Les méthodes de Monte Carlo et Dynamique Moléculaire ont été utilisées pour étudier l’argile hydratée de type Wyoming-montmorillonite. Nous avons examiné le comportement du gonflement et du dégonflement de la montmorillonite sous l'effet de la température, de le teneur en eau et de la nature des différents cations. Un phénomène d’hystérésis a été constaté pendant le processus d’hydratation/déshydratation ainsi que durant le processus de chauffage/refroidissement. La quantité de cations et leur capacité d’attraction ont un rôle important dans ce comportement du gonflement et du dégonflement de l’argile. En outre, l’étude de la diffusion de l’eau et des cations dans l’espace interlamellaire de la montmorillonite montre l’effet du degrés d’hydratation et de la nature des cations (en particulier les polluants) sur la diffusion de l’eau, ainsi que l’effet de la température sur la diffusion des particules présentes dans les inter-couches. Le bon accord de nos résultats de simulation concernant l'argile contenant Li+, Na+, K+ et Ca2+ avec les résultats expérimentaux et d’autres travaux théoriques nous a permis d'étendre notre étude à l'argile comportant des ions métalliques lourds, comme Rb+, Cs+, Ni2+, Zn2+ et Pb2+.Enfin l’étude détaillée de l’effet de la pression sur la montmorillonite montre la variation de la rigidité en fonction de la teneur en eau et permet de prédire des transitions de phase structurales inférieures à 1 GPa. Cette étude montre particulièrement que le module de rigidité peut atteindre des valeurs extrêmement élevées à des pressions relativement modestes d’environ 20 GPa. Ce résultat ouvre clairement la voie pour une nouvelle famille de matériaux ultra-incompressible, et éventuellement ultra dures, composés d’argiles. Ce type de matériaux aux propriétés mécaniques spécifiques et de faibles coûts (abondance des argiles), peut être particulièrement attractif pour de nombreuses applications dans divers domaines. / Monte Carlo and Molecular Dynamic simulation methods have been performed to study hydrated Wyoming-type montmorillonite, a naturally occurring swelling clay, including different monovalent or divalent counterions. Simulations have been also performed to investigate the swelling and shrinking behaviour of montmorillonite under the influence of water content, temperature and different counterions. We found a general phenomenon of hysteresis during both the hydrate/dehydrate and the warming/cooling processes. The quantities of counterions together with their solvation characteristics and their attraction capacities have an important influence on the swelling and shrinking behaviours of this clay. Moreover, investigations on the diffusion behaviours of water and counterions in the interlayer space of montmorillonite show the effect of hydrated degree and different existing cations especially the pollutants on the diffusion of water, as well as the effect of temperature on the diffusion of interlayer particles. The agreement of our simulation results on the clay containing some common cations such as Li+, Na+, K+ and Ca2+, with available experiment and preceding theoretical works, permitted us to extend our study to the clay with heavy metal ions, such as Rb+, Cs+, Ni2+, Zn2+ and Pb2+, some of which have never been studied.Finally, detailed studies on the shrinking behaviours of hydrated montmorillonite under pressure allowed us to shed light on the rigidity of this clay affected by the hydrated degree, where we found a phase transition phenomenon in hydrated montmorillonite with the phase transition pressure less than 1GPa. According to our calculations the bulk modulus reaches extremely high values at modest pressures of ~20 GPa. It would open a route to a new family of ultra-incompressible materials of clay composition. Such materials, and their superior mechanical properties and possibly low cost, could be practically attractive for a number of uses.

High-resolution tower observations of turbulent transport processes in the coastal atmospheric surface layer show that the exchange coefficients for momentum, enthalpy, and moisture behave differently for different environmental and atmospheric conditions. The drag coefficient is closely tied to wind speed and turbulent intensity. The exchange coefficient for enthalpy shows a dependence on stability. Analysis of the turbulent kinetic energy budget yields a new parameterization framework that well explains the observed variation of the drag coefficient, particularly at low wind speeds.

