The detected audio signals have been identified as sensitive to several important variables in the welding process. (Reference) Figure 19 This shows examples of outlier score values ​​for different welding characteristics.

[0495] Outlier scores are obtained through a mathematical method that considers the distance of a particular data value to a known set of data values ​​that have been classified as acceptable values ​​for the property and / or sensor type data being evaluated.

[0496] One such method used in this invention is to employ a Mahalanobis distance novelty detection model by comparing the incoming audio signal value with a previously developed principal component analysis (PCA) model for sensor types and its signal and / or data values.

[0497] Signal processing involves acquiring the audio signal and applying noise reduction algorithms to the raw data. The audio signal is further processed using a short-time Fourier transform to convert the raw time-series data to the frequency domain. Statistical features are then extracted from each of a series of bandwidths spanning the frequency range of interest. In this example, the bandwidth used is 39.1 kHz, resulting in 312 features describing the acoustic signal in a given signal instance. The total feature set [312 features for each signal instance] is then optimized by removing redundant features and subsequently standardized to improve the robustness of features with smaller standard deviations.

[0498] In the initial establishment of the PCA model, the signal and thus the residual feature set generated by the above processing are established to obtain acceptable welding performance relative to the variable (in this case, the acoustic signal). Therefore, the residual feature set can be finally reduced using PCA to give the model and the principal components defining the model with respect to acceptable welding performance.

[0499] Then, a PCA model with acceptable performance can be used to consider subsequent signals. These subsequent signals undergo the same noise cancellation and other steps defined above. For each feature, the location of the eigenvalue is determined relative to the distribution of eigenvalues ​​used to establish the acceptable weld performance, and thus the eigenvalue is determined. The mean of the distribution is calculated, and the distance of the eigenvalue is measured relative to this mean. This distance provides... Figure 19 The outliers shown are quantified, and thus distance variations that are still consistent with acceptable welding conditions are established, as well as distance variations that are more anomalous and thus indicate impaired welding performance.

[0500] This method is advantageous in providing a unitless, scale-invariant quantization that takes into account the correlation of eigenvalues ​​within the distribution. The mean can be recalculated when acceptable eigenvalues ​​are added to the distribution and / or the PCA model, or the mean can be based on a fixed set of pre-existing acceptable eigenvalues ​​used in the PCA model.

[0501] Reference obtained using this process and Figure 19 The example results shown indicate that the first set of data points demonstrates favorable welding operation parameters. These parameters were independently verified as applicable. It can be seen that these parameters provide a good clustering set of data points A relative to the logarithmic outlier fraction axis.

[0502] The second set of data points illustrates excessively high welding speeds; that is, the welding equipment and the substrate are moving too fast relative to each other. Two distinct welding speed deviations are demonstrated, and these deviations give two well-clustered sets of data points B1 and B2 with few outliers.

[0503] The third set of data points illustrates the sidewall arc initiation during welding; that is, the arc is short-circuited with the sidewall rather than with the desired welding position within the weld bevel. Furthermore, these conditions yield a good cluster set of data points C with few outliers.

[0504] The fourth set of data points illustrates sidewall fusion; that is, the arc melts the sidewall and causes fusion there rather than in the weld bevel. The data points are well clustered into data point D with very few outliers.

[0505] The fifth data set indicates that the protective gas flow rate is too high; this can lead to undesirable porosity problems. Like the other data set sets, this set E is well-defined with few outliers. Similar locations can be detected when the protective gas flow rate is too low.

[0506] When present, the individual undesirable welding conditions mentioned give much higher outlier or eigenvalues ​​than when good welding conditions are present. Therefore, the threshold Th acoustic The selected value for the [Madara distance] can be set and used to differentiate between acoustic sensor types that indicate good welding conditions and those that indicate damaged welding conditions. Therefore, if the threshold Th... acoustic For a given number of data points that are breached or remain breached, the acoustic data type has the capability to issue warnings, trigger welding stoppages, etc. Therefore, it provides defect identification based on real-time acoustic signals. This analysis can be provided continuously and for each weld pass in a multi-pass weld method.

