The demand for better indoor environment has led to a wide use of heating, ventilating and air conditioning (HVAC) systems. Employing advanced HVAC control strategies is one of the strategies to maintain high quality indoor thermal comfort and indoor air quality (IAQ). This thesis aims to analyse and discuss the potential of using advanced control methods to improve the indoor occupants’ comfort. It focuses on the development of controllers of the major factors of indoor environment quality in buildings including indoor air temperature, indoor humidity and indoor air quality. Studies of the development of control technologies for HVAC systems are reviewed firstly. The problems in existing and future perspectives on HVAC control systems for occupants’ comfort are investigated. As both the current conventional and intelligent controllers have drawbacks that limit their applications, it is necessary to design novel control strategies for the urgent issue of indoor climate improvement. Hence, a concept of designing the controllers for indoor occupants’ comfort is proposed in this thesis. The proposed controllers in this research are designed by combining the conventional and intelligent control technologies. The purpose is to optimize the advantages of both conventional and intelligent control methods and to avoid poor control performance due to their drawbacks. The main control technologies involved in this research are fuzzy logic control (FLC), proportional-integral-derivative (PID) control and neural network (NN). Three controllers are designed by combining these technologies. Firstly, the fuzzy-PID controller is developed for improvement of indoor environment quality including temperature, humidity and indoor air quality. The control algorithm is introduced in detail in Section 3.2. The computer simulation is carried out to verify its control performance and potential of indoor comfort improvement in Section 4.1. Step signal is used as the input reference in simulation and the controller shows fast response speed since the time constant is 0.033s and settling time is 0.092s with sampling interval of 0.001s. The simulating result also proves that the fuzzy-PID controller has good control accuracy and stability since the overshot and steady state error is zero. In addition, the experimental investigation was also carried out to indicate the fuzzy-PID control performance of indoor temperature, humidity and CO2 control as introduced in Chapter 5. The experiments are taken place in an environmental chamber used to simulated the indoor space during a wide period from late fall to early spring. The results of temperature control show that the temperature is controlled to be varying around the set-point and control accuracy is 4.4%. The humidity control shows similar results that the control accuracy is 3.2%. For the IAQ control the maximum indoor concentration is kept lower than 1100ppm which is acceptable and health CO2 level although it is slightly higher than the set-point of 1000ppm. The experimental results show that the proposed fuzzy-PID controller is able to improve indoor environment quality. A radial basis function neural network (RBFNN) PID controller is designed for humidity control and a back propagation neural network (BPNN) PID controller is designed for indoor air quality control. Then, in order to further analyze the potential of using advanced control technologies to improve indoor environment quality, two more controllers are developed in this research. A radial basis function neural network (RBFNN) PID controller is designed for humidity control and a back propagation neural network (BPNN) PID controller is designed for indoor air quality control. Their control algorithms are developed and introduced in Section 3.3 and Section 3.4. Simulating tests were carried out in order to verify their control performances using Matlab in Section 4.2 and Section 4.3. The step signal is used as the input and the sampling interval is 0.001s. For RBFNN-PID controller, the time constant is 0.002s, and there is no overshot and steady state error. For BPNN-PID controller, the time constant is 0.003s, the overshot percentage is 4.2% and the steady state error is zero based on the simulating results. Simulating results show that the RBFNN-PID controller and BPNN-PID controller have fast control speed, good control accuracy and stability. The experimental investigations of the RBFNN-PID controller and BPNN-PID control are not included in this research and will carried out in future work. Based on the simulating and experimental results shown in this thesis, the indoor environment quality improvement can be guaranteed by the proposed controllers.

Sustainable energy systems require deployment of new technologies to help tackle the challenges of climate change and ensuring energy supplies. Future sources of energy are less economically competitive than conventional technologies, but there is the potential for cost reduction. Tools for modelling technological change are important for assessing the deployment potential of early-stage technologies. Learning curves are a tool for assessing and forecasting cost reduction of a product achieved through experience from cumulative production. They are often used to assess technological improvements, but have a number of limitations for emerging energy technologies. Learning curves are aggregate in nature, representing overall cost reduction gained from learning-by-doing. However, they do not identify the actual factors behind the cost reduction. Using the case study of onshore wind energy, this PhD study focuses on combining learning curves with engineering assessment methods for improved methods of assessing and managing technical change for emerging energy technologies. A third approach, parametric modelling, provides a potential means to integrate the two methods.