There have been numerous studies on measuring protein-protein interactions in solution using a variety of techniques including membrane osmometry, sedimentation, and static light scattering. Most of these techniques yield an osmotic second virial coefficient. The osmotic second virial coefficient has been shown to be an important parameter for protein crystallization. To date, there have been few fundamental theoretical studies of estimating second virial coefficient values using conventional models because of the diverse and complex nature of the potential of mean force. In the present study, the variation of equine serum albumin interactions was measured with respect to pH and sodium chloride salt concentration by static light scattering to determine the second virial coefficient and electrophoretic light scattering to determine the electrophoretic mobility. The main aim is to show the effect of the solution conditions such as pH and ionic strength on ESA interactions. In this thesis, I will focus on the classical theory of Derjaguin, Landau, Verwey and Overbeek (DLVO). The experimental data from electrophoretic and static light scattering measurements for equine serum albumin are compared with the DLVO model.

Rock weirs are small dam-like structures composed of large loose rock commonly found in ecological engineering design. By appearing more natural than concrete structures, rock weirs are preferred for use as hydraulic control structures in river engineering, stormwater management, and constructed wetlands. Rock weirs increase hydraulic head upstream, and facilitate fish passage, channel stabilization, floodplain reconnection, and in-stream habitat creation. When used in constructed wetlands, rock weirs play a valuable role in developing appropriate wetland hydrology. Although rock weirs are commonly used, a deficit of knowledge exists relating to the stage-discharge relationship of these structures. Therefore, the goal of this research was to determine a weir equation and corresponding discharge coefficients that improve predictions of flow through rock weirs. A flume study was conducted to develop a rock weir equation and discharge coefficients. Scaled model rock weirs were tested in a 1 m x 8 m x 0.4 m recirculating flume. Rock weirs varied by length (0.152 m, 0.305 m, and 0.457 m), depth (0.152 m and 0.305 m), and minimum rock diameter (12.7 mm, 19.1 mm, 25.4 mm). Three channel slopes were used (0%, 0.5%, 1%), and the flume discharge was varied for five water stages for each rock weir. Buckingham Pi analysis was used to develop seven dimensionless parameters. Regression analyses were then used to develop a model for discharge and the discharge coefficient. Results showed that weir length and depth play a significant role in predicting the discharge coefficient of rock weirs. / Master of Science / Rock weirs are small dam-like structures composed of large loose rock; by appearing more natural, they are preferred for use in river engineering, stormwater management, and constructed wetlands. Rock weirs increase upstream water depth, improving fish passage, channel stabilization, floodplain reconnection, and in-stream habitat creation. When used in design of constructed wetlands, rock weirs are used to establish the necessary water depths for a given type of wetland. Although rock weirs are commonly used in engineering design, there are no equations to predict water velocity or flow rate across these structures. Therefore, the goal of this research was to determine a weir equation that improves predictions of flow through rock weirs. A flume study was conducted to develop a rock weir equation. Miniature rock weirs were tested in a 1 m x 8 m x 0.4 m recirculating laboratory channel. Rock weirs varied by length (0.152 m, 0.305 m, and 0.457 m), depth (0.152 m and 0.305 m), and minimum rock diameter (12.7 mm, 19.1 mm, 25.4 mm). Three channel slopes were used (0%, 0.5%, 1%), and the water flow rate was varied for five water depths for each rock weir. Statistical analyses were conducted to determine an equation that predicts water flow through rock weirs for use in engineering design. Results showed that weir length and depth played a significant role in predicting water flow through rock weirs.

