WO2020205998A1

WO2020205998A1 - Non-destructive evaluation and weld-to-weld adaptive control of metal resistance spot welds via topographical data collection and analysis

Info

Publication number: WO2020205998A1
Application number: PCT/US2020/026218
Authority: WO
Inventors: David Fulton; Akshay HARI; Jared FRIANT; Donald J. Spinella; Michael A. Globig; David E. Coleman; Gino N. Iasella
Original assignee: Arconic Technologies LLC
Current assignee: Arconic Technologies LLC
Priority date: 2019-04-05
Filing date: 2020-04-01
Publication date: 2020-10-08
Anticipated expiration: 2021-10-05

Abstract

An apparatus and method for non-destructive evaluation of resistance spot welds uses a line laser triangulation sensor, a linear encoder, and a device for moving the welds relative to the sensor to measure the three-dimensional shape of one or both surfaces of the weld. The shape data is analyzed by an artificially intelligent system that predicts weld quality based upon the shape data. The gradient normal of points on the weld surface that are high are indicative of high slope attributable to electrode degradation which can be correlated to weld quality in making the prediction. The apparatus and technique can also be applied to the evaluation of RSR joints.

Description

NON-DESTRUCTIVE EVALUATION & WELD-TO-WELD ADAPTIVE CONTROL OF

METAL RESISTANCE SPOT WELDS VIA TOPOGRAPHICAL DATA COLLECTION AND

ANALYSIS

FIELD OF THE TECHNOLOGY

[0001] The present disclosure relates to apparatuses and methods for measuring and assessing welds. More particularly, the disclosure relates to apparatuses and methods for non-destructive assessment of resistance spot welds of aluminum alloys.

BACKGROUND

[0002] Resistance spot welding is widely used for spot welding of steel and other metals, particularly in the assembly of automobile bodies and truck bodies. An apparatus for resistance spot welding includes a pair of resistance welding electrodes. Typically, a robot weld gun fitted with a pair of electrodes is moved in stages along a continuous weld path. At each stage, the electrodes are pressed against opposing sides of the work pieces to be welded, and an electric current is passed through the electrodes in the work pieces. The electrical resistance of the metal work pieces produces localized heating which causes the work pieces to fuse at a weld site. The electrical heating at the point of pressure between the electrodes forms a molten nugget at the interface between the work pieces. Typically, welds are oval in shape. The ovality of a weld nugget is dependent on the rigidity of the welding equipment and the condition of the electrodes. The diameter of a weld is considered to be an average value for the weld and is conventionally determined to be the average of the maximum diameter and the minimum diameter of the weld nugget. Apparatuses and methods for assessing resistance spot welds (RSWs) are known, e.g., as shown in U.S. Patent Nos. 6,403,913 and 9,927,367. Notwithstanding, alternative apparatuses and methods are still of interest in the field.

SUMMARY

[0003] The disclosed subject matter relates to a method for assessing a weld joint, including the steps of: (A) providing a weld joint; (B) measuring a three-dimensional shape of a surface of the weld joint to determine shape data; (C) analyzing the shape data; (D) predicting a quality of the weld joint based upon the step of analyzing; and (E) generating an output in response to weld quality. [0004] In accordance with another aspect of the present disclosure, the output includes controlling at least one welding process parameter.

[0005] In accordance with another aspect of the present disclosure, the step of analyzing is by machine intelligence trained by training data correlating a plurality of shape data to weld quality.

[0006] In accordance with another aspect of the present disclosure, the step of measuring includes projecting a laser line on the surface; receiving reflected light from the surface in a sensor; and moving the weld joint and the sensor relative to each other from a first position to a second position.

[0007] In accordance with another aspect of the present disclosure, the shape data includes slope data and the step of analysis includes interpreting the slope data as indicative of electrode condition.

[0008] In accordance with another aspect of the present disclosure, the shape data includes roughness, skewness, root mean square (RMS), peak-to-peak distance, valley depth, peak height, and combinations thereof.

[0009] In accordance with another aspect of the present disclosure, the weld j oint is formed by resistance spot welding (RSW), has two side surfaces, and the step of measuring is conducted on one or more of the two side surfaces of the weld joint.

[0010] In accordance with another aspect of the present disclosure, the weld joint is formed by resistance spot riveting (RSR), has two side surfaces, and the step of measuring is conducted on one or more of the two side surfaces of the weld joint.

[0011] In accordance with another aspect of the present disclosure, a plurality of weld joints is provided during the step of providing and the steps of measuring and analyzing are conducted on the plurality of weld joints in a batch.

[0012] In accordance with another aspect of the present disclosure, the step of measuring is conducted a plurality of times for the weld joint.

[0013] In accordance with another aspect of the present disclosure, the weld joint joins a first aluminum member to a second aluminum member and is an RSW.

[0014] In accordance with another aspect of the present disclosure, further including the steps of obtaining additional weld measurement data and analyzing the additional weld measurement data.

[0015] In accordance with another aspect of the present disclosure, the output is indicative of a non-optimal weld and the method further includes the step of making a supplemental weld.

[0016] In accordance with another aspect of the present disclosure, the method further includes a step of referring to a reference weld pattern to determine a location of the supplemental weld.

[0017] In accordance with another aspect of the present disclosure, the weld joint is prescribed by an optimally efficient weld pattern having a minimum number of good quality welds for forming a structure and further including the step of providing a supplemental weld if the step of predicting indicates that the weld joint is discrepant.

[0018] In accordance with another aspect of the present disclosure, the step of measuring includes identifying the shape data associated with the weld joint distinct from an adjacent unwelded area, replacing outliers in the shape data with interpolated values from neighboring points, and overlaying a best-fit surface to remove macro form.

[0019] In accordance with another aspect of the present disclosure, the step of identifying includes finding a minimum Z-value within a sample of shape data, assigning the minimum Z- value to be a center of a weld dimple, and calculating the location of the remainder of the weld dimple based upon the geometry of a welding electrode that made the weld dimple.

[0020] In accordance with another aspect of the present disclosure, the step of analyzing includes calculating cartesian gradients: dz/dx and dz/dy; and computing non-negative gradient norm: (dz/dx)²+(dz/dy)²]^1/2.

[0021] In accordance with another aspect of the present disclosure, further including the steps of displaying a gradient normal image for a weld dimple; inspecting for scratches on the surface of the weld; remediating data corresponding to a scratch; computing histogram and 95% quantile of gradient normal values for the weld dimple; and assigning a gradient norm. For each weld dimple, a portion of the histogram of gradient normals represents the steepest slopes in the image and is indicative of electrode tip degradation and poor-quality welds.

[0022] In accordance with another aspect of the present disclosure, an apparatus for assessing weld quality of welds in a welded structure, includes: a line laser triangulation sensor; a linear encoder; a device for moving the welded structure relative to the line laser triangulation sensor, the linear encoder measuring the movement of the welded structure relative to the line laser triangulation sensor; a computer with a three-dimensional measuring routine runnable thereon, the computer receiving as input the outputs of the line laser triangulation sensor and the linear encoder and capable of calculating three-dimensional measurements of a surface of the weld, the computer further capable of analyzing the three-dimensional measurements to calculate surface slope data indicative of weld quality.