La conception sûre et le développement maîtrisé des systèmes humain(s)-machines(s)s est un enjeu majeur pour l’ensemble de la société et, spécialement, dans les domaines critiques comme les domaines militaire, aérospatial et médical qui sont amenés à évoluer conjointement, grâce aux progrès réalisés par l’automatisation. Parallèlement, les techniques disponibles pour assurer la mise en place de solutions répondant à ces besoins sont également en constante progression concernant leurs dimensions de puissance, efficacité énergétique, miniaturisation, ce qui correspond à la notion de convergence Nanotechniques-Biotechniques-Informatique-Sciences Cognitives.Tout cela rend nécessaire d’avoir une ergonomie théorique, non pas au sens parfois entendu d’un détachement de la pratique mais au sens d’une ergonomie fondée sur un cadre conceptuel et formel permettant de rendre intelligibles les méthodologies et normes utilisées à travers des principes généraux.Ainsi, l’ergonomie théorique n’est pas une théorie de l’ergonomie mais une théorie de l’anthropotechnique, qui cherche à développer une méthodologie et une épistémologie pragmatique, propre à apporter les fondements conceptuels à ce qui existe aujourd’hui et à permettre le développement futur de l’ergonomie comme champ scientifique. Elle repose sur l’idée centrale qu’à la base de toute Interface Humain(s)- Machine(s) se produit une interaction particulière ; l’interaction sémiotique.Cette interaction particulière, propre aux être vivants, est à la base de l’intégration entre l’individu et son environnement. C’est elle qui lui permet d’agir dans le monde en produisant de la signification à partir de la perception sensible. En envisageant l’humain comme un système biologique intégré, au sens de Chauvet, c’est à dire comme un système dont le comportement est le produit des couplages implicites à différents niveaux de structure, il est possible de fonder un paradigme postcognitiviste en psychologie générale. Le propre de ce paradigme est qu’il permet d’envisager le comportement de l’individu à partir de sa biologie et de façon non réductionniste.À partir des travaux de Chauvet et de Fass, il est possible d’établir une modélisation mathématique de ces phénomènes biologiques qui permette de formaliser les spécifications d’un système technique, à partir des exigences liées aux facteurs humains. Grâce à ce cadre conceptuel, il est possible d’analyser et de réformer certains outils précieux pour l’ergonome comme la critériologie de Bastien et Scapin dont nous proposons une généralisation à tous les systèmes techniques. Cette critériologie a pour fonction originelle de permettre qu’un système interactif réponde à un ensemble fondamental d’exigences liées aux facteurs humains. En la généralisant et en la formalisant, il devient possible d’envisager une modélisation des paramètres fondamentaux de l’interaction sémiotique.Cette approche aboutit dans une ontologie de conception, qui a pour ambition d’être un cadre formel de modélisation de l’interaction humain(s)-machine(s). Nous en proposons une application à MIDAS, outil de modélisation de la performance humaine, développé par la NASA. Le propos de cette ontologie est de permettre la modélisation de l’interaction de façon simultanée à la modélisation du comportement de l’utilisateur ce qui permet de mettre en évidence quels sont les paramètres hamartogènes d’une hypothèse de conception.Ainsi, l’usage d’ontologies de conception ouvre la voie à la construction de modèles formels, corrects par construction, de l’interaction humain(s)-machine(s) et de ses conditions optimales de mise en place, permettant ainsi une conception plus assurée des systèmes à dimension critique. / The safe design and controlled development of human(s)-machine(s) systems is a major challenge for society as a whole and, especially, in critical areas such as the military, aerospace and medical fields that are brought to evolve, thanks to the progress made by automation. At the same time, the techniques available to ensure the implementation of solutions that meet these needs are also constantly increasing with regard to their power dimensions, energy efficiency and miniaturization, which corresponds to the notion of convergence between Nanotechniques-Biotechniques-Informatics-Cognitive Sciences.All this makes necessary to have theoretical ergonomics, not in the sense sometimes understood of a detachment of the practice but in the sense of ergonomics based on a conceptual and formal framework making possible to render the methodologies and norms used through general principles intelligible.Thus, theoretical ergonomics are not a theory of ergonomics but a theory of anthropotechnics, which seeks to develop a methodology and a pragmatic epistemology, capable of providing the conceptual foundations for what exists today and to allow the future development of ergonomics as a scientific field. It is based on the central idea that at the base of any Human(s)-Machine(s) Interface occurs a particular interaction ; semiotic interaction.This particular interaction, specific to living beings, is at the basis of the integration between the individual and his environment. It is what allows it to act in the world, by producing meaning from sensible perception. By considering the human as an integrated biological system, in the sense of Chauvet, as a system whose behavior is the product of implicit couplings at different levels of structure, it is possible to found a postcognitive paradigm in general psychology. The peculiarity of this paradigm is that it makes