Reconfigurable Measuring System for Quality Control of Cross-Wire Welding Group of Products

Abstract

Quality control of welded joint is an indispensable part of the welding production process. As part of spot resistance welding group, cross-wire welding process showed great application for welding of products for everyday usage. The non-contact quality control checking is fit for purpose due to specific characteristics of welded products that consist of two cross welded wires or a combination of wires and strips. This work proposes a new method for detecting and measuring of required dimensional parameters, but also founds its applicability for other products if required. A crucial parameter of this research is the height of welded joint, which is necessary for calculating the penetration of the wire into the wire. The proposed measuring method with a reconfigurable measuring system is explained in this paper. The main component of this system is using a machine vision system, which has become an indispensable part of industrial metrology and is considered one of the industry 4.0 concepts. The calibration process for such systems could be very complicated. This work shows an elaborated calibration procedure for this kind of measuring system with referenced standards made for this purpose. Measurement results are compared with ones obtained by conventional method. The focus of vision system is a substantial part as it dictates the quality of the system. This research is done within the project in collaboration with the industrial sector and all samples are from real processes. The results of measured penetration on one product group are showing the applicability of a reconfigurable measuring system in the welding sector, and demonstrate that measurement of welding penetration based on machine vision is feasible and can ensure accuracy.

1. Introduction

Cross-wire welding (CWW) is one of the spot resistance welding (SRW) [1] methods and refers to the joining of two wires of different diameters that are perpendicular to each other [2]. The subject of research within the project “European regional development fund project” is a welded joint between two wires of a certain diameter (Figure 1). The amount of penetration of one wire into another one depends on the set welding parameters, and it is one of the quality indicators of the resulting joint. An indispensable part of the SRW production process is the quality control of obtained joint [3]. Due to the specific shape of the controlled product, this paper describes one of the non-contact measurement control procedures of wire-to-wire penetration at the joint, which is a base for further analysis of quality control of the obtained joint. Once the first phase of control is satisfied, and it is confirmed that a certain percentage of penetration has been achieved, controlled weld joint proceeds to further control procedures to check appearance of weld, strength and at last corrosion resistance as final advantage of joined materials. Over the last decades, there have been improvement in joining materials technology and anticorrosion behaviour [4,5,6,7]. The research conducted and documented in this paper provides a scientific contribution to the development and application of reconfigurable measuring systems (RMS) in testing of specifically shaped welded joint characteristics in industrial environment.

Reconfigurable systems were developed by scientists at the University of Michigan [8], and up to today, more different prototypes of such devices have been developed [9,10,11,12]. Such systems are designed in accordance with certain principles and the more they are applied, the more reconfigurable a particular system is [13]. Reconfigurability concept is gaining practical significance and has been strengthened by strong and considerable economic benefits [14]. According to the purpose, there are several types of reconfigurable machines [15], one of which is a product inspection machine such as [16,17,18]. The emphasis of such control systems is on sensors and non-contact measurement of certain characteristics. The purpose of RMS is to enable sufficiently fast and complete product quality control in accordance with the required characteristics and capabilities.

According to the principles of reconfigurability, the definition of a reconfigurable measuring system would define it as a system designed in such a way that it can be quickly adjusted to control characteristics of a particular product group or within different groups. Based on the classification of features for selected product group, the preliminary model of the measuring system was analysed and the optimal one was selected. The prototype of the measuring system equipped with the necessary measuring sensors was constructed and built based on selected preliminary model of measuring system. It is made for one or more groups of products with specific characteristics. This approach is also environmentally friendly as limited use of resources (minimum amount) is one of the key guidelines. According to the concept of reconfigurability, it is necessary to meet certain requirements in order for the system to be called such. The architecture of the RMS construction must be modular, allowing modules (de)installation to (from) the system. In order to adapt the RMS to a new product from the same product group, the locations of individual modules can be reconfigured. The main advantage of such a system is to promptly have feedback on the quality of each product during production process, and with the use of available technology within the smart factory concept, it is possible to immediately make adjustments to the production line if deemed necessary.