[0507] Importantly, the algorithm used is more complex than that using only a single previously developed principal component analysis (PCA) model for sensor type and its signal and / or data values. As seen in Figure 20b, in multi-pass welding, the weld beads gradually fill the weld groove. This means that the depth, shape, and filling volume of the weld groove vary with each weld pass. All of these, along with other variations that may occur between weld passes, affect the emitted and detected acoustic signals. Therefore, this method uses a separate model for each weld pass to provide the most accurate evaluation of the observed data against the expected data for that weld pass.

[0508] Individual models can be obtained through neural network methods, trained based on pre-existing weld beads, or learned as weld beads are made during operation, where the increasing number of individual weld beads improves the model for a given weld bead in the weld bead sequence.

[0509] The use of separate models for each weld bead has been extended to the use of separate models for repeated weld beads, for example, as part of remedial work on the weld. The weld bevel shape and the resulting acoustic effects in such repeated weld bead cases are significantly different from those of ordinary weld beads.

[0510] Laser sensor type

[0511] The next type of sensor used is a vision sensor, and this seeks to assess the geometry of the resulting weld.

[0512] Figure 20a illustrates a cross-section of a substrate 2 on which a weld 20 is formed. In this example, the weld 20 is linear, but other weld trajectories can be considered in the same way. The vision sensor type device 22 includes a housing 24 within which a light source 26 is provided, which can illuminate the substrate 2 and the weld 20 across an illumination width 28. The device 22 has an operating range 30 within which accurate imaging can be performed. Light returns to the device 22, where a receiver 32 focuses the light onto a sensor matrix 34 and generates a signal. In this example, a 2D laser profile scanner is used; other types of scanners can be substituted.

[0513] The laser, serving as light source 26, is used to establish various details of the weld and its surroundings, including the weld profile, the remaining bevel profile, the weld width, and any material deposited outside the weld bevel. A plane perpendicular to the substrate surface adjacent to the weld bevel and perpendicular to the longitudinal axis of the weld bevel can be used for inspection. Additionally, the weld profile along the weld bevel can be considered.

[0514] Figure 20b is an illustration of a typical weld bevel during a series of welding passes. Individual weld passes are added to the existing weld bevel in a predetermined order (such as numbering) to establish the entire weld. It can be seen that, if the welding is performed correctly, the weld passes contribute a predictable geometry to the weld pass itself and a predictable variation in the geometry of the weld bevel.

[0515] The signal from device 22 can be used to form a profile image across the weld track at various locations along the weld track. This actual profile can be compared to the expected profile and deviations can be annotated. The deviation can be compared with a threshold Th. profile Compare them. Therefore, if the threshold Th profile For a given number of data points that are breached or remain breached, the profile data has the capability to issue warnings, trigger welding stoppages, etc. Therefore, defect generation identification based on real-time profile signals is provided. This analysis can be provided continuously and for individual weld passes in multi-pass weld methods. Effective geometric verification is provided.

[0516] Visual sensor types

[0517] In the next sensing area, a high dynamic range camera is used to acquire images of the welding locations, including locations where a weld has not yet formed, locations where a weld is forming, and locations where the weld is solidifying and then further cooling.

[0518] Figure 21This illustrates a series of images of this type collected from different weld locations. After processing these images using a combination of artificial intelligence and conventional machine vision tools, the system can detect changes in the images related to abnormal welding conditions or visual anomalies caused by defects. This can be achieved using processing of single images from recent images as well as historical sections and weld beads, or a combination of multiple images from previous data.

[0519] One or more variables can be considered in this area, such as weld pool size (width, tail length, leader length), weld pool geometry (elliptical, teardrop, or other shapes), and weld pool temperature, which can be considered in conjunction with the analysis of pattern, shape of deposited material, and visible anomalies. Distribution maps of known judgment parameters for different types of defects can be used to compare the real-time output from various parts of the algorithm with known good values. These values ​​can be the presence or absence of specific visual features, numerical bands or thresholds, or combinations of categories. If an image is deemed to meet more than these parameters, the occurrence of one or more or all of these parameters can be compared with the expected location for one or more or all of these parameters, and the deviation can be used again to trigger a welding warning or stop.