possible to envisage the behavior of the individual from his biology and in a non-reductionist way.From the work of Chauvet and Fass, it is possible to establish a mathematical modeling of these biological phenomena, allowing to formalize the specifications of a technical system, based on the requirements related to human factors. Thanks to this conceptual framework, it is possible to analyze and reform some valuable tools for the ergonomist like the criterium of Bastien and Scapin for which we propose a generalization to all technical systems. The primary function of this criterology is to allow an interactive system to meet a fundamental set of requirements related to human factors. By generalizing and formalizing it, it becomes possible to envisage a modeling of the fundamental parameters of the semiotic interaction.This approach results in a design ontology, which aims to be a formal framework for modeling human(s)-machine(s) interactions. We propose an application to MIDAS, a human performance modeling tool developed by NASA. The purpose of this ontology is to allow the modeling of the interaction simultaneously to the modeling of the behavior of the user which makes it possible to highlight what are the hamartogenic parameters of a given hypothesis of conception.Thus, the use of design ontologies opens the way to the construction of formal models, correct by construction, of the human(s)-machine(s) interaction and its optimal conditions of implementation, allowing a more assured design of critical systems.

In order to enhance safety and prevent the recurrence of major accidents it is necessary to understand why they occur. This understanding is gained by utilising accident causation theory to explain why a certain combination of events, conditions and actions led to a given outcome: the process of accident analysis. At present, the systems approach to accident analysis is arguably the dominant research paradigm. Based on the concepts of systems theory, it views accidents as the result of unexpected and uncontrolled relationships between a system s components. Various researchers claim that use of the systems approach, via systemic accident analysis, provides a deeper understanding of accidents when compared with traditional theories. However, the systems approach and its analysis techniques are yet to be widely adopted by the practitioner community and, therefore, a research-practice gap exists. The implication of such a gap is that practitioners may be applying outdated accident causation theory and, consequently, producing ineffective safety recommendations. The aim of this thesis was to develop the current understanding of the systemic accident analysis research-practice gap by providing a description of the gap, considering its extent and examining issues associated with bridging it. Four studies were conducted to achieve this aim. The first study involved an evaluation of the systemic accident analysis literature and techniques, in order to understand how their characteristics could influence the research-practice gap. The findings of the study revealed that the systems approach is not presented in a consistent or clear manner within the research literature and that this may hinder its acceptance by practitioners. In addition, a number of issues were identified (e.g. model validation, analyst bias and limited usage guidance) which may influence the use of systemic analysis methods within industry. The examination of how the analysis activities of practitioners may contribute to the gap motivated Study 2. This study involved conducting semi-structured interviews with 42 safety professionals and various factors, which affect the awareness, adoption and usage of the systems approach and its analysis methods, were highlighted. The combined findings of Studies 1 and 2 demonstrate that the systemic accident analysis research-practice gap is multifaceted in nature. Study 3 investigated the extent of the gap by considering whether the most widely used analysis technique (the Swiss Cheese Model) can provide a systems approach to accident analysis. The analysis of a major rail accident was performed with a model based on the Swiss Cheese Model and two systemic analysis methods. The outputs and usage of the three analysis tools were compared and indicate that the Swiss Cheese Model does provide a means of conducting systemic accident analysis. Therefore, the extent of the research-practice gap may not be as considerable as some proponents of the systems approach suggest. The final study aimed to gain an insight into the application of a systemic accident analysis method by practitioners, in order to understand whether it meets their needs. Six trainee accident investigators took part in an accident investigation simulation and subsequently analysed the data collected during the exercise with the Systems Theoretic Accident Modelling and Processes model. The outputs of the participants analyses were studied along with the evaluation feedback they provided via a questionnaire and focus group. The main findings of the study indicate that the analysis technique does not currently meet the usability or graphical output requirements of practitioners and, unless these issues are addressed, will struggle to gain acceptance within industry. When considering the research findings as a whole a number of issues are highlighted. Firstly, given the benefits of adopting the systems approach, efforts to bridge the systemic accident analysis research-practice gap should be made. However, the systemic analysis methods may not be best suited to analyse every type of accident and, therefore, should be considered as one part of an investigator s analysis toolkit . Adapting the systemic analysis methods to meet the needs of practitioners and communicating the systems approach more effectively represent two options for bridging the gap. However, due to the multidimensional nature of the gap and the wide variety of individuals, organisations and industries that perform accident analysis, it seems likely that tailored solutions will be required. Furthermore, due to the differing needs of the research and practice communities, efforts to bridge the gap should focus on collaboration between the two communities rather than attempting to close the gap entirely.