[0023] In accordance with another aspect of the present disclosure, further including a welding apparatus connected to and responsive to the computer data indicative of weld quality, the computer capable of sending control signals to the welding apparatus to remediate discrepant welds.

[0024] A method for assessing a weld joint is also disclosed. The method includes projecting, via a projector, a laser line on a surface of a resistance spot welded (RSW) weld joint. Light reflected from the surface is received by a sensor. Shape data defining a three-dimensional shape of the surface based on the light reflected from the surface is produced. Based at least in part upon the shape data, quality of the weld joint is predicted. The predicting includes comparing the shape data to model shape data corresponding to a plurality of model weld joints. An output is generated by the computer in response to the quality of the weld joint as predicted.

[0025] An apparatus for assessing weld quality of weld joints in a welded structure is also disclosed. The apparatus includes a line laser triangulation sensor; a linear encoder; and a device for moving the welded structure relative to the line laser triangulation sensor. The linear encoder is configured to measure movement of the welded structure relative to the line laser triangulation sensor. A computer with a three-dimensional measuring routine is configured to receive as input, outputs of the line laser triangulation sensor and the linear encoder, and to calculate three- dimensional measurements of a surface of the weld joints. The computer is further configured to predict a weld quality based on the three-dimensional measurements relative to a plurality of model weld joints, and to output an indicator of the weld quality of the weld joints.

BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Embodiments of the present disclosure can be understood by reference to the illustrative embodiments depicted in the appended drawings.

[0027] FIG. 1 is a diagrammatic view of an apparatus in accordance with an exemplary embodiment of the present disclosure.

[0028] FIG. 2 is a computer-generated graphic of a weld dimple in accordance with another exemplary embodiment of the present disclosure.

[0029] FIGS. 3A, 3B, 3C and 4 are sets of computer-generated graphics of dimension data of a weld dimple in different stages of processing in accordance with another exemplary embodiment of the present disclosure.

[0030] FIG. 5 is a set of computer-generated graphics of dimension data of a plurality of weld dimples in accordance with another exemplary embodiment of the present disclosure.

[0031] FIGS. 6A, 6B, 6C, 6D and 6E are computer-generated graphics of dimension data for a plurality of weld dimples in different stages of processing in accordance with another exemplary embodiment of the present disclosure.

[0032] FIGS. 7A, 7B, 7C, 7D, 7E, 7F, 7G, 7H, 71, 7J and 7K are computer-generated graphics of dimension data for a plurality of weld dimples for which a gradient norm has been calculated and classified as good or discrepant welds in accordance with another exemplary embodiment of the present disclosure.

[0033] FIGS. 8A, 8B and 8C are sets of histograms of gradient norm (i.e., maximum-descent slope, in angular degrees) distribution for ten welds measured in accordance with an exemplary embodiment of the present disclosure.

[0034] FIG. 9 is a graph of weld dimple dimension data for a plurality of welds processed to calculate the 95% quantile angle in degrees vs. the start weld number of a set of 10 welds in accordance with an embodiment of the present disclosure.

[0035] FIG. 10 is a perspective view of a scanning apparatus in accordance with an embodiment of the present disclosure.

[0036] FIG. 11 is a plan view of a movable sample support in accordance with an embodiment of the present disclosure.

[0037] FIG. 12 is a side view diagram of a movable sample support with a stationary scanning head in accordance with an embodiment of the present disclosure.

[0038] FIG. 13 is a plan view of the moving sample support of FIG. 12.

DETAILED DESCRIPTION

[0039] An aspect of the present disclosure is the recognition that as automotive original equipment manufacturers (OEMs) transition from steel to aluminum, the need for reliable nondestructive evaluation (NDE) and weld-to-weld adaptive control of resistance spot welds (RSW) is imperative.

[0040] “Weld-to-weld” adaptive control refers to a process that is continuously or periodically applied to welds shortly after they are made to assess the condition of acceptability of the welds, to perceive trends indicative of the effectiveness and acceptability of the welds made, to identify and characterize welds as either good or bad (discrepant), to correlate weld assessment to welding parameters, to adjust the welding parameters in response to the assessment of welds, and to identify and eliminate discrepant welds from the finished product line at as early a stage of processing/manufacture as possible.

[0041] A“good weld” may be defined as a weld having mechanical properties, such as tensile strength, meeting the specifications of a particular application, e.g., securing two vehicle body panels together into an assembly which has predefined rigidity, weight-bearing capacity, stress resistance, etc., for a pre-defmed useful life at a pre-defmed level of usage.

[0042] In contrast, a“discrepant weld” has mechanical properties that are deficient with respect to a given standard for a particular use.

[0043] “Model welds” include a plurality of known good and discrepant welds.

[0044] The weld-to-weld approach has the potential to be more efficient than procedures based on worst-case assumptions. For example, testing may reveal that electrode tip dressing for a given application may be required between every 50 and 150 welds to achieve electrical usage efficiency and welds of acceptable quality. To accommodate this variability, a worst-case scenario would be assumed and a safety factor applied, such that a tip dressing schedule of, e.g., once every 40 welds could be used.

[0045] In contrast, continuously monitoring indicia of weld quality can avoid an unnecessarily high rate of tip dressing and still be sensitive to a weld that is characteristic of historic worst-case weld behavior, or one where quality is even less than previously historically observed, e.g., due to an unusual malfunction or other unprecedented circumstance.

[0046] In an alternative approach, a plurality of welds may be assessed in a batch, e.g., on an assembly or portion of an assembly, e.g., as a quality control measure. The batch assessment may be in addition to individual assessment, that is, the apparatus and methods of the present disclosure may be applied multiple times to a given weld.

[0047] In addition to being a more challenging material to spot weld, aluminum and its alloys do not have the decades of historical welding data to refer to as are available for RSW of steel. Existing inspection/qualification technologies that have been developed for steel do not transition well to aluminum, leading to automotive OEMs lacking confidence in spot welding aluminum. This lack of confidence directly impacts the rate of vehicle“aluminization”, due to several factors, such as, the higher cost of more trusted joining technologies for aluminum, e.g. self-pierce riveting, a short RSW electrode tip life due to over-dressing, and potentially increased equipment downtime due to the high frequency of electrode tip dressing.

[0048] In accordance with an aspect of the present disclosure a line laser triangulation sensor, when paired with a linear encoder, measures the three-dimensional shape of a target object. In one embodiment, the sensor and encoder measure the 3D topography of a metal resistance spot weld. This topographical data is then analyzed via advanced analytics methods, enabling both the prediction of weld quality, and the weld-to-weld adaptation of process parameters (current, force, time) to ensure high quality future welds.

[0049] Note that“weld quality” refers to weld button diameter, thickness, and a binary good vs. discrepant assessment. As noted above, this technique can be applied in both an in-line, real time production setting, and in an offline, batch processing environment. When used in a batch process, a set of welds that have already been made can be assessed for quality and those falling short of a given standard can be identified, e.g., pursuant to quality assurance testing of a set of finished welds. If testing of a set of welds reveals one or more discrepant welds, a decision can be made as to whether the assembly can be remediated, e.g., by adding one or more welds, or whether the assembly must be scrapped/recycled. The topography data can be combined with other evaluation measurements (ultrasound, welding current/voltage/force, etc.) and subjected to the analytics methods described below to improve predictive capability.