[0520] Voltage and current sensor types

[0521] The applied voltage affects the formation of the electric arc and the current within it. This, in turn, affects the power, and therefore the melting rate of the substrate [and the electrodes, if consumed]. These are important variables affecting weld quality. These are, for example, variables that affect weld pool size.

[0522] The welding voltage also needs to be automatically and continuously adjusted to reflect the interval between the welding equipment and the substrate being welded. This is based on the known and fixed position of the substrate and the variable but known XYZ position of the robotic arm carrying the welding equipment, and therefore on the interval between the two.

[0523] The sensing system for voltage and current monitoring is orders of magnitude faster than those incorporated into pre-existing automatic voltage control methods. The power supply system can handle 500 A and scale up to 1000 A or more, while still providing nanosecond-level voltage and current monitoring between measurements. Therefore, very detailed information on voltage and current is obtained, even taking into account transient variations.

[0524] Input voltage and current, and therefore input power, are typically detected, rather than output power, to avoid sensing / detection itself compromising output power performance.

[0525] The data and therefore the methods used to process it are similar to those described above for acoustic sensor types.

[0526] refer to Figure 22 The arc voltage is plotted as a curve V relative to time. In this example, the welding process stops after 30 seconds, so the voltage returns to zero.

[0527] Figure 22 The graph also plots the Gaussian amplitude x Gaussian center versus time. It can be seen that in the initial time period, i.e., the first 2 or 3 seconds, the value of this curve is above the acceptable threshold, thus indicating concern about weld quality. After this initial 2-3 seconds, and naturally forward to after 8 seconds, the curve is significantly below the threshold, indicating a lower quality weld relative to this set of variables.

[0528] In addition to these variables, Figure 22 It also includes an indication of when low argon is present as a protective gas. This indication is shown by plotting a low argon level as a point, and also by showing the degree of argon deficiency by the vertical position of the plotted point.

[0529] The whole process - continue

[0530] Back to Figure 17 The details of sensor type operation have been established, taking into account their data, processes, and the data conditioning steps involved. Data conditioning is provided based on inputs provided through the user interface and / or based on historical data obtained from storage. Storage may contain data from earlier times during the weld and / or data from numerous previous welds performed by the system and / or data from other systems [e.g., from calibration processes, etc.], all of which can contribute to historical data and thus to data conditions.

[0531] In the real-time analysis step, two different levels of processing can be achieved.

[0532] The first-level processing provides synchronization of data types from different sensor types. This synchronization can accommodate not only different data types from the different sensor types exemplified above, but also any number of other sensor types introduced and used to measure key characteristics directly and / or indirectly related to the welding system and the resulting weld.

[0533] Programmable Logic Controllers (PLCs) are used to synchronize different data types in time by providing a master timestamp when data collection from sensor types within the system is initiated. The PLC also continuously checks whether the various sensors of each sensor type within the system continue to provide data and whether the provided data is being collected at a consistent rate. Additional master timestamps can be applied periodically during the data collection process.

[0534] The master timestamp means that data collected from microphones that act as sensors for acoustic analysis, input power analysis that measures traction during welding, area scanning cameras that provide vision systems, and laser profilers that provide 3D profile sensors can all be aligned to represent data from the same time and therefore from the same location within the weld.

[0535] Importantly, the PLC also receives data from incremental encoders that provide position data. The encoder can be mounted on the substrate being welded and / or on the welding equipment, indicating its physical position at that time. The encoder triggers data collection and provides a consistent correlation between sensor data and the weld position on the component. Again, the data from the encoder has the same master timestamp. This means that for the same established position that is simultaneously applicable in the alignment of synchronized data, the actual position is known. The data is synchronized in time and space.

[0536] This processing enables sensor data collected at different locations on the component to be post-processed into a cohesive data structure, where the raw sensor data, post-processed information, and analytical output are correlated with the physical location of the weld. This data can then be displayed to the operator on a user monitor, allowing them to adjust the welding process based on the system's output.

