Robert Bosch (France)

In an industrial maintenance context, degradation diagnosis is the problem of determining the current level of degradation of operating machines based on measurements. With the emergence of Machine Learning techniques, such a problem can now be solved by training a degradation model offline and by using it online. While such models are more and more accurate and performant, they are often black-box and their decisions are therefore not interpretable for human maintenance operators. On the contrary, interpretable ML models are able to provide explanations for the model’s decisions and consequently improves the confidence of the human operator about the maintenance decision based on these models. This paper proposes a new method to quantitatively measure the interpretability of such models that is agnostic (no assumption about the class of models) and that is applied on degradation models. The proposed method requires that the decision maker sets up some high level parameters in order to measure the interpretability of the models and then can decide whether the obtained models are satisfactory or not. The method is formally defined and is fully illustrated on a decision tree degradation model and a model trained with a recent neural network architecture called Multiclass Neural Additive Model.

The OH*, CH* and CO 2∗ chemiluminescence signals of methane/air premixed laminar flames stabilized over a nonadiabatic porous plug burner are compared to the signals measured from a nearly adiabatic conical flame in a series of experiments. The impact of reactant stream temperature is also characterized. A numerical study based on 1-D flame models then follows to support the experimental results. It is found both in experiments and in simulations that the linear relationship between the mixture flowrate and the chemiluminescence intensities is no longer valid when flames are closely attached to the burner surface due to the heat transfer between the flame and the burner. The transition between the linear and the nonlinear regimes is identified as the gas flow velocity drops below the adiabatic laminar burning velocity calculated at the bulk temperature of the flow leaving the burner. When the mass flowrate is kept constant, preheating of the reactant stream increases the chemiluminescence intensity for a freely propagating flame, but has almost no impact for a burner-stabilized flame. It is finally found that the OH* and CH* chemiluminescence intensities correlate with the burnt gas temperature for the adiabatic but also the nonadiabatic flames. The underlying physical mechanisms are discussed. Finally, the evolution of the CH*/OH* ratio with the inlet gas velocity is discussed.

The development of engine control unit (ECU) systems for series production requires an important validation phase. In order to reduce the number of time consuming and expensive vehicle tests, the ECU is validated using hardware in the loop (HIL) test bench. The HIL simulates the engine behavior in real-time simulations to generate consistent sensor values for all engine operating points, e.g., starting phase, transient behavior, static behavior, etc. Mean value engine models (MVEM) are able to run in real time and are appropriate for HIL test systems. In this paper we present a full MVEM taking into account all engine components: air system (air filter, turbocharger, charge air cooler, throttle valve, intake and exhaust manifolds, EGR valve, and turbine), oxidation catalyst (OxiCat), and diesel particulate filter (DPF). Additionally, combustion models have been developed to simulate the influence of the injection strategies (pre, main, post, and late injections) on the exhaust temperature and the unburned hydrocarbon emission (HC). These are taken into consideration in the exothermal reactions models inside OxiCat and DPF. The results show that the model prediction in term of pressure and temperature are in good agreement with the original equipment manufacturer (OEM) project data. The after treatment temperature evolutions in the OxiCat and DPF are well reproduced by the proposed model.

A new type of assembly and interconnection technology in power modules has been developed to connect MOSFETs. These power modules, used as frequency inverters for electric feature, have an innovative design. They avoid using aluminum wire bond, often to be blamed for device failure, by using a copper clip soldered on the top side of the chip. The successful design for increased reliability of this electronic package depends on better understanding and modeling its fatigue behavior and its failure mechanisms. During Active Power Cycling tests, the chip acts as a heat source and temperature gradients develop in the package causing expansion of the different materials. To assess the reliability of those devices under thermal power cycling loads, both experiments and simulations have to be performed. Some failures were already observed after Active Power Cycling tests, but they do not correspond to failures usually observed in standard MOSFET packages, and thus are not thoroughly understood. For instance, the formation of a wave in the aluminum metallization layer, on top of the chip, caused by a high deformation grade, was never described. High operating life of more than 1 million cycles can be achieved by optimizing clip geometry and thicknesses of metal layers. Such packages are then clearly more robust compared to those using wire bond technology. In this paper, failures observed via testing are confronted with thermal and mechanical stresses distribution computed by Finite Element Analysis in order to improve the understanding of failure formation mechanisms. A 2D Finite Element model of MOSFET packages is used to analyze mechanical stresses induced by thermal loads. Simulations help in determining critical areas and then in improving the design of modules.