[0050] While the present application provides technology applicable to evaluating welds in aluminum structures, it is also applicable to assessment of welds in structures made from other materials, such as steel, other metals, and combinations of different metals, such as aluminum with steel.

[0051] FIG. 1 shows a weld assessment system 10 according to an embodiment. In the illustrated weld assessment system 10, a line laser triangulation sensor 12 includes a projector and a sensor. The projector projects a laser line P onto a spot weld W of a sample S. Reflected light R is reflected from a surface of the sample S and is received by the sensor of the line laser triangulation sensor 12.

[0052] The sample S in which the spot weld W is present has two or more layers SI, S2. The sample S moves relative to the sensor 12 as indicated by the arrows Al, A2. Unless otherwise more specifically defined below, the terms“weld,”“spot weld,””weld joint,”“weld dimple” or “dimple” may be used interchangeably herein in that a resistance spot weld joins a plurality of layers of material, e.g., SI, S2 (which layers SI, S2 may be made from independent parts or structures or may be a single part, such as a sheet, bent back on itself) resulting in a weldment or weld nugget (merged metal materials from a plurality of layers) that joins the layers forming a weld joint and exhibiting a weld dimple on one or both sides of the resultant conjoined, welded structure, the weld dimple being the impression made by the welding electrode on one or more surfaces of the structure resulting from resistance spot welding.

[0053] The present disclosure contemplates that either the sample S, the sensor 12, or both may be moved relative to the other. A resultant signal 13 corresponding to the reflected light R is transmitted with position data 15 from a linear encoder 14 sensitive to the relative position of the sample S and the line laser triangulation sensor 12, to a computer 16. The computer 16 combines the resultant signal 13 and the position data 15 to yield measurement data representing a measure of the three-dimensional shape (3D topography) of the spot weld W in the X, Y, and Z directions.

[0054] The measurement data (also referred to as“shape data” herein) may be analyzed by a weld analysis program P running in the computer 16. The computer 16 may also receive other weld data 17, such as weld measurement data obtained from additional weld measuring and assessment devices(s) 18. Examples of such weld data 17 potentially including photographic, thermal imaging, ultrasonic analysis data, welding parameter data associated with the weld W, such as current, resistance, force, and time settings from the welding machine and interactive measurements of same taken when forming the weld W, along with combinations thereof.

[0055] The analysis program P running on computer 16 may generate a variety of output data, control signals 20, or combinations thereof, such as display data, reports, or auditory signals representing the assessment of the weld W, that is received by a device 21, such as a display, printer, speaker, weld controller, robotic arm, or combinations thereof.

[0056] In one alternative embodiment, the data 20 may be used to control welding operations in response to the weld assessment conducted by the program P. For example, the current or pressure applied by the welder can be altered, the duration of the same, or shutting down the welding operation until the electrode(s) are dressed or replaced.

[0057] A record 25 of weld assessment as good or discrepant and measures of quality for each weld may be retained for later reference, e.g., to allow identification and remediation of problematic welds on an assembly and/or to customize welding procedures based upon trends discernable from the stored weld assessment data. [0058] In one embodiment, a negative or non-optimal weld assessment 20 may be responded to by triggering an additional/supplemental weld or welds to be made in the sample S to strengthen the welded assembly.

[0059] A set of pre-qualified weld patterns 22 may be referred to, as indicated by decision box 23 (which may be executed in program P) for verifying that a given weld pattern, e.g., a pattern having additional back-up welds to compensate for non-optimal welds, is of adequate overall strength to satisfy quality control criteria. If so, then the device 21 (welder and robotic positioning system) may be directed to make another weld in accordance with a supplemental pattern from the pattern set 22.

[0060] The foregoing approach may therefore initially allow using an optimally efficient weld pattern that is optimally efficient in time, energy, and equipment use in having the minimum number of welds required, assuming perfect welds, and one or more back-up weld patterns 22 with supplemental welds that compensate for welds that are deduced as being less than optimal, based upon dimensional measurements, other weld parameters, or combinations thereof. In this manner, the optimally efficient weld pattern does not need to have a safety margin (or as large a safety margin) to compensate in advance for worst-case scenarios where one or more welds is non- optimal based upon historical patterns of weld quality realization. This approach may be used to increase the distance between welds (pitch) on a welded assembly, giving an optimally efficient weld pattern.

[0061] If weld quality degrades for one or more welds specified by the optimal weld pattern, additional/supplemental welds may be applied between the welds specified by the optimally efficient weld pattern. The addition of supplemental welds may be made in reference to one or more acceptable weld patterns 22, which include supplemental welds at one or more locations.

[0062] A training set of data 26 correlating dimple dimensions to good and discrepant welds may be used to train the computer 16 programmed with artificial intelligence, such that the training data 26 may be referred to in classifying welds as good or bad, to assign a quality indicator to the welds measured, or combinations thereof.

[0063] New weld assessment data 25 may be used to expand and refine the training data 26 used in an artificially intelligent system. In this manner, the new weld assessment data 25 establishes a type of feedback loop for the weld assessment system 10. In one example, the weld assessment data 20, 25 may be utilized to ascertain a schedule of electrode dressing/replacement for a given electrode composition, the composition and thickness of the materials to be welded, the weld setting used, etc.

[0064] An aspect of the present disclosure is the recognition that the quantification of the dimensions of electrode imprints of resistance spot welded material, quantification of the erosion of an RSW electrode tip, or combinations thereof, may be used to assess weld quality, the useful life of the electrode tip remaining, and to identify discrepant and low-quality welds with an unacceptably high probability of failure. Different weld runs, e.g., as defined by the alloys to be welded, the thickness of the materials, the proximity of adjacent welds, the use of sealants, etc., may result in significantly different weld performance.

[0065] The apparatus and methods of the present disclosure can be used to monitor and adjust to welding conditions and requirements to increase confidence that good welds will be realized, and discrepant welds will be identified and removed from the finished product stream.

[0066] In one alternative, the three-dimensional shape of electrode imprints at the weld W may be analyzed by machine learning/artificial intelligence to discern correlations between topographical parameters and empirically determined weld strength to assess welds based upon the 3D shape of the weld dimple. In one example, topographical parameters can include, but are not limited to, those indicative of roughness, skewness, root mean square (RMS), peak-to-peak distance, valley depth, peak height, and combinations thereof. In one example, a subset of 100+ topographical parameters (Sa, etc. - Development of Methods for Characterization of Roughness in Three Dimensions Ken Sout et ah, May 1, 2000) or an innovative feature prescribed by an algorithm can be used with Machine Learning to detect & predict weld failures/size. The learned correlation may be binary (good or discrepant weld) or continuous (an inferred strength approximation or a weld quality measure, such as a categorization by numbers 1 to 10 with 1 being a poor weld and 10 being an excellent weld).

[0067] FIG. 2 shows a computer model of the topology of a welded sheet assembly, i.e., the shape of the top sheet near the weld W. Viewed from the top, there is a dimple or depression attributable to the welding electrode pressing down into the welded assembly. An aspect of the present disclosure is that shape attributes of the weld dimple can be discerned and utilized to assess the quality of the weld, as well as to provide insights into the state of the welding electrode and the overall process of welding.