[0537] The second-level processing is achieved through the application of data reduction and / or machine learning. This second-level processing acquires the relevant data and then feeds it to a second-level processing and analysis (using machine learning or other appropriate analytical methods) to correlate any number of features from both the raw data and the output of the sensor-type level analysis with welding defects. In practice, a decision engine supporting machine learning is employed.

[0538] In this method, a general assessment is made as to whether a defect (in the case of suspected location) falls within or exceeds acceptable characteristics (e.g., size) and therefore requires attention or remedial action. When NDT-type sensing is performed, the defect is directly measured using the imaging process. This directly reports the defect's size and other possible characteristics. However, when the consideration of a defect is based on the welding conditions occurring during welding, potential defects are considered indirectly; the question "Could these conditions have led to the defect?" is considered. In this case, a second-level processing provides for the determination of improvements to acceptable and / or unacceptable welds.

[0539] refer to Figure 24 Consider two different sensor types, Type A on the left-hand side and Type B on the right-hand side. These two sensor types can be any sensor type mentioned in this article and / or other sensor types that provide data on weld behavior or weld results.

[0540] For reference type A sensor type, a series of data points will be established within region 800 to indicate acceptable welds. These data points can be established from test runs verified by other sensing and / or NDT, or they can be modeled.

[0541] There will also be data points from the weld, such as data point 802, which are clearly outside the acceptable area and also a long distance from the unacceptable side of threshold 804. These can be used to make an initial determination as to whether the data point indicates an acceptable or unacceptable weld. Other data points, such as data point 806, are above the threshold and are also considered unacceptable according to the definition of threshold 804.

[0542] Examples of data points that are difficult to interpret include data points 808 and 810, which are outside region 800 but below threshold 804. A call based on a single sensor would lead to these data points being considered acceptable welds because they are below the threshold.

[0543] The second-level processing will yield a gain that takes into account data points [802, 806, 808, 810] spanning more than one sensor type to achieve complete determination. Referring to the right-hand side and the Type B sensor type, data point 802 again lies within the established acceptable weld area 800. Similarly, data point 806, far exceeding the threshold 804 and again originating only from this sensor type, indicates an unacceptable weld.

[0544] Turning back to data points 808 and 810, both are outside region 800 but below threshold 804. Under a single sensor type approach, these edge cases would again be considered acceptable welds. However, the second-level processing gains additional information by considering locations across multiple sensor types.

[0545] Neural networks can be used to employ both approaches that consider locations across multiple sensor types. These two approaches can be used as alternatives to each other, or one approach can be used in parallel or serially with the other.

[0546] In the first approach, labeled data is provided, and supervised learning is involved. The labeled data can come from one or both sources. First, particularly in the early stages of processing, such as during calibration or early production welding runs, labeled data can be obtained from experimental results. Therefore, continuing... Figure 24For example, the positions of both points 808 and 810 can be labeled for the operator to use in an operator-based assessment of whether the data is acceptable or unacceptable. The operator is not only provided with data from a single sensor type for consideration and invocation, but also with labeled data [data points or sequences of data points] from multiple different sensor types on that data, and thus a more granular location from which the invocation is determined. The results are used to label the data and are therefore available in the data pool from which the neural network learns. Human knowledge and interpretation are fed into the neural network through user determination, thus providing supervision.

[0547] A second way to provide labeled data is to utilize an existing database. Again, this data relates to the determination of labeling and has the necessary transfer of operator knowledge. This database is fed into a neural network classifier to establish the processing location of the database dataset. The database will cover data from multiple sensor types, thus allowing for the determination of the second-level processing type; data point locations span inspections and results across multiple sensor types. The classifier score for any data point associated with this database data can then be determined, and consequently, the possible error of the data point (or its sequence) to be inspected and the determination to be made can be quantified. The classifier can use a Bayesian-based method for classification and error quantification.

[0548] The second approach (library) can be used as the initial labeled dataset. The first approach (calibration or test runs performed on an actual welding system by operator invocation) can be used as an alternative from the outset. However, the first approach can be used to add datasets to the second approach, and thus allow the neural network to learn from a more general welding system position to a position more suited to that specific welding system.