Product-oriented environmental improvement is a great challenge for SMEs. There are two main reasons for this; firstly, a contradictory situation exists because SMEs lack the knowledge and competence to adopt the mainstream principle holistic approach of environmental product policy and regulations, and, secondly, there is a growing demand to incorporate environmental management system (EMS) and product dimension via ecodesign. To address these issues, this research focused on a product-oriented environmental management system (POEMS) that has been specifically developed to solve these problems. However, POEMS studies so far lack methodological development and focus more on how to incorporate EMS and eco-design rather than how to interpret the contents and improve POEMS as a standard tool. This research aimed to develop a new model, namely, component-based POEMS (CPOEMS). The C-POEMS model is a first attempt to define a potential standardised form of POEMS in its contents and structure, and form the basis for a useable self-help format for SMEs. A C-POEMS model, comprising stages of process, categories of functional areas/units, and elements, and a diagnosis template, as a component-based application, are developed and formulated. Through primary research, the C-POEMS conceptual model was verified and applicability of a C-POEMS component to SMEs was validated. C-POEMS made major contributions of two aspects: (1) methodological development of POEMS; (2) improvement of the format applicable for SMEs incorporating mainstream principles. Regarding methodological development of POEMS, arguments of existing POEMS methods were revealed, and the suggestions as well as analytical information in this research would provide benefit for further research in this field. The C-POEMS improved POEMS by providing clear contents and structure with predefined prior and iii correlated categories of functional area/units, which are helpful for SMEs. In addition, a diagnostic approach would help SMEs recognise their own problems and focus areas. As a result, SMEs can reduce the initial time to identify a structure of product-related environmental management, and increase opportunities to focus on major targets and product aspect for environmental improvements. Because of time constraint and underdevelopment of POEMS, there remain problems, in particular, to fully integrate EMS and eco-design, and the asymmetric situation between maintaining the broad scope of POEMS and focusing upon specific areas and user demand. However, this C-POEMS model provides a foundation for the development of EMS for SME’s and for ongoing development of POEMS.

People-centred design is a design approach that takes the intended end users into consideration throughout the development process, resulting in more appropriate design output in terms of meeting peoples needs and wants. There is recognised value in the use of user-based information, and in establishing empathy with those being designed for, yet there is a distinct lack of literature addressing both aspects and the potential for associated support mechanisms for designers. The combination of information and empathy is the focus of this research. This thesis presents studies carried out to investigate the potential for ‘supporting people-centred design through information and empathy’, focusing upon the early stages of design development. The main aims of this research were to understand designers’ processes and where users fit into these, and to suggest means of support that could promote user focus whilst remaining a practical and appropriate complement to established methods. The under-explored nature of this area required empirical research engaging in practical ways with designers, which was achieved through in-depth probe studies and follow-up interviews with 10 designers; active participation in two four-month real-life design projects; the examination and co-creation of resource tool concepts during two workshops, each with 20 design participants; and ‘MHIRROR’ (Means of Human Information Retrieval, Representation, Organisation and Reflection), a mixed media human information resource was developed and trialed with six experienced inclusive design practitioners. These qualitative explorations with designers and within real-life projects facilitated understanding of the potential for human information resources to support the design process. The thesis has made original contribution to knowledge in terms of the formation of a framework for the manipulation and integration of human information into the design process; the iterative design and embodiment of a working prototype resource MHIRROR, and it has provided insights into the value of information and empathy resource combinations and their potential to promote people-centred design.