[0068] To find and isolate electrode imprints/weld dimples at the weld W for parameter computation, a preliminary analysis is conducted on the raw dimension data to“locate” the weld W within the entire sample, i.e., to identify the image data associated with the area/volume of the sheet that is welded, as compared to the un-welded surrounding areas of the welded article, e.g., sheet metal parts, e.g., of a vehicle body.

[0069] FIGS. 3A, 3B, 3C and 4 show a plurality of computer-generated graphical models (graphs) illustrating the process of locating welds, where the measured Z dimension (depth of the dimple at that point/pixel) is expressed in a color ranging from light (Z=0) to dark (Z=too deep to measure).

[0070] Graph 122 shows multiple measurements 122M taken from a rectangular portion A2 (FIG. 3B) of surface area of a welded sheet bounded by points (320, 200) to (750, 400) and containing a weld dimple. The measurements were sequentially taken as a line laser triangulation sensor 12 was moved relative to the welded sheet, as in FIG. 1. The Z dimension is a measure of depth in the weld area W relative to a reference plane, in this instance, the average height of the unwelded portions of the composite sheet S over the surface that supports it (not shown).

[0071] Graph 124 shows the raw data with outliers removed. Outliers are typically either missing data, or large-magnitude spikes in the data, either positive or negative, that are beyond the possible range of dimensions of the sample S. Interpolation via neighboring points is used to replace outliers that were removed from the raw data.

[0072] Graph 126 shows the data depicted in Graph 124 with a best-fit surface overlaid to remove macro form.

[0073] Graph 128 shows the data depicted in Graph 126 after macro form has been removed via the best-fit surface.

[0074] Graph 130 illustrates how an algorithmic approach can be used to locate and extract individual welds within a larger dataset defined by the entire sample, i.e., by focusing on a smaller sample size with a good probability of containing at least a portion of the weld.

[0075] Graph 132 is the output of the algorithmic extraction of a single weld within a larger dataset.

[0076] Graph 134 of FIG. 4 illustrates how local form can be removed from each individual weld using a best-fit surface.

[0077] Graph 136 of FIG. 4 shows the output resulting from removing local form from an individual weld using an algorithm that finds the minimum Z-value within the (320, 200) to (750, 400) rectangle. The location of the minimum Z-value may then be used to calculate the location of the remaining points forming the weld dimple because the minimum Z-value would typically be in the center of the weld dimple and the dimple is of a shape and size determined by the known dimensions of the electrode tip that forms the dimple.

[0078] Due to the welds W being located in different places within the sample rectangles, e.g., A2 of graph 130 of FIG. 3B, it is difficult to extract the pure weld data from the non-weld data. These extraction issues may be mitigated by the flattening of the sample rectangles.

[0079] In many applications, the locations of welds are predefined, such that known weld location coordinates may be used to locate welds on a welded article and to correlate weld shape data with welds at a specific location, such that automated weld location by image analysis would not be necessary.

Data Analysis

[0080] Having identified the measurement data corresponding to the weld W, i.e., the dimple, the data can be analyzed. This may take the form of classifying the weld by an artificially intelligent computer trained by a training set of data for a plurality of known good and discrepant welds (i.e., model welds) and establishing a computer discernable correlation of the dimensional attributes of the welds to their quality (i.e., model shape data). Having established the correlation of dimple dimensions to good and discrepant weld classification, welds to be evaluated can be measured using the device 10 of FIG. 1. The measurement data can be evaluated, and the welds classified utilizing the training data. The laser line triangulation sensor (device 12) may be used to measure topography of a previously untested production weld, and the model generated from the training data would classify this previously untested weld as either good or discrepant.

[0081] Gradient normal is first calculated for the extracted weld area. Next, a circle from the original weld area is removed from the center of the weld area and used for further processing. This gradient normal and weld height data is fed through principal components analysis (PCA) and then into a cross validation classification learner alongside the Amada Myachi time-series data. Different permutations of these data sets were used to find an ideal set for maximum accuracy. Different classification learners were tested to find the best model to learn. The two most promising models were both of K-nearest neighbor ensembles using either pure raw weld data or gradient norm data with time series data.

[0082]

Raw Data

[0083]

Gradient Norm and Times Series

[0084] FIG. 5 shows a set of computer-generated graphics of dimension data of a plurality of sequential weld W samples 140, 142, 144, 146, 148, 150 illustrating topography changes occurring as a function of tip wear. These topography changes illustrate that a predictive model can be based upon measurement of topography changes.

Measurements of weld dimple dimensions indicating electrode erosion

[0085] The present disclosure recognizes that over a series of welds, welding parameters, such as electrode shape, will change giving rise to an evolving shape and surface topography of the “dimple” made by the top electrode in the upper surface of the metal welded.

[0086] The shape and surface topography of the weld dimple may be predictive of the weld nugget volume, quality, and the probability of a good weld having been formed. Changes in dimple shape and topography may not follow an obvious and/or predictable pattern. For example, the electrode degradation rate is not constant, and electrodes can even revert (“self-clean”). In such circumstances, continuous machine learning has utility.

[0087] In one embodiment, the identification of dimensions indicating electrode degradation may signal an opportunity to self-clean the electrodes by, e.g., reversing the polarity of the electrodes, adjusting the downslope or make changes to the force profile (the amount of force pressing the electrodes against the structure to be welded) and thereby avoid resorting to mechanical electrode tip dressing. In conducting an operation of this sort, one or more welds may be non-optimal during the process of attempts to self-clean the electrode tip or otherwise compensate for the non-optimal weld. To compensate for this, the welded structure may be reinforced by supplemental welds made therein, as described above. In some circumstances, the cost of producing supplemental welds may be less than that of interrupting production, e.g., to dress the electrode tip, such that overall process efficiency is increased.

[0088] As noted above, a first issue to address in measuring weld topography is data governance, i.e., isolating the data corresponding to the weld, as distinct from other areas of the welded structure. This data governance criteria includes eliminating void pixels due to method failure, e.g., attributable to the weld W dimple being too reflective to allow the surface topography to be measured or steep slope or due to local, complex form (i.e., b/g topography), where planar & paraboloid correction is insufficient, and the macro form is inadequately corrected. In one embodiment, gradient norm quantile data is correlated to weld button diameter and quality. In the collection of such measurement data, the selection of line laser triangulation sensors 12 may be compared in terms of speed and resolution. Line laser triangulation sensors 12 may be obtained from a variety of sources such as Alicona (Alicona.com) and Keyence (Keyence.com), e.g., model no. LJ-V7060. Operating parameters for each of these devices follows.

[0089] Keyence

[0090] Lateral Resolution: 20um - repeatable to 5um

[0091] Z-resolution: 0. lum - repeatable to 0.4um

[0092] Time per Weld: seconds

[0093] FOV: can scan 10 welds in a strip

[0094] In use, the following issues are encountered when using this model 1.) flatness of strip

(form) and 2.) measurement/pixel drop-outs at points due to the slopes of the measured surfaces that are too steep for optical resolution and registration. Registration (position) of the weld in the field of view changes from one weld to the next because weld sequencing is not precise, i.e., there are variations in the X and Y relative offsets of the center of each successive weld dimple in the horizontal and vertical dimensions from the previous weld. Additionally, as the electrode tip wears, it becomes increasingly difficult to define the center of the weld dimple.

