Marek Fidali

Methods of image processing in vision system for assessing welded joints quality

The paper deals with methods of image processing applied in a vision system for assessing welded joints quality. The vision system consists of two CCD and one infrared camera. The base of elaborated approaches to joint assesment is a set of images taken in infrared and visible light during welding process. Images taken in visible light are sources of information about outer conditions of the joint, while thermograms let us to obtain information concerning the joint interior. In the paper results of the application of processing of these two types of images are presented.

Welding process evaluation on the basis of video and thermal images

The progress of the welding process has a decisive impact on the quality of weld joints. Obtaining welded joints of high quality, both in terms of their mechanical and aesthetic properties, requires selection of appropriate welding process parameters and, most importantly, maintaining their stability during joint formation. While the choice of welding parameters is not a problem for an experienced welder, ensuring stability of the process is not an easy task. The stability of the welding process is affected by many interfering, often random factors, such as disruption of stability of filler feed, dirt and inaccuracy of edge preparation of the joined materials, deformation of the joined elements due to thermal expansion of the material. These factors cause welding unconformities, such as excessive melting, lack of penetration, cracks, burns, too large risers or depressions in the face of the weld, presence of bubbles, and other factors. Ensuring that high quality of welded joints is possible by controlling the welding parameters. The method of controlling welding parameters can be based on the measurement of welding parameters (e.g. current, arc voltage, and shielding gas flow). Another way to control the welding process and the welded joints is the use of visual inspection of the welding process, based on recording images of the formed welds in the visible spectrum, as well as in the nearand far-infrared spectrum, and their subsequent analysis and diagnosis. This approach is mentioned in the literature as one of the most promising ways of assessing the welding process and the welded joints. The most important aspect of using visual methods is that their use in volume production enables rapid identification of disturbances in the welding process parameters, errors in positioning of the joined elements and limiting the number of unconformities that appear throughout the series of manufactured components. Visual welding process control clearly provides increased joint quality and reduced production losses.

The application of evolutionary algorithms in the search of relevant statistical features of infrared images

A thermographic image can be a source of diagnostic information. This information can be obtained using a variety of image analysis methods. Unfortunately, informational noise resulting from the large number of features can cause problems with the efficient assessment of object state. There are, however, methods which allow one to search for relevant features useful in diagnostics. In this paper an application of evolutionary algorithms (EAs) is presented for selection of optimal (from diagnostics point of view) a set of statistical features of infrared images. The infrared images recorded during the active diagnostic experiment are analysed using a variety of statistical texture analysis methods. As the result of the analysis, a set of 259 diagnostic features for each of the five regions of interest is obtained for the 840 recorded images. Two cases of the evolutionary computation are applied in order to search for relevant features. In the first case, the algorithms search for an assumed number of features. In the second case, a number of features are selected automatically by the algorithms. Three types of computational strategies are used. The first one uses a single EA in opposition to the second and third where concepts of weak-strong and external-internal EAs are used. The concepts of weak-strong and external-internal are original ideas of the author. The assessment of feature relevance is performed based upon the results of classification. The classification is performed by a neural network classifier in all evolutionary computation test cases. Finally, the results are compared to those using the classical method of feature selection. The comparison shows that the EA made it possible to obtain higher classification performance by use of a smaller subset of selected relevant features.

Application of fused visual and thermal images in monitoring and evaluation of welding processes

This article presents the application results of different methods of merging images recorded using vision cameras and thermal imaging cameras. The images were recorded during the evaluation of a welding process. The object of observation was the welding arc and the welded joint during its cooling phase. Images recorded during GMA welding on two different welding stations have been taken into account. The observations have been conducted using cameras equipped with different optics systems. This article presents an analysis of the influence of different parameters of the optics systems on the resulting fused images. The analysis also considers the pre-processing method to ensure correct matching of images. The study shows that the fused image contains aggregated information useful in online monitoring and quality evaluation of the welding process.

Concept of Database Architecture Dedicated to Data Fusion Based Condition Monitoring Systems

In the paper a concept of object-relational database dedicated to monitoring and diagnostics systems is presented. Important feature of proposed solution is the treatment of database as an integral sub-system dedicated to collect, manage and share data required for configuration and correct operation of the diagnostic system. It was assumed that diagnostic system has a modular structure, what determined the database architecture. Each module of the monitoring system and all operations performed in the system are reflected in the database by instances of designed objects connected with relations (references). In order to limit amount of stored data the database will collect only primary data like e.g. raw signals, configuration constants and meta data (rules of data management and parameters of data processing implemented in the diagnostic system). The concept of proposed database was implemented in limited form in commercial database system Oracle 12c using object-oriented features introduced in the SQL:1999 dialect.