Erosion and deposition in gas turbine engines are functions of particle/wall interactions and the Coefficient of Restitution (COR) is a fundamental property of these interactions. COR depends on impact velocity, angle of impact, temperature, particle composition, and wall material. In the first study, a novel Particle Tracking Velocimetry (PTV) / Computational Fluid Dynamics (CFD) hybrid method for measuring COR has been developed which is simple, cost-effective, and robust. A Laser-Camera system is used in the Virginia Tech Aerothermal Rig to measure microparticles velocity. The method solves for particle impact velocity at the surface by numerical methods. The methodology presented here characterizes a difficult problem by a combination of established techniques, PTV and CFD, which have not been used in this capacity before. The current study characterizes the fundamental behavior of sand at different impact angles. Two sizes of Arizona Road Dust (ARD) and one size of Glass beads are impacted on to 304-Stainless Steel. The particles are entrained into a free jet of 27m/s at room temperature. Mean results compare favorably with trends established in literature. This technique to measure COR of microparticle sand will help develop a computational model and serve as a baseline for further measurements at elevated, engine representative air and wall temperatures. In the second study, ARD is injected into a hot flow field at temperatures of 533oK, 866oK, and 1073oK to measure the effects of high temperature on particle rebound. The results are compared with baseline measurements at ambient temperature made in the VT Aerothermal Rig, as well as previously published literature. The effects of increasing temperature and velocity led to a 12% average reduction in total COR at 533oK (47m/s), a 15% average reduction at 866oK (77m/s), and a 16% average reduction at 1073oK (102m/s) compared with ambient results. From these results it is shown that a power law relationship may not conclusively fit the COR vs temperature/velocity trend at oblique angles of impact. The decrease in COR appeared to be almost entirely a result of increased velocity that resulted from heating the flow. / Ph. D.

The topographical variations of the seas around Taiwan are great, which make the tides complicated. Taiwan is located in the juncture of the tropical and subtropical area. Geographically, it is located within the region of northwestern Pacific typhoon path. These seasonal and geographical situations causing Taiwan frequently threaten by typhoons during summer and autumn. In addition to natural disasters, the coastal area is over developed for the last few decades, which destroys the balance between nature and man. Storms and floods constantly threaten the lowland areas along the coast. An accurate and efficient storm surge model can be used to predict tides and storm surges. The model can be calibrated and verified with the field observations. Data measured by instruments at the tidal station constituting daily tidal variations and storm surge influences during typhoons. The model can offer both predictions to the management institutions and to the general public as pre-warning system and thus taking disaster-prevention measures. This study implements the numerical model, developed by Yu (1993) and Yu et al. (1994) to calculate the hydrodynamic in the seas around Taiwan. The main purpose of this study is to make a calibration and sensitivity analysis of the model parameters. Tidal gauge data around Taiwan coastal stations collected from June to October 2005 are used for the analysis and the comparison between the modeled data and the observations. Two steps have been taken for the model calibration and sensitivity analysis. First step is to calibrate the model for accurate prediction of the astronomical tide, and then the compound tide with meteorological influences. For the calibration of the astronomical tides, sensitivity analysis has been carried out by adjusting the horizontal diffusion coefficient and the bottom friction coefficients used in the model. The sensitivity of the time-step size used in the model and model grids fitted to coastlines are also checked. A depth dependent Chézy numbers are used in the model to describe bottom friction. The model has a better result when the Chézy value varied within 65 to 85. Modifying grids fitted to the coastline has improved the model results significantly. By improving the dynamic phenomenon brought about by the land features, the model calculation fits the real tidal phenomenon better. The analysis has shown that the model is less sensitive to the horizontal diffusion coefficient. Data from 22 tidal stations around Taiwan have been used for the comparisons. The maximum RMSE (root-mean-square error) is about 10 cm at WAi-Pu, whereas the minimum RMSE is about 1 cm for the stations along eastern coast. The calibration of the compound tide is divided into three cases. The first case is to calibrate the forecasted wind field. This has been done by comparing the forecasted wind field from the Central Weather Bureau with the satellite data obtained from QuikSCAT¡XLevel 3. The satellite wind speed has been applied to adjust the forecasted wind speed. The adjusted forecast wind field has shown improvement to the model predictions in the tidal stations south of Taichung, slightly improved in the eastern coast. The second case is tuning the drag coefficient on sea surface used by the hydrodynamic model. Several empirical formulas to describe the sea surface drag have been tested. The model result has shown little influence using various drag formulations. The third case is to single the influences by the meteo-inputs, i.e. the wind field and the atmospheric pressure. The tidal level is more sensitive to the variation of the atmospheric pressure through out the tests carried out during typhoon periods. The model simulation for 2006 using the best selected parameters has shown that the model is consisted with good stability and accuracy for both stormy and calm weather conditions.