[0095] Solutions for the above issues associated with Keyence line laser triangulation sensors include: i) Use of gradient slope distributions to eliminate issues of form, weld-dimple registration & drop-out (steep slope), ii) limiting handling damage, i.e., to avoid scratches or scuff marks) or iii) use of image pre-processing to filter out damage artifacts.

[0096] Alicona

[0097] Lateral Resolution: 3.5um or less

[0098] Z-resolution: 0.004um

[0099] Time per Weld: minutes

[00100] FOV: a few cm - single-weld

[00101] Issues: Flatness of strip (form), drop-outs, registration. The same solutions to these issues as described above for Keyence may be used.

Data Processing and interpretation

[00102] In accordance with one embodiment of the present disclosure, a process for classifying welds based upon electrode wear includes the follows steps:

[00103] 1) Acquiring top-surface electrode dimple data from a z-topography measurement system, e.g., Keyence or Alicona. The measured region must include an entire dimple area as well as unmarred (un-dimpled) surrounding sheet area. The measured region may be, e.g., from -700x700 pixels to 2500x2500 pixels per weld dimple (it is to be appreciated that these numbers are examples and the actual values can vary within the scope of the principles described herein);

[00104] 2) Fill in missing z values: (a) by interpolation if simple drop-outs; or (b) by max if due to steep slope;

[00105] 3) For each non-border pixel, a) compute the gradient norm, based on a 3x3 or 5x5 local pixel region, i.e., for a pixel at position q,r; the gradient norm can be based on all 9 pixels in the 3x3 square with pixel q,r at the center, or on all 25 pixels in the 5x5 square with pixel q,r at the center; b) compute cartesian (signed) gradients: dz/dx and dz/dy; and c) compute non-negative gradient norm (i.e., 3D local max slope): [(dz/dx)²+(dz/dy)²]^1/2;

[00106] 4) Display gradient norm image for each weld dimple; [00107] 5) Ensure that the surface is not marred by scratches (Visibility is enhanced in gradient norm image - this can be ascertained visually, or by using an image-analysis-based scratch-finding algorithm.);

[00108] 6) If the surface is marred by scratches, eliminate scratch data by filling in, omitting, or changing the field of view (FOV); (These image analysis methods are available in publicly available image processing software, such as in SPIP or Fiji software.); and

[00109] 7) Compute histogram and 95% quantile of gradient norm values (500K to 6MM values per weld). In this last step, each pixel is assigned a gradient norm, i.e., a maximum descent or “steepest slope”. For each weld dimple (or collection thereof), the upper tail of the histogram of gradient norms represents the steepest slopes in the image. As the electrode tip wears, there is an increasing likelihood of metal plucking from and depositing in the weld dimple, leading to an excess of steep slopes relative to the initial“clean paraboloid dimple”).

[00110] The present disclosure recognizes that welding electrode dimple degradation from use due to metal plucked and deposited can be correlated to higher slope quantiles and poorer quality weld nuggets.

[00111] FIGS. 6A, 6B, 6C, 6D and 6E show a plurality of graphical computer modeling depictions of dimension data for weld dimples in each of graphs 234, 236, 238, 240, and 242.

[00112] Graph 234 of FIG. 6A shows raw dimension data for a series of 10 welds on a sample strip, wherein the Z dimension is graphically conveyed by color spectrum as shown in the legend on the right. The digital image includes voids attributable to failure of the laser metrology method and visible as white regions in Graph 234.

[00113] Graph 236 of FIG. 6B shows the data of graph 234 with the voids filled using 3D linear interpolation.

[00114] Graph 238 of FIG. 6C shows the data of graph 236 after planar leveling using the ordinary least squares best-fit plane.

[00115] Graph 240 of FIG. 6D shows the data of graph 238 after local leveling using a maximum-dimension (599x599) levelizing sliding filter, which forced mean=0 for all such squares).

[00116] Graph 242 of FIG. 6E shows the data of graph 240 with the computed gradient norm (absolute slope) at each measured point/pixel calculated and depicted in a color shown by the legend at the right. The slope at each point is the gradient norm as calculated according to: [dz/dx)²+(dz/dy)²]^1/2. An area of erosion of the electrode that is impressed into the surface of the dimple or a deposit of electrode material on the surface of the dimple would result in a significant change in slope and a slope with a high magnitude at the edge of that dimple feature. Similarly, an area of welded substrate pickup on the electrode that is impressed into the dimple area or an area of material missing from the dimple would exhibit a slope with a high magnitude. In both instances, this would be an indication of electrode wear or pickup that can be correlated to electrode degradation and resultant poor-quality welds, with the greater the area of the dimple exhibiting these attributes, the greater the degradation and the poorer the weld quality. The reason for this phenomenon is that with steeper slopes there is less area of physical contact between the electrode tip and the metal surface, resulting in fewer and less uniform melt points, each with higher energy density. The latter condition has been shown to be correlated with poorer weld quality.

[00117] FIG. 7A shows a high-brightness/high-contrast digital image representation 344A of calculated gradient norm data for a plurality of weld dimples from a set of 10 welds.

[00118] FIGS. 7B through 7K show a plurality of digital images 344B, 344C, 344D, 344E, 344F, 344G, 344H, 3441, 344J, 344K showing ten additional sets of welds in grouped sample sets of 10 sequential welds, i.e., 1-10, 41-50, 81-90, etc. made without changing or dressing the electrodes that formed them. The sample sets depicted in graphs 344B, 344C, 344D, 344E, 344F, 344G, 344H, 3441, 344J, 344K therefore extend over a total sequence of 290 welds.

[00119] The welds in the graphs 344B through 344K have a variety of gradient norm Z ranges and are characterized as either good welds (“Weld Pass”) or discrepant welds (“Weld Fail”) based on the standard criterion for minimum acceptable weld dimensions, i.e., the thickness of the weld nugget.

[00120] The first 170 welds shown in graphs 344B, 344C, 344D, 344E and 344F are good welds. The next five sets of ten welds (from 201-290) shown in graphs 344G, 344H, 3441, 344J and 344K show the progression of electrode tip degradation with repeated use - as revealed by more prominent, steep slopes in the weld dimple.

[00121] As can be appreciated from the graphs 344G, 344H, 3441, 344J and 344K, welds formed after 200 prior welds with the same, undressed electrodes were not acceptable. Accordingly, FIGS. 7 A through 7K show the correlation between electrode degradation, steep slopes and poor welds.

[00122] FIGS. 8 A, 8B and 8C show a plurality of histograms 448B, 448C, 448D, 448E, 448F, 448G, 4481, 448 J, and 448K, the 95% quantile for the gradient norms of the ten sample weld strips shown in graphs 344B, 344C, 344D, 344E, 344F, 344G, 344H, 3441, 344J, and 344K of FIGS. 7B through 7K, respectively. Each histogram 448B, 448C, 448D, 448E, 448F, 448G, 4481, 448J, and 448K is an empirical distribution for the gradient norm (i.e., maximum-descent slope, in angular degrees) for a strip of 10 welds - identified in the histogram header by sequence number. For example, the histogram 448 J at the upper left-hand comer of FIG. 8C has a header title“2-281 FILTER RESULTS HISTOGRAM.BCR” indicating that this is the histogram for welds 281 to 291. The triangle T2 identifies Q95% = the approximate 95% quantile for the distribution. Hence, to the right of the triangle T2, 5% of the local slopes are steeper than Q95%.