This paper demonstrates the influence of the trailing end of the piston, or inboard, pad on the propensity of a brake to generate audible squeal. The unique technique was employed to measure the dynamic CoP at the pad/disk interface during a normal braking operation. This novel technique uses an embedded pressure sensitive film within the pad. The paper also presents the co-planar analysis of the pad/caliper contact regions and its influence on the position of the disk/pad CoP. The analysis shows how the trailing end of the piston pad influences the position of the CoP and goes on to explain why the centre moves as the pad abutment vibrates against the abutment. The paper includes additional work with a 12-piston opposed caliper where the initial CoP may be varied both along the pad and radially. Results show a very definite movement of the CoP as the brake pressure is increased.

We are considering the problem of observer design for a general class of state-affine systems in presence of output delay and sampling. Two novelties characterize the considered class of systems: (i) the state equation is subject to output injection involving future output values (not accessible to measurement); (ii) the injected output values come in the state equation not only through a driving term but also through the state matrix. These novel characteristics entail the loss of the system state-affine nature and lead to a new observer design problem never investigated so far. The solution we develop in this paper is a sampled-data based observer of Kalman-like type, augmented with inter-sample predictors and signal saturations. Using both Lyapunov and small-gain arguments, we show that the observer is exponentially convergent for sufficiently small sampling interval and delay, provided an observability condition is satisfied.

Generally speaking, the behaviour of viscoelastic material is more complicated than the behaviour proposed by classical models as Voigt, Maxwell or Zener. The stiffness of such materials is a frequency dependent complex function. Above 1000Hz, classical measurements techniques are unable to achieve accurate measurements of the stiffness. In this paper, a new Dynamical Mechanical Analysis (DMA) tester is presented. It allows the characterization of the shear stiffness of preloaded viscoelastic materials between 200 and 3500Hz and without using frequency-temperature equivalences. Then the Generalized Maxwell model is used to describe behaviours measured with the DMA tester. A new iterative identification method of the parameter of the Generalized Maxwell model is described. This identification method is based on the asymptotes of the model.

International audience

Dans le cadre de la simulation des systèmes de freinage, la mise en évidence de phénomènes tels que le crissement fait appel à des méthodes spécifiques. On s'intéressera ici à l'analyse de stabilité par le calcul des pulsations et modes propres. L'utilisation de dispositifs amortissants "shims" pour la réduction du bruit de crissement implique que les simulations prennent en compte les effets amortissants des matériaux. On utilise pour cela des modèles de matériaux viscoélastiques et en particulier le modèle de Maxwell généralisé. Les travaux ont consisté à choisir et valider des outils (modèles viscoélastiques, formulations éléments finis, ...) et à mettre en place une méthode de simulation. Ils se sont divisés en plusieurs étapes. Dans un premier temps, il a fallu sélectionner, parmi plusieurs modèles mathématiques permettant de décrire la viscoélasticité, le plus adapté à la modélisation de l'amortissement dans les systèmes de freinage. On a étudié les modèles à module complexe, de Kelvin-Voigt et de Maxwell généralisé. Dans un second temps, il a fallu trouver une formulation du problème éléments finis adaptée à l'analyse modale. On a alors utilisé une formulation en modèle d'état. Une hypothèse importante concernant l'égalité des pôles du modèle de Maxwell a permis de limiter la taille du modèle tout en simplifiant sa formulation. Enfin, les méthodes mises en place ont été appliquées à une plaquette de frein dans un premier temps et à un frein complet ensuite. Des analyses paramétriques sur le nombre de modes de projection et l'ordre des modèles de Maxwell ont été menées sur ces deux cas test. Dans le cas de la plaquette de frein, on a effectué une analyse modale complexe en prenant en compte la viscoélasticité du matériau de friction. L'objectif était d'obtenir un recalage de la plaquette, en terme de fréquences et d'amortissements, par rapport à une mesure FRF. Dans le cas du modèle de frein complet, on a effectué une analyse modale complexe en prenant en compte la viscoélasticité des matériaux de friction et du "shim". Le modèle de frein a également servi de support à la mise en évidence du lien entre la répartition de l'énergie de déformation dans le frein et l'amortissement observé par analyse modale complexe. Pour conclure, on a développé une méthode de simulation utilisant Abaqus®, Matlab® pour la phase d'analyse modale complexe avec effets viscoélastiques et Python pour l'interfaçage entre Abaqus® et Matlab®. Cette méthode permet de traiter des problèmes d'analyse de stabilité avec la prise en compte de plusieurs matériaux viscoélastiques, sur des modèles de grande dimension, sans surcoût en terme de temps de calcul.

Augmented Reality (AR) for industry has become a significant research area because of its potential benefits for operators and factories. AR tools could help to collect data, create standardized representations of industrial procedures, guide operators in real-time during operations, assess factory efficiency, and elaborate personalized training and coaching systems. However, AR is not yet widely deployed in industries, and this is due to several factors: hardware, software, user acceptance, and companies’ constraints. One of the causes we have identified in our factory is the poor user experience when using AR assistance software. We argue that adding computer vision and deep learning (DL) algorithms into AR assistance software could improve the quality of interactions with the user, handle dynamic environments, and facilitate AR adoption. We conduct a preliminary experiment aiming to perform 3D pose estimation of a boiler with MobileNetv2 in an uncontrolled industrial environment. This experiment produces insufficient results that cannot be directly used but allow us to establish a list of challenges and perspectives for future work.