Approved for public release; distribution is unlimited. / Ever-increasing demands for accuracy and range in modern warfare have expedited the optimization of projectile design. The crux of projectile design lies in the understanding of its aerodynamic properties early in the design phase. This research first investigated the aerodynamic properties of a standard M549, 155mm projectile. The transonic speed region was the focus of the research as significant aerodynamic variation occurs within this particular region. Aerodynamic data from wind tunnel and range testing was benchmarked against modern aerodynamic prediction programs like ANSYS CFX and Aero-Prediction 09 (AP09). Next, a comparison was made between two types of angle of attack generation methods in ANSYS CFX. The research then focused on controlled tilting of the projectile’s nose to investigate the resulting aerodynamic effects. ANSYS CFX was found to provide better agreement with the experimental data than AP09.

Thesis (MEng)--Stellenbosch University, 2015. / ENGLISH ABSTRACT: Bioconversion of hydrocarbons to value-added intermediates and products has significant industrial potential using both prokaryotic and eukaryotic organisms. In particular, alkanes can be converted to an expansive range of commercially important products using aerobic bioprocesses under mild process conditions. Coupled with the relative abundance of alkanes derived from gas to liquid (GTL) technologies, such as those employed by SASOL, South Africa, the commercial potential for bioconverison of alkanes is large. However, unlike carbohydrate substrates, alkane feedstocks are devoid of oxygen in their molecular structure. This means that the entire oxygen demand needs to be met by oxygen transfer. Furthermore, a decline in oxygen transfer in aqueous-hydrocarbon dispersions with increasing alkane concentration has been observed to result from depression of the overall volumetric oxygen transfer coefficient (KLa). Therefore, understanding KLa and the fundamental parameters underpinning its behaviour is critical to ensuring the bioprocess is kinetically, rather than transport, limited in terms of both operation and scale-up. Previous studies have examined KLa in aerated-alkane-aqueous systems. In light of the importance of oxygen transfer in bioprocesses, this study expands on the KLa understanding in 3-phase studies by including a fourth solid phase, thus more closely representing a hydrocarbonbased bioprocess. The project aimed to determine the impact of agitation, alkane concentration and solid loading on the Sauter mean bubble diameter (DSM), gas hold-up and specific interfacial area (a) and correlate these parameters to KLa. This ultimately determined which parameter was dominant over a range of process conditions. Furthermore, concurrent measurement of the KLa and interfacial area meant the behaviour of the liquid side oxygen transfer coefficient (KL) could be defined, providing further insight into how changes in the process conditions impact on KLa. Experiments were conducted in a 5 litre stirred tank bioreactor containing n-C14-20 straight chain alkane, sparged with air at 0.8 vvm. In line with process conditions typical of a hydrocarbonbased bioprocess, KLa and a were measured for agitation rates from 450 to 1000 RPM, alkane concentrations from 2 to 20% v/v and yeast solids from 1 to 10 g/l. KLa was measured using the gassing out procedure using a dissolved oxygen (DO) probe which measured the response of the system to a step change in the sparge gas oxygen pressure. The probe response lag ( P), equal to the time taken for the probe to reach 63.2% of the saturation DO concentration, was determined for every set of process conditions. The inverse of P, KP was taken into account when calculating KLa from the DO probe response. The area was calculated from DSM and gas hold-up. DSM was quantified using high speed photography and image analysis was performed in Matlab® using bespoke routines. Elimination of optical distortion and the development of an adequate light source was key to acquiring clear images. Both KLa and interfacial area were found to be