[00123] FIG. 9 shows a graph of weld dimple dimension data for a plurality of welds processed to calculate the 95% quantile angle in degrees vs. the start weld # and whether the weld was good (below the horizontal line at 10.50) or bad (above the horizontal line at 10.50) demonstrating the correlation between electrode wear over a series of welds and weld quality. 95% quantile angles greater than 10.5 are strongly predictive of discrepant welds and characteristic of electrodes that have been used 200 or more times.

[00124] The present disclosure teaches apparatuses and methods to aid in reliable NDE and weld-to-weld adaption of RSW technology to increase confidence in aluminum spot welds and the rate of vehicle aluminization.

[00125] As described in U.S. application 14/967,777 published as US 2016-0167158, entitled, Resistance Welding Fastener, Apparatus and Methods for Joining Similar and Dissimilar Materials and U.S. application 15/469,161, published as PCT/US17/24093 and entitled, Resistance Welding Fastener, Apparatus and Methods for Joining Dissimilar Materials and Assessing Joints Made Thereby, both of which applications are incorporated by reference herein, a fastener may be used to attach layers of material that are not weld compatible to the fastener to a layer of material to which the fastener can be resistance welded. For example, aluminum can be attached to steel using a steel fastener that passes through an aluminum layer and welds to a steel layer. This technology is known as resistance spot riveting (RSR). As noted in these applications, the quality of the joint made by RSR may be assessed by image processing and analysis of the joint, e.g., by observing the pattern of ejecta from vents in the cap of the fastener.

[00126] An RSR joint has two sides like an RSW joint, with the positive or negative electrode being placed against the cap of the fastener and the electrode of opposite polarity placed against the sheet to which the fastener is welded. The resultant RSR joint therefore has two sides which may be measured and analyzed in accordance with the teachings of the present application, viz., the cap side and the opposite side. The image processing data developed in accordance with the disclosure of the 15/469, 161 application may optionally be analyzed in conjunction with dimension/shape data obtained from a line laser triangulation sensor 12. For example, the imaging data captured in accordance with the 15/469,161 application may be used to locate the weld joints and to provide a first level of information as to the conformance with standard ejecta patterns on the cap side of the joint that are indicative of weld quality. This information and the conclusions drawn therefrom can be confirmed and refined by use of the dimensional data captured and analyzed by the techniques disclosed in the present application, e.g., by measuring the ejecta pattern and the weld dimple on the opposite side of the joint (opposite to the cap) with a line laser triangulation sensor and analyzing such measurements in accordance with the disclosure of the present application, e.g., by finding the quantity of high slope areas on the weld dimple and correlating that to weld quality and/or subjecting ejecta measurements to classification by artificial intelligence analysis based on training data that correlates weld quality with historical ejecta measurements.

[00127] FIG. 10 shows a scanning apparatus 610 having a slide beam 612 supporting a line laser triangulation sensor 614 relative to a sample 61 I S with multiple welds W supported on a support surface SR and held by retainers Rl, R2. The sensor 614 is movable along the beam 612, e.g., by the turning of a threaded rod by a stepper motor or by the motion of a cable, chain or linear motor (not shown). As described above, the sensor 614 can be used to obtain dimension data from the welds W as it is moved relative to the sample S.

[00128] FIG. 11 shows a movable sample support 710 with a pair of support beams 712, 714 disposed at right angles to one another. Support beams 712, 714 each have a threaded rod 712R, 714R and a stepper motor 712M, 714M for moving platforms 712P and 714P, respectively.

[00129] A sample like sample S of FIG. 1 may be placed on platform 714P and then moved to a given position within the range of combined motions of platforms 712P and 714P under control of the stepper motors 712M, 714M actuating the threaded rods 712R, 714R. The stepper motors 712M, 714M may be controlled by a computer, such as computer 16 of FIG. 1. The scanning apparatus 610 of FIG. 10 and the sample support 710 may be used alone or in combination to induce relative motion between a line laser triangulation sensor and a sample to be measured, e.g., to measure a given set of welds, e.g., for developing a training data set.

[00130] FIGS. 12 and 13 show a movable sample support 810 with a base 812 upon which is mounted a first linear slide 814. The linear slide 814 has a stepper motor 816 for moving a platform 818 (in dotted lines) upon which a second linear slide 820 is mounted at right angles to the first linear slide 814, as in the movable sample support 710 of FIG. 11. The second linear slide 820 is similarly equipped with a moving platform 822 (in dotted lines) movable by stepper motor 824. A sample support in the form of a table or platform 826 can receive and retain a sample (not shown), the sample being positionable at a selected position below a scanning head 830. The range of motion of the table 826 is shown in dotted lines in FIG. 13.

[00131] In the present disclosure, the figures constitute a part of the specification and include illustrative embodiments and illustrate various objects and features thereof. The figures are not necessarily to scale, and some features may be exaggerated to show details of particular components. In addition, any measurements, specifications and the like shown in the figures are intended to be illustrative, and not restrictive. It is contemplated that elements and features of one embodiment may be beneficially incorporated in other embodiments without further recitation. Therefore, specific structural and functional details disclosed herein are not to be interpreted as limiting, but merely as a representative basis for teaching one skilled in the art to variously employ the present invention.

[00132] Detailed embodiments are disclosed herein. However, it is to be understood that the disclosed embodiments are merely illustrative and may be embodied in various forms. In addition, each of the examples given in connection with the various embodiments which are intended to be illustrative, and not restrictive.

[00133] Throughout the specification and claims, the following terms take the meanings explicitly associated herein, unless the context clearly dictates otherwise. The phrases“in one embodiment” and“in some embodiments” as used herein do not necessarily refer to the same embodiment s), though it may. Furthermore, the phrases“in another embodiment” and“in some other embodiments” as used herein do not necessarily refer to a different embodiment, although it may. Various embodiments of the invention may be readily combined, without departing from the scope or spirit of the invention.

[00134] The term“based on” is not exclusive and allows for being based on additional factors not described, unless the context clearly dictates otherwise. In addition, throughout the specification, the meaning of“a,”“an,” and“the” include plural references. The meaning of“in” includes“in” and“on.”

[00135] While a number of embodiments have been described, it is understood that these embodiments are illustrative only, and not restrictive, and that many modifications may become apparent to those of ordinary skill in the art, including that the inventive methodologies, the inventive systems, and the inventive devices described herein can be utilized in any combination with each other. Further still, the various steps may be carried out in any desired order (and any desired steps may be added, and/or any desired steps may be eliminated).

[00136] Aspects will now be described with reference to the following numbered clauses:

EXEMPLARY EMBODIMENTS

[00137] 1. A method for assessing a weld joint, comprising the steps of: (A) providing a weld joint; (B) measuring the three-dimensional shape of a surface of the weld joint to determine shape data; (C) analyzing the shape data; (D) predicting a quality of the weld joint based upon the step of analyzing; and (E) generating an output in response to weld quality.

