Search

Industry 4.0 is revolutionizing decision making processes within the manufacturing industry. Among the technological portfolio enabling this revolution, the late literature has capitalized on the potential of data analytics for improving the production cycle at different stages, from resource provisioning to planning, delivery and storage. However, such a promising role of data analytics has been so far explored without a proper, quantitative inspection of the cost-improvement trade-off, nor has the process of acquiring sensors and extracting valuable information from their captured data formalized in a series of methodological steps. This paper introduces the Analyze, Sense, Preprocess, Predict, Implement and Deploy (ASPPID) methodology, an iterative decision workflow that spans from the acquisition of sensing equipment to the quantitative assessment of the contribution of their captured data to enhance the production step under focus. By placing the data scientist at the core of the workflow, this methodology helps improvement teams make informed decisions about which parts of the process need to be sensed, and how to exploit this information towards a verifiable improvement of the production cycle. The implementation of this methodology is exemplified in a real use case within the automotive industry, where the detection of defects in an annealing process can be modeled as a classification problem over a highly imbalanced dataset. Results obtained after applying the proposed ASPPID methodology show that the scrap ratio is reduced by sensing the correct part of the process at minimal investment costs, thus highlighting the crucial role of the data scientist in the management team of manufacturing plants. • ASPPID: an iterative data-based decision workflow for cost-efficiently sensing smart industries. • A comprehensive description of how data analytics should play a central role within the proposed methodology. • A real case study of the application of ASPPID to defect detection in a manufacturing industry. [ABSTRACT FROM AUTHOR]

Abstract: Fingerprint minutiae extraction is a critical issue in fingerprint recognition. Both missing and spurious minutiae hinder the posterior matching process. Spurious minutiae are more frequent than missing ones, but they can be removed by post-processing. In this work, we study the usage of a state-of-the-art minutiae extractor, MINDTCT, and we analyze its major drawback: the presence of spurious minutiae lying on the borders of the fingerprint and out its area. In order to overcome this problem, we use two different filtering approaches based on the convex hull of the minutiae and the segmentation of the fingerprint. We will analyze, supported by an exhaustive experimental study, the efficacy of these methods to remove spurious minutiae. We will evaluate both the effect on different state-of-the-art matchers and the goodness of the minutiae, by comparing the extracted minutiae with the ground-truth ones. For this purpose, the experiments have been performed on several databases of both real and synthetic fingerprints. The filters used allow us to remove spurious minutiae, resulting in more accurate results even in the case of robust matchers. The EER is improved up to 2% for good quality databases, and up to 25% for FVC databases. Additionally, the matching time is accelerated, since less minutiae are processed, attaining up to a 60% runtime reduction for the tested database. [Copyright &y& Elsevier]

Modern machine learning algorithms excel in a great variety of tasks, but at the same time, it is also known that those complex models need to deal with uncertainty from different sources. Consequently, understanding if the model is indeed making accurate predictions or simply guessing at random is not trivial, and measuring the confidence bounds becomes very important. Bayesian machine learning seems to provide the solution, however, many of the state-of-the-art Bayesian algorithms use rigid parametric representations of the uncertainty where the learning process depends on the gradient of a predefined cost function. In this article, a new gradient-free training algorithm based on Approximate Bayesian Computation by Subset Simulation is proposed, where the likelihood function and the weights are defined by non-parametric formulations, resulting in a flexible and fairer representation of the uncertainty. The experiments, specially the engineering case study on composite materials subject to fatigue damage, show the ability of the proposed algorithm to consistently reach accurate predictions while avoiding gradient related instabilities, and most importantly, it provides a realistic and coherent quantification of the uncertainty represented by confidence bounds. All this may lead to a reduction of safety factors in engineering problems, and in general, allows us to make well-informed decisions in situations with a high degree of uncertainty and risk. A comparison with the state-of-the-art Bayesian Neural Networks is also carried out. • Neural networks trained with approximate Bayesian computation. • Accurate and flexible representation of the uncertainty in the observed data. • Stability of predictions thanks to the gradient-free nature of the algorithm. • Non-parametric weights and likelihood function provide adaptability to data. • Appropriate when decisions are dependent on the level of uncertainty. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, we propose the use of weighted linguistic fuzzy rules in combination with a rule selection process to develop accurate fuzzy logic controllers dedicated to the intelligent control of heating, ventilating and air conditioning systems concerning energy performance and indoor comfort requirements. To do so, a genetic optimization process considering an efficient approach to perform rule weight derivation and rule selection is developed. This allows the tuning of the system to be developed at the rule level. The proposed technique was tested considering a physical modelization of a real test site. [Copyright &y& Elsevier]

Detecting the dependency between integration test cases plays a vital role in the area of software test optimization. Classifying test cases into two main classes – dependent and independent – can be employed for several test optimization purposes such as parallel test execution, test automation, test case selection and prioritization, and test suite reduction. This task can be seen as an imbalanced classification problem due to the test cases' distribution. Often the number of dependent and independent test cases is uneven, which is related to the testing level, testing environment and complexity of the system under test. In this study, we propose a novel methodology that consists of two main steps. Firstly, by using natural language processing we analyze the test cases' specifications and turn them into a numeric vector. Secondly, by using the obtained data vectors, we classify each test case into a dependent or an independent class. We carry out a supervised learning approach using different methods for handling imbalanced datasets. The feasibility and possible generalization of the proposed methodology is evaluated in two industrial projects at Bombardier Transportation, Sweden, which indicates promising results. • In a manual testing procedure, all testing artifacts are written in a natural text, employing natural language processing techniques might provide highly useful information for test optimization purposes. • The ratio of dependent and independent test cases might suffer from an imbalanced distribution due to the testing level and complexity of the system under test. • Doc2Vec proves to be a good tool when transforming the manual test cases into feature vectors. • IFROWANN performs well when splitting dependent and independent test cases as an imbalance learning algorithm. [ABSTRACT FROM AUTHOR]