Empirical knowledge is lacking on reclined seating postures. To unify such data, a biomechanical model is needed that accurately predicts posture, the relative position of the pelvis, the point of load transfer to the seat, internal and external forces, and the motion paths of the support surfaces. The overall aim of this investigation was, therefore, to create and validate a biomechanical model of reclined seating postures, and to evaluate in vivo measured and predicted data. A two-dimensional biomechanical model was developed, validated and applied. A comprehensive set of biomechanical data was collected from fifteen gender and age diverse subjects to examine the foundational principles for reclined seating ergonomics. The model agreed with 98.8% of measured data on posture across the seated test conditions. There was a significant relationship between modelled and measured force (p < .001, r = .92), which improved after normalisation (p < .001, r = .97) with an 8% full scale error. The model was robust across height and gender. Significant differences in interface pressure (peak pressure, average pressure and area), stature, back muscle activity and spinal curvature were found between all of the seated test postures. Significant relationships were found between the model predictions and all of the experimental data. This research is unique in creating a framework around reclined seating postures which connects previously disparate areas of seating research. The biomechanical model, experimental results, and theories developed from this research have potential implications in research, and design, for applications including backcare chairs, seating for long-term care and patients with neuromotor deficits, wheelchairs and airline seating. Furthermore, this study exists at the interface of anthropometric and biomechanical modelling, and therefore may have cross over potential to digital humans, where their integration with biomechanical models is at the cutting edge of the field.

Cette recherche en ergonomie, aborde l’étude du travail au travers de la notion du « prendre soin ». Elle porte sur le processus de production coopérative du soin et plus particulièrement sur le soutien de la gestion du flux de travail, ou workflow, dans la production du soin radiothérapeutique. Cette recherche s’est déroulée dans le cadre d’un projet d’innovation industrielle pour la conception d’un outil informatique de workflow en radiothérapie. Ce traitement contre le cancer est conçu par un collectif transverse dans un processus comprenant plusieurs étapes qui présente des exigences de coopération. L’objectif de cette thèse est d’apporter des éléments de compréhension sur le « prendre soin » du travail à partir de l’étude de la gestion du flux de travail. Des observations systématiques dans quatre centres de radiothérapie français permettent de modéliser des dispositifs de gestion de flux de travail (DGF) en situation nominale. Dans un des centres, 32 entretiens portant sur la reconstitution du DGF local à partir de traces de l’activité, permettent de comprendre la gestion du flux de travail en situation d’exception (incidentelle ou de variabilité). Les résultats mettent en évidence une « invention organisationnelle » issue de la « mise en main » et la mobilisation d’un environnement physique sémiotisé dans le soutien de la gestion collective du flux de travail. Par la mobilisation du DGF en situation nominale les professionnels visent la production de traitements sûrs et efficaces ainsi que le « prendre soin » du travail d’articulation. En situation d’exception, des pratiques de régulation émergente permettent de produire des traitements sûrs et efficaces et de « prendre soin » du patient, du travail de l’autre et de son propre travail.Les retombées de notre travail portent sur le développement et l’outillage du processus du flux de travail et la conceptualisation de la qualité en milieu du soin. Ainsi, la qualité du processus de production du soin est composée des dimensions d’efficacité et de sécurité du traitement (cure) ainsi que de la dimension du « prendre soin » (ou care) des patients, du travail coopératif et plus largement, du travail collectif. Enfin, cette recherche permet de penser les dimensions du « prendre soin » comme partie intégrante de la qualité du travail individuel et collectif. / This research in ergonomics addresses the "taking care" dimension of work. It deals with the cooperative processes involved in healthcare production, and more particularly with supporting the workflow in radiotherapy. The research takes place within an industrial innovation project aiming at developing a workflow software. Radiotherapy is a treatment against cancer by the administration of ionizing radiation. This type of treatment is produced by a transverse collective in a multi-step process with cooperation requirements. The objective of this thesis is to better understand the “taking care” dimension of work by analyzing workflow management. Systematic observations in four French radiotherapy services have allowed the modelling of Workflow Management Devices (WMD) in nominal situations. In a radiotherapy center, 32 interviews were conducted with professionals, by using a restitution of the local WMD composed by traces of activity. This helped to understand the professionals’ activity in exceptional situations (non nominal). The results highlight a “tailored”, “semioticized” physical environment, that supports the collective workflow management in order to ensure a safe and efficient treatment and to “take care” of the articulation work. This environment supports the normative model of the radiotherapy process. Furthermore, in exceptional situations, emerging regulation strategies help to produce safe and efficient treatments and to take care of the patient, the work of colleagues and of one’s own work.The impact of our work concerns the development of workflow processes and tools and the conceptualization of quality in healthcare. Healthcare quality integrates the safety and effectiveness of treatments (cure) as well as the “taking care”, not only of patients, but of cooperative and collective work. Finally, this research helps in conceptualizing that “taking care” of cooperative work is to be considered as an additional dimension of the quality of individual and collective work.

With more legislation being enforced to achieve a reduction in road transport CO2 emissions, automotive companies are having to research and develop technologies that deliver greener driving . Whilst emissions from passenger vehicles have dropped over recent years, there has been an increase in emissions from light commercial vehicles (LCVs). The nature of LCV delivery work is a routine of ingress/egress of the vehicle, changing from a standing to a seated posture repetitively throughout the day. One research focus is packaging occupants in to a smaller vehicle space, in order to reduce the amount of vehicle emissions over its lifecycle. For LCVs, benefits from space saving technology could be an increase in overall loading space (with the same vehicle length) or a reduction in the overall length/weight of the vehicle. Furthermore, an elevated seat posture could reduce the strain on drivers during ingress/egress, as it is closer than that of a conventional seat to a standing posture. Whilst space saving technology has obvious benefits, current driving conventions and standards are not inclusive of new and novel seated postures when packaging a driver in to a vehicle. The fundamental purpose of a vehicle driver s seat is to be comfortable and safe for the occupant and to facilitate driving. It has been shown that a seat needs both good static and dynamic factors to contribute to overall seat comfort. Additionally, comfortable body angles have been identified and ratified by studies investigating comfortable driving postures; however, this knowledge only applies to conventional driving postures. For an elevated posture , defined as having the driver s knee point below the hip point, there is little research or guidance. The overall aim of this thesis is to identify the ergonomic requirements of a wide anthropometric range of drivers in an elevated driving posture for LCVs, which was investigated using a series of laboratory based experiments. An iterative fitting trial was designed to identify key seat parameters for static comfort in an elevated posture seat. The results showed that in comparison with a conventional seat: Seat base length was preferred to be shorter (380mm compared with 460mm); Seat base width was preferred to be wider (560mm compared with 480mm); Backrest height was preferred to be longer (690mm compared with 650mm). These findings provided a basis for a seat design specification for an elevated posture concept seat, which was tested in two subsequent laboratory studies. A long-term discomfort evaluation was conducted, using a driving simulator and a motion platform replicating real road vibration. Discomfort scores were collected at 10-minute intervals (50-minutes overall) using a body map and rating scale combination. The results indicated that in comparison with the conventional posture, the elevated posture performed as well, or better (significantly lower discomfort for right shoulder and lower back; p<0.05, two-tailed), in terms of long-term discomfort. Furthermore, the onset of discomfort (i.e. the time taken for localised discomfort ratings to be significantly higher than the baseline ratings reported before the trial) occurred after as little as 10 minutes (conventional posture) and 20 minutes (elevated posture) respectively. A lateral stability evaluation was conducted using low-frequency lateral motion on a motion platform (platform left and right rolls of 14.5°). Stability scores were reported after each sequence of rolls, comparing scores on a newly developed lateral stability scale between three seats: Conventional posture seat; Elevated posture concept seat (EPS1); Elevated posture concept seat with modifications aimed at improving stability (EPS2). Participants reported being more unstable in EPS1, compared with the conventional posture seat (p<0.05, Wilcoxon). However, the EPS2 seat performed equally to the conventional posture seat. These findings suggest that the elevated posture seat developed in this research is a feasible and comfortable alternative to a conventional posture seat. Furthermore, the final elevated seating positions showed that real space saving can be achieved in this posture thus allowing for more compact and lighter vehicles and potentially reducing strain on drivers during ingress/egress.