[00138] 2. The method of Clause 1, wherein the output includes controlling at least one welding process parameter.

[00139] 3. The method of Clause 1 or 2, wherein the step of analyzing is by machine intelligence trained by training data correlating a plurality of shape data to weld quality.

[00140] 4. The method of any of Clauses 1 to 3, wherein the step of measuring includes projecting a laser line on the surface; receiving reflected light from the surface in a sensor; and moving the weld joint and the sensor relative to each other from a first position to a second position.

[00141] 5. The method of any of Clauses 1 to 4, wherein the shape data includes slope data and the step of analysis includes interpreting the slope data as indicative of electrode condition.

[00142] 6. The method of any of Clauses 1 to 5, wherein the weld joint is formed by resistance spot welding (RSW), has two side surfaces and the step of measuring is conducted on one or more of the two side surfaces of the weld joint.

[00143] 7. The method of any of Clauses 1 to 6, wherein the weld joint is formed by resistance spot riveting (RSR), has two side surfaces and the step of measuring is conducted on one or more of the two side surfaces of the weld joint. [00144] 8. The method of any of Clauses 1 to 7, wherein a plurality of weld joints is provided during the step of providing and the steps of measuring and analyzing are conducted on the plurality of weld joints in a batch.

[00145] 9. The method of any of Clauses 1 to 8, wherein the step of measuring is conducted a plurality of times for the weld joint.

[00146] 10. The method of any of Clauses 1 to 9, wherein the weld joint joins a first aluminum member to a second aluminum member and is a resistance spot weld.

[00147] 11. The method of any of Clauses 1 to 10, comprising: the steps of obtaining additional weld measuring data and analyzing the additional weld measurement data.

[00148] 12. The method of any of Clauses 1 to 11, wherein the output is indicative of a non- optimal weld and further comprising the step of making a supplemental weld.

[00149] 13. The method of Clause 12, comprising: the step of referring to a reference weld pattern to determine the location of the supplemental weld.

[00150] 14. The method of any of Clauses 1 to 13, wherein the weld joint is prescribed by an optimally efficient weld pattern having the minimum number of good quality welds for forming a structure and further comprising the step of providing a supplemental weld if the step of predicting indicates that the weld joint is discrepant.

[00151] 15. The method of any of Clauses 1 to 14, wherein the step of measuring includes identifying the shape data associated with the weld joint distinct from an adjacent un-welded area, replacing outliers in the shape data with interpolated values from neighboring points and overlaying a best-fit surface to remove macro form.

[00152] 16. The method of Clause 15, wherein the step of identifying includes finding the minimum Z-value within a sample of shape data, assigning the minimum Z-value to be a center of a weld dimple and calculating the location of the remainder of the weld dimple based upon the geometry of a welding electrode that made the weld dimple.

[00153] 17. The method of any of Clauses 1 to 16, wherein the step of analyzing includes calculating cartesian gradients: dz/dx and dz/dy; and computing non-negative gradient norm: [(dz/ dx)²+(dz/ dy)²] ^1/2.

[00154] 18. The method of Clause 17, comprising: the steps of displaying a gradient normal image for a weld dimple; inspecting for scratches on the surface of the weld; remediating data corresponding to a scratch; computing histogram and 95% quantile of gradient normal values for the weld dimple; assigning a gradient norm, for each weld dimple, a portion of the histogram of gradient normals representing the steepest slopes in the image, indicative of electrode tip degradation and poor quality welds.

[00155] 19. An apparatus for assessing weld quality of welds in a welded structure, comprising: a line laser triangulation sensor; a linear encoder; a device for moving the welded structure relative to the line laser triangulation sensor, the linear encoder measuring the movement of the welded structure relative to the line laser triangulation sensor; a computer with a three-dimensional measuring routine runnable thereon, the computer receiving as input the outputs of the line laser triangulation sensor and the linear encoder and capable of calculating three-dimensional measurements of a surface of the weld, the computer further capable of analyzing the three- dimensional measurements to calculate surface slope data indicative of weld quality.

[00156] 20. The apparatus of Clause 19, comprising: welding apparatus connected to and responsive to the computer data indicative of weld quality, the computer capable of sending control signals to the welding apparatus to remediate discrepant welds.

[00157] 21. A method for assessing a weld joint, comprising: projecting, via a projector, a laser line on a surface of a resistance spot welded (RSW) weld joint; receiving, by a sensor, light reflected from the surface; producing shape data defining a three-dimensional shape of the surface based on the light reflected from the surface; based at least in part upon the shape data, predicting a quality of the weld joint, wherein the predicting includes comparing the shape data to model shape data corresponding to a plurality of model weld joints; and generating, by the computer, an output in response to the quality of the weld joint as predicted.

[00158] 22. The method of Clause 21, comprising: measuring, using a position sensor, a location of the sensor relative to the weld joint, wherein the shape data is determined based on the light reflected from the surface and the location of the sensor relative to the weld joint.

[00159] 23. The method of one of Clauses 21 or 22, comprising: moving one of the sensor or the weld joint from a first position to a second position.

[00160] 24. The method of any one of Clauses 21-23, wherein the output comprises at least one of: display data; an auditory signal; a control signal to modify one or more of a current or a pressure applied by a welder forming the weld joint; and a control signal to stop a welding operation.

[00161] 25. The method of any one of Clauses 21-24, wherein the model shape data includes a plurality of shape data correlated to weld quality based on machine intelligence trained with training data including the plurality of shape data correlated to weld quality.

[00162] 26. The method of any one of Clauses 21-25, wherein the shape data includes roughness, skewness, root mean square (RMS), peak-to-peak distance, valley depth, peak height, and combinations thereof.

[00163] 27. The method of any one of Clauses 21-26, wherein the shape data includes slope data and the predicting is based on the slope data.

[00164] 28. The method of any one of Clauses 21-27, wherein the weld joint has two side surfaces and the surface of the weld joint is one of the two side surfaces.

[00165] 29. The method of Clause 28, comprising: projecting, via the sensor, the laser line on the other of the two side surfaces of the weld joint; receiving, by the sensor, light reflected from the other of the two side surfaces of the weld joint; determining second shape data defining a three- dimensional shape of the other of the two side surfaces of the weld joint based on the light reflected from the other of the two side surfaces of the weld joint; predicting a quality of the weld joint based upon the second shape data of the weld joint, wherein the predicting includes comparing the second shape data of the weld joint to model shape data corresponding to a plurality of model weld joints; and generating, by the computer, another output in response to the quality of the weld joint as predicted.

[00166] 30. The method of any one of Clauses 21-29, comprising: a second resistance spot welded weld joint, the method including: projecting, via the sensor, the laser line on a surface of the second weld joint; receiving, by the sensor, light reflected from the surface of the second weld joint; determining shape data defining a three-dimensional shape of the surface of the second weld joint based on the light reflected from the surface of the second weld joint; predicting a quality of the second weld joint based upon the shape data of the second weld joint, wherein the predicting includes comparing the shape data of the second weld joint to model shape data corresponding to a plurality of model weld joints; and generating, by the computer, another output in response to the quality of the second weld joint as predicted.

[00167] 31. The method of any one of Clauses 21-30, comprising: repeating the method for a plurality of weld joints in a batch process.