affected by changes in agitation, alkane concentration and yeast loading. An increase in agitation increased the KLa over the entire range of alkane concentration and yeast loading. Similarly, an increase in agitation resulted in an increase in interfacial area, underpinned by a decrease in the DSM. It is therefore likely that the interfacial area plays a dominant role in defining KLa when considering an increase in agitation. Increases in alkane concentration resulted in a peak in KLa between 2.5 and 5% alkane concentration while further increases in alkane concentration depressed KLa. This peak was not observed in interfacial area, where an increase in alkane concentration resulted only in a decrease in interfacial area, thus indicating a positive influence of KL on KLa at low alkane concentrations. Further increases in alkane concentration beyond those creating the peak KLa resulted in KLa depression, suggesting that the increasing viscosity imparted by the alkane decreases both KL and interfacial area. Increased yeast loading had opposing effects at low and high agitation rates. At low agitation rates, increased loadings were observed to increase KLa, while increased loadings at high agitation rates caused a decrease in KLa. This behaviour was also evident in interfacial area, suggesting that in this regime KLa was defined by interfacial area behaviour. Increased yeast loading was observed to depress the KLa for all alkane concentrations when examined at a constant midpoint agitation rate. This trend was not evident in interfacial area, which increased with increasing yeast loading at the same agitation rate. The positive influence of yeast on interfacial area was likely caused by adhesion of the yeast particles to the bubble surface, lowering the DSM by preventing coalescence. The disagreement between the KLa and interfacial area results suggested that yeast loading impacted negatively on KL, which had an over-riding negative impact on KLa. The use of reliable methods for the determination of both interfacial area and KLa were demonstrated for application in model hydrocarbon-based bioprocesses. The combined results offer a unique insight into how changes in the process conditions impact independently on KL and interfacial area, which when combined ultimately defined the KLa behaviour. Quantification of the relative magnitude of the impact each parameter had on KLa contributed toward a fundamental understanding of oxygen transfer in model hydrocarbon-based bioprocesses. / AFRIKAANSE OPSOMMING: Biologiese omsetting van koolwaterstowwe na produkte met finansiële waarde het beduidende industriële potensiaal met behulp van beide prokariotiese en eukariotiese organismes. In die besonder, kan alkane omgeskakel word na ’n uitgebreide reeks van kommersieel belangrike produkte met behulp van aerobiese bioprosesse onder ligte proses voorwaardes. Tesame met die relatiewe oorvloed van alkane afgelei van GTL tegnologie, soos dié van Sasol, Suid-Afrika, die kommersiële potensiaal vir bioconverison van alkane is groot. Maar, in teenstelling koolhidrate substrate, alkaan voerstowwe is beroof van suurstof in hul molekulêre struktuur. Dit beteken dat die hele suurstof vereiste moet nagekom word deur suurstof oordrag. Verder het ’n afname in suurstof oordrag in waterige-koolwaterstof dispersies met toenemende alkaan konsentrasie waargeneem te lei van depressie van die algehele volumetriese suurstofoordragkoëffisiënt (KLa). Daarom verstaan KLa en die fundamentele parameters onderliggend sy gedrag is van kritieke belang om te verseker dat die bioprocess is kineties, eerder as vervoer, beperk in terme van beide werking en skaal-up van bioprosesse. Vorige studies het KLa in deurlug-alkaan-waterige stelsels ondersoek. In die lig van die belangrikheid van suurstof oordrag in bioprosesse hierdie studie brei uit op die KLa begrip in driefase studies deur die insluiting van ’n vierde soliede fase, dus meer nou wat ’n koolwaterstofgebaseerde bioprocess. Die doel van die projek is om die impak van vermengingstempo, alkaan konsentrasie en soliede inhout op die Sauter gemiddelde borrel deursnee (DSM), gas-vasvanging en