Two perpendicular wires welded by the CWW process represent one group of products that will be controlled on a specially designed RMS that can be adapted also for other product groups when needed. The group of products studied in this paper consists of two circular cross-section wires with a certain diameter. Based on characteristics of welded pieces other groups of products shall be also defined, e.g., steel strips and wires product group or steel strips and strips group, both with variations in items dimensions. When it is necessary to test a completely different product, new product groups will be established based on product characteristics.

The factory where items for analysis (Figure 1) are produced is equipped with a robotic cell for the SRW/CWW process (Figure 2), which enables and facilitates the monitoring and change of welding parameters.

2. Materials and Methods

Steel grade of wire material is S235 and this kind of material is widely used in everyday life and can be found in the form of grids, gratings, fencing systems, baskets, etc.

It is desirable that quality control systems are accurate and versatile enough to allow measurement of produced parts and to prove more information related to performed measurements at the same time. The designed RMS uses machine vision to obtain the necessary measurements with the help of photogrammetry [19,20]. The camera has found an important place in product quality control as part of industrial photogrammetry, industrial metrology and the industry 4.0 concept in general.

Today’s industrial production without machine vision is unthinkable, and can be found in all types of industry, production and within product quality control, subsequently, it is increasingly represented in the application of welding and welding control.

Simple optical mechanisms allow a wide range of optical methods that allow the measurement of various objects regardless of size. Optical measurement systems can be divided according to [21].

The price of components and software is a limiting factor in the field of vision control. Many manufacturing systems do not have the resources to implement quality vision systems as part of full product control, but can only afford low-budget devices [22].

Processing of 2D obtained photos is mostly used for detection of objects and their parts, measuring dimensions and reading of symbols and text, etc.

According to the requirements for quality control and preliminary research, the method of control of individual dimensions of a given product was chosen, by non-contact measurement. With the development of non-contact measurement, much more reconfigurability has been enabled by moving optical systems to areas of interest.

Machine vision technology includes image collection, image, and data processing while the machine vision system consists of hardware and software [23]. The main components of hardware are the computer as an essential part, cameras with mounted corresponding lenses, light, stand, and cables. For this purpose, the sensor used for measurement is selected according to the measured subject dimensions and measurement requirements. As the measurement subject is of small dimensions and complex shape, the industrial camera was chosen together with the telecentric lens. The telecentric lens has proven to be the best choice for precise measurements of small dimensions and it has a constant magnification regardless of the object’s distance from the lens. For this reason, it is often used in metrology and vision systems that require high measurement accuracy. Telecentric lenses have a high resolution, low distortion, and stable magnification [24]. For the purpose of illuminating subjects that are dark and reducing the unfavourable reflection (because of the material of wire—steel S235) that occurs, LED lighting is most often used as a light source. It shall be noted that it is important to pay attention to the appearance of shadows and the lifespan of a lighting system. LED lighting with direct front lighting provides a high contrast between the subject and the background [25]. To avoid the appearance of shadows, the ring shape of LED lighting was used in combination with LED compact flat lighting.

Image processing software is required. Most cameras come with applications used to display images and configure the camera. In this research, Basler cameras with associated software for image acquisition were used. Special applications and specific image processing tasks require specialized software that can be purchased or developed. Development of specialized software can be a complex operation, so ”off the shelf” software is used more often. Solutions are already available for many standard imaging tasks, such as Halcon or LabView.

The used RMS consists of standard modules that are easily available, such as the listed optical devices, computers, light sources, software, accessories that perform certain operations specified by the program, etc. Used modules and their main characteristics are shown in

Table 1. Main components of reconfigurable measuring system with main specifications.

