Dans l’optique de produire de maîtriser les conditions d’expansion ex vivo de greffons hématopoïétiques produits à partir de sang périphérique en homéostasie, l’objectif de ce projet a été de caractériser fonctionnellement, métaboliquement et transcriptomiquement les cellules souches hématopoïétiques (CSH). Compte tenu de l’impossibilité technique de sélectionner spécifiquement les CSH humaines, nous avons utilisé un modèle cellulaire enrichi en CSH, le modèle Side Population(SP). Dans un premier temps, nos travaux ont confirmé que les CSH du sang périphérique étaient majoritairement dans la population SP et qu’elles possédaient des caractéristiques fonctionnelles proches des CSH des autres compartiments hématopoïétiques. Nous avons également démontré l’implication des basses concentrations d’O2 sur le maintien des CSH du sang périphérique. Dans un second temps, nos résultats ont prouvé que les CSH du sang périphérique utilisaient à la fois la glycolyse et la phosphorylation oxydative pour produire l’énergie nécessaire à leur maintien. Enfin, ce projet a permis d’apporter des résultats préliminaires concernant les régulations transcriptomiques des CSH du sang périphérique. Ces données montrent donc que le sang périphérique en homéostasie pourrait constituer une source potentielle de cellules pour la production de greffons hématopoïétiques tout en apportant les premiers éléments de compréhension de la physiologie de ces cellules, afin, dans un plus long terme de maîtriser leur maintien ou leur différenciation ex vivo. / To evaluate the possibility to control ex vivo expansion conditions, a key point to produce hematopoietic graft from steady state peripheral blood (SSPB), the objective of this project to characterize the functional properties, the metabolism and the transcriptomic regulations of hematopoietic stem cell (HSC) from SSPB. Due to the lack of strong HSC’s marker in human, we choose to use the Side Population (SP) model, previously described as enriched in HSC in other hematopoietic compartments. In a first part of our work, we showed that HSC from SSPB are mainly inside the SP population. Indeed, SP cells from SSPB exhibit functional properties very closed from HSC. In addition, we found they strongly affected by low O2 concentrations, as HSC from bone marrow. In a second part, our results showed that HSC from SSPB use as much glycolysis as oxidative phosphorylation to produce energy they need to maintain their properties. All together, these data give some interesting information about HSC regulation and needs. They also suggest that HSC from SSPB could be considering as a potential source of hematopoietic graft for therapy.

Side-channel analysis (SCA) is a prominent tool to break mathematically secure cryptographic engines, especially on resource-constrained devices. SCA attacks utilize physical leakage vectors like the power consumption, electromagnetic (EM) radiation, timing, cache hits/misses, that reduce the complexity of determining a secret key drastically, going from 2¹²⁸ for brute force attacks to 2¹² for SCA in the case of AES-128. Additionally, EM SCA attacks can be performed non-invasively without any modifications to the target under attack, unlike power SCA. To develop defenses against EM SCA, designers must evaluate the cryptographic implementations against the most powerful side-channel attacks. In this work, systems and techniques that improve EM side-channel analysis have been explored, making it lower-cost and more accessible to the research community to develop better countermeasures against such attacks. The first chapter of this thesis presents SCNIFFER, a platform to perform efficient end-to-end EM SCA attacks. SCNIFFER introduces leakage localization – an often-overlooked step in EM attacks – into the loop of an attack. Following SCNIFFER, the second chapter presents a practical machine learning (ML) based EM SCA attack on AES-128. This attack addresses issues dealing with low signal-to-noise ratio (SNR) EM measurements, proposing training and pre-processing techniques to perform an efficient profiling attack. In the final chapter, methods for mapping from power to EM measurements, are analyzed, which can enable training a ML model with much lower number of encryption traces. Additionally, SCA evaluation of high-level synthesis (HLS) based cryptographic algorithms is performed, along with the study of futuristic neural encryption techniques.

With more and more ICs being used in sectors requiring confidentiality and integrity like payment systems, military, finance and health, there is a lot of concern in the security and privacy of ICs. The widespread adoption of Intellectual Property (IP) based designs for modern systems like system on chips has reduced the time to market and saved a lot of money for many companies. But this has also opened the gates for problems like product piracy, IP theft and fraud. It is estimated that billions of dollars are lost annually to illegal manufacturing of Integrated Circuits. A possible solution to this problem of IP theft is to insert small circuits which are like unique IDs that only the owner or the registered verifier will know and detect in case of any conflict. The circuits that are inserted are called watermarks and are in some cases kept very small so as to be hidden. In such cases, we would need detection schemes that work well even with very small watermarks. In this work, we use Electro-Magnetic (EM) based side-channels for the detection of watermarks. Since the 90s, Side-channel Analyses have attracted significant attention within the cryptographic community as they are able to obtain secret information from smart cards and ICs. The power side-channel analysis is a very powerful method but EM side-channels are very useful as they will not need a resistor in series to the power supply and just needs passive observation of the EM radiations emanated by the IC. This passive monitoring will be a big advantage in the case of automated watermark detection used by a verifier. In this work, we start with EM side-channel analysis on FPGA for smaller designs. We insert watermarks on a Micro-controller, Smartcard and an FPGA and detect these watermarks using EM side-channel information emanated from the Design under Test. We used environments with different levels of noise interference. We compare the watermarking application using EM side-channels and Power side-channels in these different setups. These watermarks are very small and are hard to attack or remove by an attacker through reverse engineering or side-channel information. Due to the robustness against such attacks and the easy access of EM side-channels when compared to power side-channels, the EM side-channel based watermarks will be a very good solution for the IP theft problem. EM side-channel based watermark detection supports automation which companies of IP cores can make use of. We also extended this work to EM Side-channel Trojans as the concepts are similar