[0549] Return to Figure 24 An operator might determine that data point 808 is close enough to the acceptable region 800 to be considered acceptable, but data point 810 across multiple sensor types might be considered so close to a threshold that it is unacceptable. This could lead the neural network to make small adjustments to the acceptable region 800 and / or the threshold 804 [value or format]. Over time, this type of repeated determination can result in more explicit and optimized corrections to the boundaries of the acceptable region 800 and / or the threshold 804. For example, the acceptable region 800 might be expanded and / or the threshold might be tightened. The same results for data points 808 and 810 in later, more advanced learning stages of this process might be labeled as acceptable and unacceptable, respectively, depending on where they lie within the corrected acceptable region for data point 808 and above the corrected threshold for data point 810.

[0550] An example of this analysis in a real-world scenario could be the left-hand side being associated with the type of visual image sensor, and one or more consecutive images suggesting an issue with the sidewall proximity of the weld. Considering the type of acoustic sensor as the right-hand side could also suggest a sidewall proximity issue, thus validating the overall determination that the weld is unacceptable.

[0551] While the examples above involve locations from both the sensor type and other sensor types that provide an acceptable or unacceptable determination, this determination can be more detailed than simply defining the nature of the problem based on welding conditions. Therefore, data from two sensor types can inform the nature of the problem, whereas data from only one sensor type might merely suggest the issue.

[0552] It is important to note that inputting historical data from other welding conditions and environments is not a strong starting point for determining acceptable weld performance. This is because other welding operations and environments have a wide variety of other variables that can affect the data from those operations. For example, in the case of visual imaging sensors, lighting and lighting angles, substrate properties, torch angles and spacing, etc., in that welding environment can all affect the data and cause it to be inconsistent with the data sought, for example, in another environment with different lighting.

[0553] As an alternative to or complement to the first approach using labeled data and supervised learning, the need for library-type data and / or supervised learning can be reduced or avoided by leveraging the ability of neural networks to perform clustering or grouping-based processing [finding similarities and / or anomalies in the data]. Various techniques exist, such as K-means clustering, which enable the establishment of centroids of clusters and a degree of certainty regarding the distances established around these centroids. Many other clustering methods are applicable. These methods can be used to establish acceptable regions at 800 and / or threshold locations at 804, and to modify them with additional data and learning, without requiring a large library of labeled data to begin with.

[0554] As previously mentioned, the first and second methods can be used in combination or as alternatives. Thus, the first method can be used to initiate the neural network during learning, and then the second method can take over after the first method has advanced the learning process. The neural network can also learn from both methods simultaneously to maximize the supplied data, particularly in cases where the library is being used for welding extensions from welding systems other than the specific welding system under consideration. Learning can also be gleaned from welding systems with similar configurations and operations.

[0555] Over time, especially when the method is used in production versions that perform a large number of welds, the amount of data, the accuracy of the evaluation, and the complexity of different scenarios for successfully evaluated data increase.

[0556] Any and all results from real-time analysis and / or data reduction and / or machine learning of the data may be copied and fed into storage for future use or to contribute to the knowledge of the system and / or similar systems.

[0557] A key step in the analysis is thresholding, which compares data values ​​or locations against one or more thresholds set for that data type. Examples of thresholding methods are provided in different sections for a specific sensor type, but they are broadly applicable to various sensor types and the data types they generate. The location relative to a threshold can be considered an indication of a defect, i.e., a defect indicator, and is therefore a crucial part of the weld quality results, which are then displayed to the user via a user display that also receives and displays the received sensor data.

[0558] Figure 23 This is an example of a user display. The user display can provide real-time output of analysis for all sensor types, allowing them to review the welding process as it unfolds and adjust the process if the analysis output identifies any poor-quality welds. Additionally, they will be able to examine the overall weld data after each individual weld pass and query that data to identify any areas requiring remediation or adjustment before or during subsequent welding.

[0559] The weld quality results step can stop the welding process and / or provide a warning to the operator, for example, via a user display, when an unacceptable weld is encountered.