Hardware Components	Preferences
Computer	USB 3.1 Gen
	Intel Core i5 9th Gen Core i5-9300HQ
	Nvidia GeForce GTX 1050 RAM 8GB, SSD 512 GB
Sensor	12.4 mm × 9.8 mm
	acA2500-60uc-Basler ace CMOS 2590 px × 2048 px
Lens	TC2MHR016-C
	max. distortion 0.04%
	C-mount
	F16
RMS Stand	Bosch standard modules
Light	Ring light LED
	Compact light LED

. Individual modules can be (de)installed to (from) at the required position depending on the product planned to be measured as shown in Figure 3. Positions of a particular module depend on the characteristics of the product group that needs to be measured and inspected at this point. This configuration of RMS enables checking of multiple features at the same time, such as the dimension of each wire, dimension of wire to wire penetration, and visual inspection for further analyses. The main advantage of the presented RMS is that in order to adjust it to the other group product, the locations of the individual modules can be reconfigured in such a way that they will be placed in a position that will allow inspection of required characteristics for a new product, considering special calibration procedure.

2.1. Calibration Procedure

To assure that measuring equipment is reliable, it is necessary to do calibration procedure in regular intervals or once the need arises (due to some change in the system, etc.). In order to make calibration as simple and fast as possible, it is necessary to use a reference object, such as glass objects with incised markings, metal parts machined with specified dimensions, or metal sheets with precisely drilled holes [26]. To complete the project task, it is necessary to determine the dimension of the wire in the wire penetration. The first step in determining the exact dimension obtained by the vision system is the RMS calibration procedure [27]. This includes a process that sets a ratio (the conversion factor) between image pixel and real distance, which is required in order to scale the recorded image to metric units. Appropriate referenced object with familiar dimensions is used to set conversion ratio in order to relate size image pixels with recorded. The image is made of pixels. Each pixel contains a set of information and has an assigned location in the 2 axis coordinate system. Digital images have a certain number of pixels and through the process of image digitization, there is a loss of a certain amount of information, as the real image must be summarized to a certain number of pixels. After selecting the appropriate object, it is necessary to know the procedure of image recording and measurement [26].

Calibration of the RMS system must be performed in the same conditions in which further measurements will be performed, and this includes the same camera settings, the same lighting, the same equipment, and the same environmental conditions.

It is best to calibrate equipment using an object that is similar in size and shape to the object for which such a system will be used. Due to their price, reference objects are more often one-dimensional plates with inscribed markings, but they can also be precisely produced 3D objects with regular and rectangular planes [28]. In this study, calibration is performed by using a reference object whose geometry in 3D space is known with very good precision. There are also other techniques like in [29]. The reference object according to its surface characteristics must not affect the accuracy of the results. The error of the reference object itself must be less than the possible error of the system so as not to affect the accuracy of the results. In this research reference object has been designed for this product group and for this RMS (Figure 4). Advantage of this reference object is that it can be used for different product groups and different modules if there is a requirement for change. The goal is to have one reference object (or set of objects) that will be used to calibrate RMS and to avoid new recalibration of RMS due to changes in the system.

For testing of the measuring system accuracy and confirming of calibration process a 6mm measuring standard shown in Figure 5. was used, respecting the same conditions.

During the calibration process, the most important thing is to accurately determine the actual edge of the object. The edges of objects can be defined as abrupt transitions in the grey scale. Many authors deal with the detection of the edge of objects in images and use different filters [28,29,30]. It is also necessary to ensure that the equipment and the reference object stand stably and firmly. The circular bubble level with high sensibility was used for this purpose.

If the camera focus is not set properly the contrast in the image is not optimal and calibration and measurement errors are more likely to occur. Focus allows the sharpening of the image. Best focus can only be achieved if a camera is installed at a certain distance from the subject. The dependence of the object distance and the focal plane is shown by the authors [31]. The lens must be precisely focused on the plane of measurement. In an industrial environment, the distance between the subject and the camera is generally fixed, so manual focus is used instead of automatic focus. Small changes in the distance must not affect the change in the mm/px ratio, and if this happens, a telecentric lens must be used as it is more resistant to changes in focus settings. The human eye can easily estimate if an image is well focused but cannot estimate if it is optimal for image processing [32].

Before setting up the object to be controlled, the equipment must be set up correctly, depending on the conditions that must be met. Images must be of the best possible quality to obtain the necessary information from the image. The camera resolution must be acceptable.