spesifieke gas-vloistof oppervlakarea (a) te kwantifiseer en korreleer met KLa gedrag. Dit sou defineer die dominante parameter oor ’n verskeidenheid van proses voorwaardes. Verder, gelyktydige meting van die KLa en oppervlakarea kan die gedrag van die vloeistof-kant suurstofoordragkoëffisiënt (KL) gedefinieer. Dit sal verskaf verdere insig in hoe die veranderinge in die proses voorwaardes impak op KLa. Eksperimente was uitgevoer in ’n 5 liter belugte geroerde tenk bioreaktor bevat n - C14-20 reguitketting alkane, met lug met lug deurgeborrel by 0.8 VVM. In lyn met die proses voorwaardes tipies van ’n koolwaterstof-gebaseerde bioprocess, KLa en a was gemeet vir vermengignstempos van 450-1000 RPM, alkaan konsentrasies van 2-20 % v/v en gis vastestowwe van 1 tot 10 g / l. KLa is gemeet deur die vergassinguit prosedure met behulp van ’n suurstofmeter wat die reaksie van die stelsel na ’n stap verandering in die voer gas suurstof druk gemeet het. Die suurstofmeter reaksie vertraging ( P), gelyk aan die tyd wat dit neem vir die suurstofmeter 63.2 % van die versadiging DO konsentrasie te bereik, is bepaal vir elke procesopset. Die inverse van P, KP is in ag geneem by die berekening van KLa uit die suurstofmeter reaksie. Die gas-vloistof oppervlak is bereken vanaf DSM en gas hold-up. DSM is gekwantifiseer met behulp van hoë spoed fotografie en beeld analise is uitgevoer in Matlab ® roetines. Uitskakeling van optiese vervorming en die ontwikkeling van ’n voldoende ligbron was die sleutel tot die verkryging van helder beelde. Beide KLa en grens oppervlakarea gevind geraak word deur veranderinge in vermengignstempo, alkaan konsentrasie en gis laai. ’N toename in geroer het die KLa verbeter oor die hele reeks van alkaan konsentrasie en gis laai. Net so, ’n toename in geroer het gelei tot ’n toename in grens oppervlak, ondersteun deur ’n afname in die DSM. Dit is dus waarskynlik dat die grens oppervlak speel ’n dominante rol in die definisie van KLa by die oorweging van ’n toename in roering. Stygings in alkaan konsentrasie gelei tot ’n hoogtepunt in KLa tussen 2.5 en 5 % alkaan konsentrasie terwyl verdere verhogings in alkaan konsentrasie druk die KLa af. Die piek was nie in oppervlakarea duidelik, waar ’n toename in alkaan konsentrasie gelei net tot ’n afname in oppervlakarea, dus dui op ’n positiewe invloed van KL op KLa teen lae alkaan konsentrasies waargeneem. Verdere stygings in alkaan konsentrasie verder as die skep van die piek KLa gelei tot KLa depressie, wat daarop dui dat die toenemende viskositeit meegedeel deur die alkaan verminder beide KL en grens oppervlak. Verhoogde gis laai het opponerende effekte teen ’n lae en hoë vermengingstempo. By lae vermengingstempo, ’n verhoging in gis laai waargeneem KLa te verhoog, terwyl ’n verhoging in gis laai op ’n hoë vermengingstempo veroorsaak ’n afname in KLa . Hierdie gedrag was ook duidelik in grens oppervlak, wat daarop dui dat daar in hierdie regime KLa gedefinieer deur grens oppervlak gedrag. Verhoogde gis laai waargeneem die KLa te onderdruk vir alle alkaan konsentrasies wanneer ondersoek teen ’n konstante middelpunt vermengingstempo. Hierdie tendens was nie duidelik in tussenvlak gebied, wat verhoog met toenemende gis laai op dieselfde geroer koers. Die positiewe invloed van gis op grens oppervlak is waarskynlik veroorsaak deur adhesie van die gis deeltjies aan die borrel oppervlak, die verlaging van die DSM deur die voorkoming van die saamsmelting van gasborrels. Die meningsverskil tussen die KLa en grens oppervlakarea resultate voorgestel dat gis laai negatiewe uitwerking op KL, met ’n dominante negatiewe impak op KLa. Die gebruik van ’n betroubare metodes vir die bepaling van beide oppervlakarea en KLa gedemonstreer vir toepassing in model koolwaterstof-gebaseerde bioprosesse. Die gekombineerde resultate bied ’n unieke insig in hoe die veranderinge in die proses voorwaardes impak onafhanklik op KL en oppervlakarea, wat wanneer gekombineer gedefinieer die KLa gedrag. Kwantifisering van die relatiewe grootte van die impak elke parameter het op KLa bygedra tot ’n fundamentele begrip van suurstof oordrag in model koolwaterstof-gebaseerde bioprosesse.

