Search

Quality control has long been one of the most challenging fields of manufacturing. The development of advanced sensors and the easier collection of high amounts of data designate the machine learning techniques as a timely natural step forward to leverage quality decision support and manufacturing challenges. This paper introduces an original dataset provided by the automotive supplier company VALEO, coming from a production line, and hosted by the Ecole Normale Superieure (ENS) Data Challenge to predict defects using non-anonymised features, without access to final test results, to validate the part status (defective or not). We propose in this paper a complete workflow from data exploration to the modelling phase while addressing at each stage challenges and techniques to solve them, as a benchmark reference. The proposed workflow is validated in series of experiments that demonstrate the benefits, challenges and impact of data science adoption in manufacturing.

In this paper an example is discussed concerning the possibility of using metrological concepts, like uncertainty of measurement, systematic effects, statistical distribution of measurements, in order to improve the interpretation capability of a specific failure mechanism in a controlled field of influences. This approach has been also used to set an interpolation in a certain range of condition when the temperature of tests is changed. The case study refers to the degradation assessment of ultrasonic sensors for distance measurement, which is ascribed to the quality decay of solder joints due to cracks or remarkable shrink holes. The components are subjected to ALT cyclic thermal stresses. The results of tests allowed to gaining information about the validation of a model of fault, useful to design accelerated tests at higher temperatures.

This paper proposes a methodology to verify products reliability through its life time using financial information from warranty cost for a long period at 8 and 15 years. It will present the challenge that automotive industry faces and study the potential improvement to be considered in order to improve the reliability during the design and validation phases. The formulas developed in this research permit one to define the reliability of the product without detailed warranty information from the field and as a consequence simplify the treatment of the data. In addition, this approach allows to assess a financial risk for warranty extension.

In the case of very reliable materials, failures are rare events, and it is often difficult to find a sample of the product that includes instances of failure merely by analyzing feedback from experience (or performing dedicated reliability tests) in order to estimate the reliability parameters. Success often varies wildly as a function of the size of this sample. Thus, the need to understand the behavior of a long-life product before it is commissioned frequently leads to so-called accelerated life testing during the development, qualification or even production phases. This chapter presents the methodology of accelerated life testing, as well as the methods and tools used to exploit accelerated life tests, and gives a selection of examples.

The reliability of high-power mechatronic systems fundamentally depends on the reliability of film capacitors.

The purpose of this chapter is to provide the reader with an overview of the methods and modeling used in reliability prediction, that is to say, in terms of pre-project reliability. How can we quantify a failure rate for a product that is not yet on the market?

The reliability of the power module is mainly related to that of its electronic components (IGBT power chips and diodes). These components undergo severe and varied stresses. Indeed, the electrical pulses due to the internal operations of these components cause thermal loadings inducing mechanical deformations which can go beyond the thresholds desired. Furthermore, these electronic components are subjected to vibrations which can affect the state the solder joints and the electrical wire connections, causing damage by thermomechanical fatigue. To this end, knowledge of the mechanical response of the power module is vital for the electronics industry as these modules fulfill strategic functions, as is the case for electric and hybrid vehicles. Knowledge of the mechanical behavior of this power module requires the modeling of several physical phenomena. Multi-physical modeling aims at considering the interdependencies and interactions between different physical phenomena such as electrical, thermal and mechanical. In this study, multi-physical coupling is performed using ANSYS software.

The idea of highly accelerated testing is already a few decades old, and belongs to the family of pro-active testing strategies: it aims to maximally exploit the potential of the technology available at any given moment. Approaching this point means that we are approaching the international state-of- the-art, or in other words what we refer to as product excellency. This chapter describes the methodology of highly accelerated testing at a theoretical level. Readers may refer to the Appendix “HALT-HASS methodology” for a detailed description of how the highly accelerated life tests (HALT) and highly accelerated stress screening (HASS) approaches are executed in a laboratory that offers this type of service. The chapter concludes by comparing accelerated life testing and highly accelerated testing.

As a theme that has been predominant for a few years now, innovation is widely used in corporate and government communications. Considered as one of the spearheads of business competitiveness amongst companies and in particular as a one of the best responses to the growing competition emerging from developing countries, innovation is today at the heart of European territorial issues.

The phenomenon of globalization and the intense need for innovation in our daily lives suits the automotive world well. In order to fully understand the issues and challenges facing this atypical industrial sector, it is necessary to review the evolution of the very concepts of the automobile and the notion of reliability over time.

This study illustrates a methodology used in our company (car components manufacturer) which enables to generate, starting from any given customer vibration specification, a fatigue damage equivalent Power Spectrum Density (PSD) signal. This PSD can be used to evaluate the mechanical durability of engine cooling or engine-mounted components. The methodology permits to accelerate durability bench tests, to compare the severity of different vibration specifications and to perform structural simulation. One practical example will illustrate the experimental validation of the procedure, based on strain-damage measurements. The main limitations of the methodology are related to the nature of the specified signal. In the case of specifications composed of purely random signals, the method is always applicable.

Suppliers of shaker controllers often use a technique called ‘sigma clipping’ to prevent the amplifier from encountering too high amplitude peaks. Sigma clipping is most commonly used to make sure a shaker system will be sufficiently powerful to reproduce the required random spectrum. The level of clipping is measured via the crest factor, which is expressed as a ratio between the absolute maximum of the signal and its root-mean-square (RMS) value. Such clipped drive signals are therefore no longer Gaussian, since their probability distribution functions (PDF) are truncated at high values. The clipped drive signal is then amplified and filtered by a system made up of the shaker itself followed by the unit under test. Although the local stress and strain responses do not show any obvious clipping, their statistics may be affected. This can feed through to fatigue damage, which is known to be exponentially related to strain amplitudes. A legitimate question is therefore “how does clipping affect damage?” This paper studies and quantifies the influence of clipping experimentally, numerically and theoretically. There are many objectives and underlying questions to consider. For the test technicians, what level of clipping to choose in order to represent the damage due a Gaussian environment? Whereas for the simulation engineers, how to enhance damage estimations by taking into account the level of clipping?