Component installation procedure:

• Ensure that the colour of the background is in a contrast with the object
• Ensure that the camera is perpendicular to the surface of the recorded subject
• Ensure satisfactory lighting: shades of the object are not projected on the surface, no reflection, the object is properly illuminated to show necessary characteristics planned to be controlled, and there is sufficient contrast between the surface on which the object is located and the edge of the object once illuminated
• If the image is taken at an angle, ensure that the used software can compensate for the error that occurs
• Ensure proper system calibration before starting the measurement.

2.2. Measurement Experiment

The wires that need to be welded are not of exactly the same dimensions, their diameters vary from 3.97 up to 4 mm. In this case, it is difficult to determine achieved penetration without measuring the dimensions of each wire separately, and then their overlap.

The penetration is not measured, although it can be easily calculated if the dimensions of each wire and the height of welded joint is known, as it is shown in Figure 6. It is not possible to physically measure dimension “A” before welding, but it can be calculated as the sum of two wire diameters (wire 1 and 2). After the welding is performed, dimension “B” can be easily checked with the proposed method.

A special module used to hold and support the wire during measuring penetration with machine vision (Figure 3b) was made for this RMS using additive manufacturing. The module is installed in front of the lens and the measuring object is inserted.

The camera, lighting, and device settings are shown in Figure 3a. Attention shall be also pointed to precision in modules installation as this also takes part in the quality of measuring process (e.g., the wire holding module has to be changed depending on the wire properties, position of the module that holds the camera with the lens is adjusted depending on the wire size and focus change). Due to the specifics of the penetration measurement and the shape of the group of objects, a telecentric lens and LED lighting was selected for the previously mentioned camera. For installation of all measuring system parts, levelling tool with sensitivity up to 2 mm/m indicates if the surface is horizontal.

The object of measurement shall be recorded in such a way that the object is in the focus of the telecentric lens so that the edges visible on the image are sharp as much as possible and clearly visible.

In addition to data related to penetration and each wire’s properties, the image recorded respecting described procedure provides also a set of data used for storing of weld visual appearance that will be used in further CWW quality control tests.

2.3. Calibration Procedure of the RMS

In order to properly relate dimensions shown on recorded images to corresponding ones, in reality, it is necessary to properly relate pixel with a unit of measure. Certain number of pixels will be associated with a corresponding number of measure units (e.g., with mm), and performing this action a conversion ratio factor will be determined. This procedure is known as the calibration of the measuring system.

The first step in calibration is to take an image of a reference object with known dimensions while respecting certain conditions listed in the previous section.

The image of the reference object must be taken under the same environmental conditions, with the same equipment at the site where future measurement activities will be performed. Image quality must be satisfactory and the edges of the measured object shall be clearly visible. If the first condition is not met, it is required to confirm that all components are installed in the designed way (settings that shall be rechecked are the ones of lighting, lens, camera, computer, and the software itself). If the image still does not satisfy quality requirements, it shall be checked what are the possibilities of the image analysis software and which image processing techniques are used [33].

The most common errors that occur are the ones caused by improper placement of the object of measurement, if the position of the camera is not perpendicular to the surface of the object (the cause may be the background, the object itself, the camera stand or the camera itself). Nonlinear distortions can occur due to aberrations of the camera lens [33].

The reference object made for this purpose and used for calibration of the RMS is shown in Figure 4. The dimensions of the used reference object are 8 × 8 × 10 mm. Once the system was calibrated, it was validated with a measuring standard by controlling 6mm dimension as shown in Figure 5. The standard was photographed under the same conditions in which other measurements were carried out. The 0.02 mm/px ratio was determined and further measurements were performed. When setting up the calibration software, the instructions shall be followed depending on the chosen software. In this study, LabView NI Vision software is used for image processing, which uses its developed procedure for image processing and calibration, as shown in [34].

3. Results

The diagram in Figure 7 shows results obtained by measuring dimension “B” in 81 different samples and comparing results for each sample obtained with two different measurement methods. The first method was conventional and measurement was performed manually with a micrometer, while RMS was the second method. If RMS measurement results are compared with ones obtained with a micrometer, it can be noted that RMS provides acceptable results. The average value of the difference is 0.02 mm between these two methods and the median is 0.019 mm.