L'analyse ergonomique des postes de travail reste le point de départ de toute politique de prévention des risques de maladies professionnelles. De nombreux travaux scientifiques s'attachent à quantifier les déterminants à risque pour aboutir le plus souvent à un score de pénibilité. La difficulté actuelle des méthodes de cotation ergonomique se situe au niveau de la capture de ces déterminants. La majorité des systèmes se limitent à une collecte de données souvent subjective et très influencé par la personne effectuant la cotation. La volonté de l'entreprise par le biais de ce stage est d'objectiver l'analyse ergonomique des postes de travail par une capture du mouvement de l'opérateur. Le principale défi est le passage d'outil et de méthode scientifique à une utilisation de terrain avec toutes les contraintes qu'elle induit.L'apport des avancées technologiques et scientifiques encourage ce passage par des outils utilisable dans le contexte industriel. Les deux principaux objectif de ce stage se sont situer premièrement sur la limitation des biais de capture pour amener un précision et un standardisation de la mesure de terrain, ainsi que sur l'accès à de nouvelles données notamment l'aspect temporel de la tâche effectuée. Le matériel utilisé est le capteur de profondeur Kinect développé par Microsoft. Cet appareil fait l'objet d’études scientifique dans différents domaines et plus particulièrement dans son utilisation pour de la capture de mouvements.Lors de ce stage, nous nous somme attaché à traiter le signal émis par la Kinect, pour obtenir des données permettant le remplissage automatique de grille de cotation. Le bruit de mesure fût travailler à l'aide d'un filtre récursif passe bas utilisé fréquemment en laboratoire d'analyse du mouvement. Le traitement des données spatiales brut des articulations de l'opérateur pour obtenir des angles fit l'objet d'un grande partie du travail, car ne nombreux paramètre entre en jeu comme la position du capteur.La réussite du stage à permis de limiter la subjectivité de la mesure mais à également donnée l'accès à de nouveaux indices comme les pourcentage de temps de cycle passé à des angulations dangereuses pour l'opérateur. Le passage d'outil de laboratoire au terrain mérite encore d'être travaillé notamment dans la robustesse des systèmes développés et doit s'appuyer sur des expérimentations de laboratoire. / Evaluation of potential risks of musculoskeletal disorders in real workstations is challenging as the environment is cluttered, which makes it difficult to correctly and accurately assess the pose of a worker. Most of the traditional motion capture systems cannot deal with these workplace constraints. Being marker-free and calibration-free, Microsoft Kinect is a promising device to assess these poses, but the validity of the delivered kinematic data under work conditions is still unknown. In this thesis we first propose an extensive validation of the Kinect system in an ergonomic assessment context with sub-optimal capture condition. As most of the large inaccuracies come from occlusions, we propose a new example-based method to correct unreliable poses delivered by the Kinect in such a situation. We introduced the Filtered Pose Graph structure to make the method select the most relevant candidates before combination. In an ergonomics context, we computed RULA scores and compared them to those computed from an optoelectronic mocap system. We also propose to challenge our method in real workplace environment and compared its performance to experts' evaluation in the Faurecia company. Finally, we evaluated the relevance of the proposed method to estimate internal joint torques thanks to inverse dynamics, even if occlusions occur. Our method opens new perspectives to define new fatigue or solicitation indexes based on continuous measurement contrary to classical static images generally used in ergonomics. The computation time enables real-time feedback and interaction with the operator.