Girolamo Fornarelli

Application of neural networks in optical inspection and classification of solder joints in surface mount technology

The defect detection on manufactures is extremely important in the optimization of industrial processes; particularly, the visual inspection plays a fundamental role. The visual inspection is often carried out by a human expert. However, new technology features have made this inspection unreliable. For this reason, many researchers have been engaged to develop automatic analysis processes of manufactures and automatic optical inspections in the industrial production of printed circuit boards. Among the defects that could arise in this industrial process, those of the solder joints are very important, because they can lead to an incorrect functioning of the board; moreover, the amount of the solder paste can give some information on the quality of the industrial process. In this paper, a neural network-based automatic optical inspection system for the diagnosis of solder joint defects on printed circuit boards assembled in surface mounting technology is presented. The diagnosis is handled as a pattern recognition problem with a neural network approach. Five types of solder joints have been classified in respect to the amount of solder paste in order to perform the diagnosis with a high recognition rate and a detailed classification able to give information on the quality of the manufacturing process. The images of the boards under test are acquired and then preprocessed to extract the region of interest for the diagnosis. Three types of feature vectors are evaluated from each region of interest, which are the images of the solder joints under test, by exploiting the properties of the wavelet transform and the geometrical characteristics of the preprocessed images. The performances of three different classifiers which are a multilayer perceptron, a linear vector quantization, and a K-nearest neighbor classifier are compared. The n-fold cross-validation has been exploited to select the best architecture for the neural classifiers, while a number of experiments have been devoted to estimating the best value of K in the K-NN. The results have proved that the MLP network fed with the GW-features has the best recognition rate. This approach allows to carry out the diagnosis burden on image processing, feature extraction, and classification algorithms, reducing the cost and the complexity of the acquisition system. In fact, the experimental results suggest that the reason for the high recognition rate in the solder joint classification is due to the proper preprocessing steps followed as well as to the information contents of the features

A Neurofuzzy Method for the Evaluation of Soldering Global Quality Index

Recently, surface mount technology is extensively used in the production of printed circuit boards due to the high level of miniaturization and to the increase of density in the electronic device integration. In such production process several defects could occur on the final electronic components, compromising its correct working. In this paper a neurofuzzy solution to process information deriving from an automatic optical system is proposed. The designed system provides a global quality index of a solder joint, starting from the assessment of a human inspector. This target is achieved by reproducing the modus operandi of the expert, evaluating the area, the shape and the barycentre position of a solder joint. The proposed architecture is constituted by three supervised neural networks and two fuzzy rule-based modules which automate experts work and provide a refined evaluation of the quality. The considered solution presents some attractive advantages: a complex acquisition system is not needed, equipment costs could be reduced by shifting the assessment of a solder joint on the fuzzy parts. Moreover, intermediate variables used in the method could be employed as control parameters in the production process under analysis.

A feature extraction unsupervised neural network for an environmental data set

Environmental data sets are characterized by a huge amount of heterogeneous data from external fields. As the number of measured points grows, a strategy is needed to select and efficiently analyze the useful information from the whole data set. One efficient way of obtaining the validation-compression of data sets is the adoption of a restricted set of features that describe, with an assigned accuracy a subset of the whole data set. One characteristic feature of the environmental data is time dependency: in the medium and long term they are not stationary data sets. The aim of this work is to propose a feature extraction technique based on a new model of an unsupervised neural network suitable to analyze this kind of data. The paper reports the results obtained utilizing the above extraction and analysis procedure on a real data set on chemical pollutants. It is shown that the proposed neural network is able to identify correctly human and/or meteorological effects in the environmental data set.

A Fuzzy Method for Global Quality Index Evaluation of Solder Joints in Surface Mount Technology

In recent years, the requirement of compact devices caused an increasing use of Surface Mount Technology. This technology guarantees the reduction of the size of electronic packages by exploiting solder joint interconnection technology. Nevertheless, parameter variations can occur during the deposition and printing of the soldering paste on a board, compromising its correct working. In this paper, it is proposed a fuzzy architecture for computing an index which provides a quantitative refined assessment about the quality of the soldered interconnections. This task is performed by reproducing the modus operandi of the human experts during their assessments. The proposed architecture consists of three modules connected in series: a feature extraction block and two fuzzy ones. The presented solution keeps the benefits of a neurofuzzy system previously proposed in literature, like the reduction of equipment and computational costs. Moreover, it implies two further advantages: the influence of the human experts in its design is reduced and its implementation is reasonable. Experimental results confirm such advantages, in fact, the architecture approximates the human assessments reliably.