Testing measurements were performed with different combinations of focus settings. The following graph in Figure 8 shows a comparison of results obtained with RMS for different settings of focus position and the ones obtained with conventional methods. The first set of measurements were carried out by positioning the focus in the middle of two wires, while the second set is done by positioning the focus on the wire which is closer to the lens. Results are shown in Figure 8, and a small deviation in results is present depending on the focus setting, although in further measurement focus can be taken in the middle of the object and results will be acceptable. When comparing results obtained for a different setting of focus it can be noticed that the results are acceptable for a case where there is present only a small change of focus and in that case, it is possible to measure dimension in the middle of the wire even if the focus is not there. This finding is important as it proves that RMS allows users to measure the diameter of an individual wire while also measuring penetration by calculating it.

Figure 9 shows an example of measuring the dimension of two cross-welded wires with RMS focus on the middle of the wires. The result appears in the left bottom of the screen.

All other results are made for the cases where the focus is on the middle of the crossing. Figure 10 shows a comparison of measurement results obtained by RMS and conventional micrometer for 81 objects. Mean value of RMS measurement results is 7.019 mm while for micrometer measuring that value is 7.003 mm. This shows that the error of RMS is less than one pixel size. The standard deviation of results obtained by RMS is 0.054 mm, while for the case of micrometer measuring it is 0.046 mm. This shows that measuring results using RMS can be considered as valid as results are within set criteria of acceptability.

Figure 11 represents individual control charts for RMS and for micrometer measured values. In Figure 11a it shows that measured values with RMS are all between upper and lower control level and the range of levels are 0.29 mm. In micrometer, measured values control chart shows that some measured values are below lower control level, but the range of levels is 0.22mm. Also, the box plot in Figure 12 can be a good representation of the results, that RMS values are in a greater range but they are more stable.

4. Discussion

Penetration of two cross-welded wires should be measured for an analysed group of products which makes special requirement of quality control and makes this group-specific. Apart from those requirements there is also a set of visual requirements that each weld shall satisfy. To satisfy the required set of requirements, RMS was designed especially for this purpose. The main advantage of a designed RMS is that it can be reconfigured if another group of products with a different set of requirements needs to be controlled. In this paper, RMS is presented and validated with an experiment performed for one product group. The first step of measuring with machine vision is to calibrate the system. This system uses a telecentric lens with some specifics, and camera focus settings can be high lightened. There is also statistical data collected in the process, analysed, and presented in diagrams that show validation and comparison of RMS with the conventional measuring system. RMS system satisfied all set requirements and further work will follow in order to analyse welded product images to classify weld category.

The main advantage is that the group of products is calibrated using a new reference object made only for this purpose, and due to the specificity of the telecentric lens, new calibration of RMS is not required even if the group of products changes, although this will require only to change the camera focus in order to get a more accurate measurement result. Also, these reference objects are adequate for other lens and sensor if there is a requirement for equipment change.

5. Conclusions

A reconfigurable measuring system was designed to perform the initial part of the quality control of welds obtained by the CWW process. The first step in quality control is to accurately determine the dimension of wire-to-wire penetration at the welded joint. The second step is to use obtained images for visual inspection of the weld. Weld penetration and the visual look of the weld directly affect weld strength.

Measurement of penetration of small dimensions circular cross-section objects is very complex. With designed RMS, penetration data is accessible on the computer, and at the same time, the settings of the robot cell used for welding can be instructed to intervene in order to change welding parameters if required.

The experimental part of the research includes the calibration of the measuring system and performance of measuring tests, where results obtained with the RMS method were compared to the ones obtained with the conventional method (use of micrometer). The average difference of measured values obtained by the two aforementioned methods is 0.02mm, which presents less than 1% of the measured dimension. Testing was performed with different focus settings and achieved results are acceptable for small changes of focus, which allows measuring of target dimensions in a single image.

This method directly contributes to improvement of welded joint quality obtained by the CWW method.