Critical equipment plays an essential role in industry because of its lack of redundancy. Failure of critical equipment results in a major economic burden that will affect the profit of the enterprise. Lack of redundancy for critical equipment occurs because of the high cost of the equipment usually combined with its high reliability. When we are analyzing the reliability of such equipment, as a result, there are few opportunities to crash a few pieces of equipment to actually verify component life. Reliability is the probability that an item can perform its intended function for a specified interval of time under stated conditions and achieve low long-term cost of ownership for the system considering cost alternatives. From the economical standpoint, the overriding reliability issue is cost, particularly the cost of unreliability of existing equipment caused by failures. Classical questions about reliability are: · How long will the equipment function before failure occurs? · What are the chances that a failure will occur in a specified interval for turnaround? · What is the best turnaround interval? · What is the inherent reliability of the equipment? · What are the risks of delaying repair/replacements? · What is the cost of unreliability? · … We will try to answer these questions for a critical reciprocating compressor, which has been in service for only 4 years and has undergone only few failures. Professionals in all industries are faced with the problems of performing maintenance actions and optimizing maintenance planning for their repairable systems. Constructing stochastic models of their repairable systems and using these models to optimize maintenance strategies require a basic understanding of several key reliability and maintainability concepts and a mathematical modeling approach. Therefore, our objective is to present fundamental concepts and modeling approaches in the case of a critical reciprocating compressor. We developed a stochastic model not to simulate a reciprocating compressor with a complete set of components but mainly to optimize the overhaul period taking into account the main failure modes only. How to lower the cost? How to reduce or remove maintenance actions that are not strictly necessary? How to improve the long-term profitability of ageing plants with the strict respect of Health-Safety-Environment HSE requirements? A reciprocating compressor is a complex machine that cannot be described with a single reliability function. A compressor has several failure modes. Each failure mode is assumed to have its own Weibull cumulative distribution function. The compressor is then a system with several Weibull laws in series. We will extend the usual procedure for minimizing the expected total cost to a group of components. Different components may have different preventive maintenance “needs”, but optimizing preventive maintenance at the component level may be sub-optimal at the system level. We will study also the reliability importance indices that are valuable in establishing direction and prioritization of actions related to a reliability improvement plan, i.e. which component should be improved to increase the overall lifetime and thus reduce the system costs. When considering a large system with many items that are maintained or replaced preventively, it is advantageous to schedule the preventive maintenance in a block such that the system downtime is kept as small as possible. This requires that the resources are available so that the maintenance of components can be performed simultaneously or according to a well-defined sequence. The result of the stochastic model optimization came as a surprise. We thought to find a new mean-time-between-failure MTBF, larger than the actual overhaul period. Actually, the model showed that there is no economical interest to schedule a systematic preventive maintenance for this reciprocating compressor. Nevertheless, we cannot wait for a failure (and the associated corrective maintenance) because the loss-of-production cost is too high and this compressor has no spare. Preventive maintenance is not the optimum strategy, but predictive maintenance is. But what means predictive maintenance? It is a maintenance policy to regularly inspect equipment to detect incipient changes or deterioration in its mechanical or electrical condition and performance. The idea behind this is to perform corrective maintenance only when needed, before the occurrence of failure. We need to find how to detect performance deterioration of the compressor with a couple of weeks or days notice before failure. So it is possible to schedule a right maintenance activity at the optimum moment. To summarize, the main findings of this thesis are · a new method to estimate the shape factor of a Weibull distribution function, · a stochastic model demonstrating that we have to move from systematic preventive maintenance to predictive maintenance, · a low cost system based on thermodynamic approach to monitor a reciprocating compressor, · an automatic detection of performance deterioration.

No description available.