Adaptive particle swarm optimization for CNN associative memories design

In this paper particle swarm optimization is used to implement a synthesis procedure for cellular neural networks autoassociative memories. The use of this optimization technique allows a global search for computing the model parameters that identify designed memories, providing a synthesis procedure that takes into account the robustness of the solution. In particular, the design parameters can be modified during the convergence in order to guarantee minimum recall performances of the network in terms of robustness to noise overlapped to input patterns. Numerical results confirm the good performances of the designed networks when patterns are affected by different kinds of noise.

Angular and axial evaluation of superficial defects on non-accessible pipes by wavelet transform and neural network-based classification

In this paper an effective procedure that allows evaluating the dimensions of corrosive flaws on non-accessible pipes is presented. The method is based on the propagation of ultrasound waves, analyzing the informative content of echoes reflected by defects. The approach exploits the properties of the wavelet transform to represent signals by a reduced form. The coefficients of this representation are selected properly by making use of a filter method followed by a genetic algorithm and, then, they feed a neural network classifier which evaluates the dimensions of defects on the pipe under test. Numerical results show low error rates in the evaluation of both angular and axial extension of each flaw. The main advantage offered by the method consists of analyzing long lines of non-accessible pipes, realizing an automatic evaluation of the dimensions of superficial flaws in pipelines.

Nondestructive evaluation of defects in concrete structures based on finite element simulations of ultrasonic wave propagation

Concrete structures require periodic inspections and quality control to assess their structural integrity. For this task, several methodologies based on the use of different technologies have been previously proposed. In particular, nondestructive evaluations, based on the use of ultrasonic wave propagation, have proved to be attractive due to the possibility to perform reliable assessments of concrete structures. In this paper, an approach exploiting ultrasonic propagation characteristics is proposed. This approach is developed by using data obtained from numerical simulations, whereas proper simulation settings to perform the diagnosis are obtained by preliminary studies. Subsequently, an automatic inspection method to determine the position of defects is developed. The aim of the method consists of determining the position of a defect by the computation of flight times related to signals reflected by anomalies in the structure. Such computation is based on a preliminary classification of defect positions that combines a genetic algorithm for a feature selection and a statistical approach for classification. The performances of the proposed method are evaluated in a specific study case, showing satisfactory numerical results, which show that this approach can be used to identify the position of defects.

An unsupervised multi-swarm clustering technique for image segmentation

Abstract Methods based on Particle Swarm Optimization represent efficient tools to solve a wide class of problems. In particular, they have been successfully applied to data clustering and image processing. In this paper a multi-swarm clustering technique to perform an image segmentation is proposed. The search of the gray levels segmenting the image is carried out by a two-stage procedure. The former is performed by a traditional swarm population, moving in the search space according to a minimum distance criterion. The latter exploits a structure composed by identical swarms that refine the solution of the previous step. The combination of the two swarm approaches allows to tackle the drawbacks of the classical paradigm without making use of a complex implementation. The method is unsupervised, since it identifies the actual number of gray levels to segment the image automatically. Such characteristic is fundamental in the application of image segmentation to real cases, where generally the optimal number of centers is not known a priori and the algorithms are required to face possible environment variations. The conducted experiments show that the proposed technique is able to yield adequate segmentations with a limited computational time, proving to be an interesting tool to face cases in which urgent time constraints have to be satisfied.

Particle swarm optimization-based approach for accurate evaluation of upconversion parameters in Er 3+ -doped fibers

In this Letter, a method for recovering homogeneous upconversion coefficients (HUCs) in Er(3+)-doped glasses and erbium-activated devices is illustrated. It is based on a particle swarm optimization (PSO) approach. The HUCs are calculated on the basis of known values of optical gain evaluated in different pumping conditions. The obtained numerical results proof that the proposed technique provides solutions that are very close to the expected values. Therefore the method constitutes a tool for the design and optimization of efficient rare-earth doped lasers and optical amplifiers. This approach can be considered a feasible and valid alternative method in the field of material science and optical engineering for determining HUCs and avoiding the employment of expensive equipment for the measurement of ion-ion interaction parameters.

PSO-Based Cloning Template Design for CNN Associative Memories

In this brief, a synthesis procedure for cellular neural networks (CNNs) with space-invariant cloning templates is proposed. The design algorithm is based on the use of the evolutionary algorithm of the particle swarm optimization (PSO) with the application to associative memories. The proposed synthesis procedure takes into account requirements in terms of robustness to parametric variations. Numerical results show that the networks also guarantee good performances in terms of correct recall in the presence of noisy patterns.