[00168] 32. The method of any one of Clauses 21-31, wherein the weld joint joins a first member to a second member, and at least one of the first member and the second member are made of aluminum or an aluminum alloy. [00169] 33. The method of any one of Clauses 21-32, wherein the predicting the quality of the weld includes classifying a weld as either good or discrepant.

[00170] 34. The method of Clause 33, wherein in response to predicting the weld is discrepant, the output includes modifying one or more weld process parameters.

[00171] 35. The method of Clause 33, wherein in response to predicting the weld is discrepant, the output includes performing a supplemental weld.

[00172] 36. The method of Clause 35, wherein a location of the supplemental weld is based on a reference weld pattern.

[00173] 37. The method of any one of Clauses 21-36, wherein determining the shape data includes separating a first portion of the shape data associated with the weld joint from a second portion of the shape data associated with an adjacent un-welded area and replacing outliers in the shape data with interpolated values from neighboring points.

[00174] 38. The method of Clause 37, wherein the separating includes finding a minimum Z- value within a sample of the shape data, assigning the minimum Z-value to be a center of a weld dimple, and calculating a location of a remainder of the weld dimple based upon geometry of a welding electrode that made the weld dimple.

[00175] 39. An apparatus for assessing weld quality of weld joints in a welded structure, comprising: a line laser triangulation sensor; a linear encoder; a device for moving the welded structure relative to the line laser triangulation sensor, the linear encoder configured to measure movement of the welded structure relative to the line laser triangulation sensor; a computer with a three-dimensional measuring routine runnable thereon, the computer configured to receive as input, outputs of the line laser triangulation sensor and the linear encoder, and to calculate three- dimensional measurements of a surface of the weld joints, the computer further configured to predict a weld quality based on the three-dimensional measurements relative to a plurality of model weld joints, and to output an indicator of the weld quality of the weld joints.

[00176] Clause 40. The apparatus of Clause 39, comprising: a welding apparatus communicatively coupled to the computer, wherein in response to the output of the indicator being a discrepant weld, the computer is configured to modify one or more control parameters of the welding apparatus.

[00177] It is to be appreciated that any one of Clauses 1-18 can be combined with any one of Clauses 19-20, 21-38, and 39-40. Any one of Clauses 19-20 can be combined with any one of Clauses 21-38 and 39-40. Any one of clauses 21-38 can be combined with any one of Clauses 39- 40.

Claims

CLAIMS What is claimed is:

1. A method for assessing a weld joint, comprising:

projecting, via a projector, a laser line on a surface of a resistance spot welded (RSW) weld joint;

receiving, by a sensor, light reflected from the surface;

producing shape data defining a three-dimensional shape of the surface based on the light reflected from the surface;

based at least in part upon the shape data, predicting a quality of the weld joint, wherein the predicting includes comparing the shape data to model shape data corresponding to a plurality of model weld joints; and

generating, by the computer, an output in response to the quality of the weld joint as predicted.

2. The method of claim 1, comprising:

measuring, using a position sensor, a location of the sensor relative to the weld joint, wherein the shape data is determined based on the light reflected from the surface and the location of the sensor relative to the weld joint.

3. The method of claim 2, comprising:

moving one of the sensor or the weld joint from a first position to a second position.

4. The method of claim 1, wherein the output comprises at least one of:

display data; an auditory signal; a control signal to modify one or more of a current or a pressure applied by a welder forming the weld joint; and a control signal to stop a welding operation.

5. The method of claim 1, wherein the model shape data includes a plurality of shape data correlated to weld quality based on machine intelligence trained with training data including the plurality of shape data correlated to weld quality.

6. The method of claim 1, wherein the shape data includes roughness, skewness, root mean square (RMS), peak-to-peak distance, valley depth, peak height, and combinations thereof.

7. The method of claim 1, wherein the shape data includes slope data and the predicting is based on the slope data.

8. The method of claim 1, wherein the weld joint has two side surfaces and the surface of the weld joint is one of the two side surfaces.

9. The method of claim 8, comprising:

projecting, via the sensor, the laser line on the other of the two side surfaces of the weld joint;

receiving, by the sensor, light reflected from the other of the two side surfaces of the weld joint;

determining second shape data defining a three-dimensional shape of the other of the two side surfaces of the weld joint based on the light reflected from the other of the two side surfaces of the weld joint;

predicting a quality of the weld joint based upon the second shape data of the weld joint, wherein the predicting includes comparing the second shape data of the weld joint to model shape data corresponding to a plurality of model weld joints; and

generating, by the computer, another output in response to the quality of the weld joint as predicted.

10. The method of claim 1, comprising:

a second resistance spot welded weld joint, the method including:

projecting, via the sensor, the laser line on a surface of the second weld joint;

receiving, by the sensor, light reflected from the surface of the second weld joint;

determining shape data defining a three-dimensional shape of the surface of the second weld joint based on the light reflected from the surface of the second weld joint;

predicting a quality of the second weld joint based upon the shape data of the second weld joint, wherein the predicting includes comparing the shape data of the second weld joint to model shape data corresponding to a plurality of model weld joints; and

generating, by the computer, another output in response to the quality of the second weld joint as predicted.

11. The method of claim 1, comprising:

repeating the method for a plurality of weld joints in a batch process.

12. The method of claim 1, wherein the weld joint joins a first member to a second member, and at least one of the first member and the second member are made of aluminum or an aluminum alloy.

13. The method of claim 1, wherein the predicting the quality of the weld includes classifying a weld as either good or discrepant.

14. The method of claim 13, wherein in response to predicting the weld is discrepant, the output includes a control signal to modify one or more of a current or a pressure applied by a welder forming the weld joint.

15. The method of claim 13, wherein in response to predicting the weld is discrepant, the output includes a control signal to perform a supplemental weld.

16. The method of claim 15, wherein a location of the supplemental weld is based on a reference weld pattern.

17. The method of claim 1, wherein determining the shape data includes separating a first portion of the shape data associated with the weld joint from a second portion of the shape data associated with an adjacent un-welded area and replacing outliers in the shape data with interpolated values from neighboring points.

18. The method of claim 17, wherein the separating includes finding a minimum Z-value within a sample of the shape data, assigning the minimum Z-value to be a center of a weld dimple, and calculating a location of a remainder of the weld dimple based upon geometry of a welding electrode that made the weld dimple.

19. An apparatus for assessing weld quality of weld joints in a welded structure, comprising:

a line laser triangulation sensor;

a linear encoder;

a device for moving the welded structure relative to the line laser triangulation sensor, the linear encoder configured to measure movement of the welded structure relative to the line laser triangulation sensor;

a computer with a three-dimensional measuring routine runnable thereon, the computer configured to receive as input, outputs of the line laser triangulation sensor and the linear encoder, and to calculate three-dimensional measurements of a surface of the weld joints, the computer further configured to predict a weld quality based on the three- dimensional measurements relative to a plurality of model weld joints, and to output an indicator of the weld quality of the weld joints.

20. The apparatus of claim 19, comprising:

a welding apparatus communicatively coupled to the computer, wherein in response to the output of the indicator being a discrepant weld, the computer is configured to modify one or more control parameters of the welding apparatus.