[0560] Based on the user's display location, the operator or even the system itself can adjust one or more variable parameters used to control and perform the welding process in real time.

[0561] Weld quality results are supplied to weld qualification for the entire weld. In the case of defects identified by NDT, the size and location of the defects are cross-checked against appropriate standards to ensure that the defects are acceptable at that level. If not, remedial action is taken. If acceptable, the 3D lifetime record of the data is maintained within the weld qualification so that it is available throughout the weld life and can be incorporated into any necessary decommissioning support.

[0562] Data processing and storage

[0563] Besides considering individual data types individually, viewing data types as composite data streams has another advantage. To achieve this, each dataset is timestamped, allowing all datasets to be synchronized by synchronizing their timestamps.

[0564] This means that all datasets can be combined and then saved as a single data file. This also allows for reuse at a later time.

Claims

1. A method for detecting electromagnetic interference in ultrasonic imaging of a substrate, comprising the steps of: transmitting a plurality of ultrasonic signals into the substrate using an ultrasonic transducer, wherein the plurality of ultrasonic signals are separated in time by a pulse repetition period; detecting a return signal for each of the plurality of ultrasonic signals and recording the return signals as signal traces; rectifying each of the signal traces to produce a plurality of rectified traces; identifying a minimum value in all of the plurality of rectified traces at each sample; for each of the plurality of rectified traces, deriving a difference between the value at each time sample and the minimum value at each time sample; when the difference exceeds a threshold value in any of the rectified traces, flagging the value of the time sample in that rectified trace as interference.

2. The method of claim 1, wherein, the threshold value is a constant value at all time samples.

3. The method of claim 1, wherein, the threshold value is adaptive and is determined by multiplying the minimum value at the time sample by a defined factor.

4. The method according to any one of the preceding claims, wherein, the step of rectifying each of the signal traces comprises performing envelope detection on each of the signal traces.

5. The method of claim 1, wherein, the ultrasonic transducer is a phased array ultrasonic transducer comprising a plurality of elements, wherein each element of the ultrasonic transducer sequentially transmits a plurality of ultrasonic signals separated by a pulse repetition period; and wherein each of the plurality of elements detects a set of return signals corresponding to each transmitted signal in parallel.

6. A method for suppressing electromagnetic interference in ultrasonic imaging of a substrate, comprising the method of detecting electromagnetic interference of claims 1 to 4, further comprising the steps of: rejecting the value in each rectified trace identified as interference; for each time sample, averaging any remaining values in each rectified trace to give an average value; and forming a processed trace from the average values.

7. The method of claim 6, wherein, the step of forming a processed trace from the average values comprises taking the average value in the unrectified signal traces corresponding to the remaining values at the same time sample in each rectified trace, and forming an unrectified processed trace from the average values in the unrectified signal traces.

8. A method for suppressing electromagnetic interference in ultrasonic imaging of a substrate, comprising the method of detecting electromagnetic interference of claims 1 to 4, further comprising the steps of: rejecting the value in each rectified trace identified as interference; taking the minimum value from any remaining values at each sample; and forming a processed trace from the minimum values.

9. A method for suppressing electromagnetic interference in ultrasonic imaging of a substrate, comprising the method of detecting electromagnetic interference of claim 5, wherein the method further comprises the steps of: selecting a set of return signals as a representative set of signals, wherein the steps of rectifying each of the signal traces, identifying a minimum value at each time sample, deriving a difference between the value at each time sample and the minimum value, and flagging the value at the time sample as interference when the difference exceeds a threshold value are performed on the representative set of signals; and flagging values in other sets of return signals as interference corresponding to the same time samples flagged as interference in the representative set of signals.

10. The method of claim 9, further comprising the steps of: rejecting values in each return signal that are marked as interference; and forming a processed signal trace by combining any remaining values at each time sample that are not marked as interference.

11. The method of claim 10, wherein, The return signals and the processed signal trace are unrectified.

12. The method of claim 10, wherein, The pulse repetition rate is selected to minimize aliasing patterns.

Images