Le succès de la Logan de Renault intrigue. N’est-ce pas un succès par erreur, destiné à être éphémère ? À moins qu’il n’annonce une révolution semblable à celle qu’a introduite Ford au siècle dernier‚ ou à celle du low cost dans le transport aérien.

Innovation is one of the aims of many organisations. It is a mandatory condition for durable industrial activities, even though it bears specific issues, such as the involvement of substantial resources, temporary low profitability, high risks and failure rates, which need to be overcome to succeed. Therefore we investigated the notion of innovation performance, as it is critical to analyse the reach of organisational innovation-related objectives and the allocation and use of resources. The measurement of such a performance is generally done through the observation of organisation Research and Development (R&D) investments as well as the quantity of patent applications. However, it was shown that even if the R&D investments are an assessment element of the innovation process inputs, the patents do not show a fair image of the innovation capacity, namely because of their relative value and use [1]. Some authors focused on the assessment of the innovation capacity (or innovation potential), studying the factors that enable innovation. We propose to synthesise the fields covered by these factors and will investigate two of the most advanced models that offer different approaches and perspectives [2], [3]. Based on these elements, we analyse the interactions between classical TRIZ and the organisational innovation capacity factors in order to determine TRIZ potential. This methodological and theoretical approach is then confronted to a field example: the design of a hybrid-heating appliance (compact electrical heat-pump combined with a gas condensing boiler) within Robert Bosch company. This study is marked by the context of regulation evolutions such as the European Directive on Energy-related Products (ErP) and strong technical challenges to overcome. Results shows the outcomes that companies can expect out of TRIZ, which are seen as vectors of expansion for the understanding and the use of the theory in the industrial world. It also emphasizes what are the points that cannot be covered by classical TRIZ within all the different fields of the innovation capacity (Strategy, Organisation, Process, Resources, Tools and Culture). The reasons for the deviations between theoretical approach and the studied case are analysed and perspectives are given on a new approach combining TRIZ and the Blue Ocean Strategy.

EcoDesign is not easily fitting in the Innovation process, mostly owing to the dedicated time and resources required for learning and applying the EcoDesign-related knowledge when fast decision-making is occurring in early stages of product development and as there is only a low level of information on the new product to be made. In this paper, we propose a way to integrate EcoDesign in the early steps of the innovation process, while taking into account available processes and tools. A pre-assessment-based approach is brought using product typology, flow study and internal Functional Analysis (FA). We illustrate our proposition with an example on a new heating appliance design. Using this case study, we show how the innovation requirements could be met and how standard tools can be employed along the product development phases (Idea Management, Feasibility Assessment and Time To Market).

<div class="htmlview paragraph">Nowadays, the sever customer targets in term of squeal require an increasing number of different tests. This paper addresses how to develop a new test bench procedure which helps to fulfill this new target with more efficiency.</div> <div class="htmlview paragraph">With a single standard test procedure, we want to:</div> <div class="htmlview paragraph"> <ul class="list disc"> <li class="list-item"><div class="htmlview paragraph">find out a maximum number of frequencies in a minimum number of stops,</div></li> <li class="list-item"><div class="htmlview paragraph">improve road correlation with adapted braking conditions including detection of the low pressure risks,</div></li> <li class="list-item"><div class="htmlview paragraph">improve the stability of the results to get more valuable test comparison.</div></li> </ul> </div> <div class="htmlview paragraph">A statistical analysis performed over three years of chassis dyno matrix test shows that noise risks in low pressures conditions are usually not well estimated in standard procedures. Indeed, pressure, speed and temperature steps are arbitrarily constant in most standard procedures and do not take into account the non-linear influence of these parameters on the noise generation.</div> <div class="htmlview paragraph">The objective of this new chassis dyno procedure is to improve the distribution of braking pressure conditions in a single procedure without lengthening the test duration.</div> <div class="htmlview paragraph">The large scale statistical study performed on three years of tests permits to determine an optimal pressure distribution of tested conditions. Some of the results are presented in this paper.</div>