Detection of Welding Process Instabilities Using Acoustic Signals

This work deals with problems related to the monitoring of welding process. This is an important issue, especially from the point of view of operation of automated welding systems. One of the sources of diagnostic information about welding process is sound. The sound generated during the welding process includes acoustic audible and inaudible parts. In this paper, the author focused on analysis of a ultrasound part of acoustic energy emitted during welding. Acoustic signals were collected during a few series of experiments conducted in laboratory conditions. The signals were gathered using a condenser microphone with measurement range 20–100 kHz. The acquired signals were high-pass filtered and processed using statistical measures such as RMS, IQR (Interquartile range) and Kurtosis, in order to detect welding instabilities. The best detection results obtained for signals in band 40–60 kHz was obtained with use of IQR parameter. These findings confirm the usefulness of ultrasound signals for detection of welding faults.

Improvement of Bonded Joint Defects Visibility by Use of Selected Infrared Image Processing Methods

This paper deals with the problems of adhesively bonded joints quality assessment. Author’s research are focused on development of the specialized method based on active thermography dedicated to detection of flaws in adhesively bonded structures for automotive industry purposes. During the research, a series of active thermography experiments on samples with different defects were performed. The infrared images acquired during the experiments were the basis for development of an IR image processing algorithms allowing to obtain the best defects detectability. Image processing algorithm requires conduction of the following operations: a pre-processing using noise filtration and contrast enhancement operations in order to improve image quality; transformation of spatiotemporal structure of images for further improvement of defects visibility; an image segmentation, image fusion, and pattern recognition for identification and classification of the defects. The paper presents the results of the first step of an infrared image processing algorithm. The algorithm applies multiple methods such as TSR (Thermographic Signal Reconstruction), PCT (Principal Component Thermography) or DTT (Dynamic Thermal Tomography). Obtained results show that algorithm combined TSR and pulse phase thermography is very promising for effective revealing of defects of adhesively bonded joints.

Image Processing Method for the Improvement of Visibility of Adhesive Path Defects

The increase of interest in adhesive joints is strongly connected with simple joint creation, high joint strength and its flexibility. Due to this, assurance of high-quality joints in many responsible applications is now a vital task. Because of the automatization of joint creation, autonomous vision systems are sometimes used for the control of adhesive path quality. There is a number of environmental obstacles that limit the applicability of simple off-the-shelf systems. The aim of the research is to develop an image processing method which could be applied for the enhancement of visibility of adhesive path features for further application in an adhesive path quality control algorithm implemented in a robot vision system. The solution of the posed problem requires not only identification of adhesive path edges, but also detection of bonded part edges and suppression of different light conditions influences. To achieve appropriate image quality, a series of image processing operations like sharpening filtration, edge detection, background detection and removal, binarization and morphological operations were applied. Obtained results prove that the presented approach of image processing and analysis can be used in a robotized adhesive path quality control system.

Application of Active Thermography for Detection of Soft-Soldered Lap Joints Inner Defects

Soft soldiering is a joining technique commonly used in many areas of industry. It is especially useful for joining components susceptible to damage caused by high temperature. One of the vital issues associated with this type of joining is the possibility of the formation of hidden defects, like: cracks, lack of connection, gas voids etc. In the paper, a study of active thermography application for detection of common defects in soft-soldered lap joints is presented. Lap joints of thin copper, aluminum and high-alloy steel palates were made using solder paste in a soldering oven. Internal defects were intentionally introduced into the joints. To reveal defects, an active lock-in thermography method was applied. Various heating parameters, IR images acquisition, and processing methods were tested. In order to compare the obtained defect detection results with a benchmark, tested samples were radiographically analyzed. Results prove that active thermography methods can be applied for the detection of defects in soft-soldered lap joints. Based on the research performed, the size of defects that can be detected was estimated.

Fault Detection of Railway Point Machine Using Diagnostic Models

Railway point devices are critical elements of railway infrastructure. Point device failure can significantly affect railway operations, with potentially disastrous consequences. Therefore, early detection of anomalies is critical for monitoring and managing the condition of rail infrastructure. The article presents a concept of a simple fault detection system dedicated to health monitoring of point machines in a remote and automatic way. A fundamental element of the system is a fault detection module, which operates on the basis of residual signal calculated on the basis of output of object models. Autoregressive (ARX) and neural (ANN) models were identified using real data collected during 1 year at the turnout localized in Poland. Obtained results show that simple analytical ARX model is a sufficient solution for efficient fault detection of the investigated point machine and can be easily applied in the proposed condition monitoring system.