This study is part of a new product design project. Its scope is the enhancement of a hybrid boiler, which is an innovative product using different energy sources (natural gas and electricity with a share of renewable energy). One of the key points in this project is to make this boiler more communicative by wiring it to an information system. This system called “Thermibox” gives us a way of monitoring boiler parameters when it works on the field in a real use case. Moreover, as a complete information system, the Thermibox could be a significant link between the products, the customer and our company which designs and manufactures the product. Within TRIZ framework, the goal of this study is to show a new way of documenting expected contradictions and finding physical and technical contradictions that were not expected while designing the product by using a smart monitoring system. The main purpose is to define a set of parameters that could have a correlation and then find how they interact by monitoring them and using algorithms taken from data mining. The system should also be able to classify the contradictions according to their frequency or their intensity and help us finding the available resources to solve them. On the one hand, this tool could give us the necessary information to design a new product as soon as we have a functional prototype, or to improve an existing one for a new version. The created information should be an adequate input to roll out the TRIZ process and work on contradictions that would have been hard to identify in the normal way. On the other hand, if the contradictions we brought to light can be formalized easily, the balance between system parameters could be used in an active way to optimize the regulation and parameterization of the product.

International audience

National audience

National audience

The main goal of the PIONEERS project is to improve the safety of Powered-Two-Wheelers by providing an integrated approach to rider protection considering on-rider (Personal Protective Equipment) and on-board systems. The implementation of the PIONEERS' main results will contribute to reducing PTW fatalities and injuries by defining test methods to develop protective systems and on-board systems to reduce impact severity. In order to assess if this major conclusion of the PIONEERS project is being fulfilled and to quantify the benefit, an Impact Evaluation has been developed in this document. This evaluation concerns both economical and safety benefits (in terms of avoided or mitigated accidents, reduction of morbidity and severity of injuries) of the following proposed PTW safety countermeasures that have been developed in PIONEERS: PreCrash Braking System, Airbag jacket and the PTW-PPE communication system, Motorcycle and scooter leg protector. Concerning the PCB, the evaluation was obtained via computer simulations of a set of 60 real-world in-depth crashes. The effects were assessed in terms of a reduction of the absolute and relative impact speed of the PTW. A parametric approach in which PCB intervention parameters were varied (field of view, range, deceleration, fade-in jerk, triggering strategy...) was adopted to compute the effects of the system for different conditions. Such approach led to the identification of three combinations of parameters to represent typical system effects assuming a pessimistic (low efficiency), average, and an optimistic approach (high efficiency). Depending on the set of parameters, benefits in terms of speed reduction can go until a median value of 15km/h. Then societal benefits have been calculated in terms of casualties' reduction or crashes using Injury Risk Function and the new relative impact speed distribution. Results show that global benefits for slight, serious or fatal injuries are included between -4% to -31%. Finally, economic benefits in terms of cost were evaluated using the SafetyCube software. Several configurations were considered like the implementation rate of PCB among all motorcycles in Europe, the cost to promote such systems, the horizon, etc. In a 5 years period, this cost will be balanced by savings from health and social expenses, and in a high rate of implementation (with the average evaluation) Europe could save, at least, 30 human lives in this period. With an optimistic configuration, the net value of the benefits could reached between approximately 500k€ until more than 90 M€. Concerning the Airbag jackets, the societal benefits calculation was first based on the establishment of Injury Risk Function for slight, serious and fatal injuries on the trunk without the airbag. In order to establish these IRF, four accident databases were considered and 382 accidents concerned a rider with at least one injury in trunk body region. Then, according to the work performed in others WP's of the Pioneers project, three hypotheses have been considered for the level of protection and the reduction of injuries. They considered that the airbag jacket provides a reduction of an AIS-1 for speed impact lower than 20km/h, respectively 30km/h and 40km/h. Results of the societal evaluation show that between 1,3% until 19% of injuries could be avoided in function of the configuration. From an economical point of view, with a high implementation rate (6%) in the next 5 years, a net value of more than 140M€ could be saved. Concerning the Lateral Protectors, Injury Risk Function of the Lower Leg in lateral impact configuration were first established based on 81 accidents. Then, based on the findings from the lateral protection devices that have been developed in the PIONEERS project (in particular WP3), two hypotheses have been retained for the level of protection of such systems: one considered as a « low-speed » countermeasure (only effective until 15km/h) and one considered as « medium-speed » countermeasure (only effective between 15 to 30km/h). Global societal and economic benefits have been observed but results have to be taken very cautiously. Indeed, the economic evaluation show that an amount of more than 5M€ for the net benefits could be saved with the best configuration. But it is important to remind that these systems are not still available in the market, so these evaluations have to be considered as a theoretical exercise. In conclusion, this work allowed to evaluate safety and economic benefits of several protective systems: Pre-Crash Braking, airbag jackets, Lateral protectors. But because some of these systems are prototypes and not yet on the market, all the evaluations have to be taken very carefully. In particular, results from societal and economical evaluations are sensible and have to be considered cautiously. Beyond these specific evaluations, theroretical methodologies have been defined and could be applied on other systems.

An abstract is not available for this content so a preview has been provided. As you have access to this content, a full PDF is available via the ‘Save PDF’ action button.