Embedded devices almost involve every part of our lives, such as health condition monitoring, communicating with other people, traveling, financial transactions, etc. Within the embedded devices, our private information is utilized, collected and stored. Cryptography is the security mechanism within the embedded devices for protecting this secret information. However, cryptography algorithms can still be analyzed and attacked by malicious adversaries to steal secret data. There are different categories of attacks towards embedded devices, and the side-channel attack is one of the powerful attacks. Unlike analyzing the vulnerabilities within the cryptography algorithm itself in traditional attacks, the side-channel attack observes the physical effect signals while the cryptography algorithm runs on the device. These physical effects include the power consumption of the devices, timing, electromagnetic radiations, etc., and we call these physical effects that carry secret information side-channel leakage. By statistically analyzing these side-channel leakages, an attacker can reconstruct the secret information. The manifestation of side-channel leakage happens at the hardware level. Therefore, the designer has to ensure that the hardware design of the embedded system is secure against side-channel attacks. However, it is very arduous work. An embedded systems design including a large number of electronic components makes it very difficult to comprehensively capture every side-channel vulnerability, locate the root cause of the side-channel leakage, and efficiently fix the vulnerabilities. In this dissertation, we developed methodologies that can help designers detect and fix side-channel vulnerabilities within the embedded system design at low cost and early design stage. / Doctor of Philosophy / Side-channel leakage, which reveals the secret information from the physical effects of computing secret variables, has become a serious vulnerability in secure hardware and software implementations. In side-channel attacks, adversaries passively exploit variations such as power consumption, timing, and electromagnetic emission during the computation with secret variables to retrieve sensitive information. The side-channel attack poses a practical threat to embedded devices, an embedded device's cryptosystem without adequate protection against side-channel leakage can be easily broken by the side-channel attack. In this dissertation, we investigate methodologies to build up comprehensive side-channel resistant embedded systems. However, this is challenging because of the complexity of the embedded system. First, an embedded system integrates a large number of components. Even if the designer can make sure that each component is protected within the system, the integration of the components will possibly introduce new vulnerabilities. Second, the existing side-channel leakage evaluation of embedded system design happens post-silicon and utilizes the measurement on the prototype of the taped-out chip. This is too late for mitigating the vulnerability in the design. Third, due to the complexity of the embedded system, even though the side-channel leakage is detected, it is very hard to precisely locate the root cause within the design. Existing side-channel attack countermeasures are very costly in terms of design overhead. Without a method that can precisely identify the side-channel leakage source within the design, huge overhead will be introduced by blindly add the side-channel countermeasure to the whole design. To make the challenge even harder, the Power Distribution Network (PDN) where the hardware design locates is also vulnerable to side-channel attacks. It has been continuously demonstrated by researchers that attackers can place malicious circuits on a shared PDN with victim design and open the opportunities for the attackers to inject faults or monitoring power changes of the victim circuit. In this dissertation, we address the challenges mentioned above in designing a side-channel-resistant embedded system. We categorize our contributions into three major aspects—first, we investigating the effects of integration of security components and developing corresponding countermeasures. We analyze the vulnerability in a widely used countermeasure - masking, and identify that the random number transfer procedure is a weak link in the integration which can be bypassed by the attacker. We further propose a lightweight protection scheme to protect function calls from instruction skip fault attacks. Second, we developed a novel analysis methodology for pre-silicon side-channel leakage evaluation and root cause analysis. The methodology we developed enables the designer to detect the side-channel leakage at the early pre-silicon design stage, locate the leakage source in the design precisely to the individual gate and apply highly targeted countermeasure with low overhead. Third, we developed a multipurpose on-chip side-channel and fault monitoring extension - Programmable Ring Oscillator (PRO), to further guarantee the security of PDN. PRO can provide on-chip side-channel resistance, power monitoring, and fault detection capabilities to the secure design. We show that PRO as application-independent integrated primitives can provide side-channel and fault countermeasure to the design at a low cost.

Side Channel Attacks (SCA) are a class of passive attacks on cryptosystems that exploit implementation characteristics of the system. Currently, a lot of research is focussed towards developing countermeasures to side channel attacks. In this thesis, we address two challenges that are an inherent part of the efficient implementation of SCA countermeasures. While designing a system, design choices made for enhancing the efficiency or performance of the system can also affect the side channel security of the system. The first challenge is that the effect of different design choices on the side channel resistance of a system is currently not well understood. It is important to understand these effects in order to develop systems that are both secure and efficient. A second problem with incorporating SCA countermeasures is the increased design complexity. It is often difficult and time consuming to integrate an SCA countermeasure in a larger system. In this thesis, we explore that above mentioned problems from the point of view of developing embedded software that is resistant to power based side channel attacks. Our first work is an evaluation of different software AES implementations, from the perspective of side channel resistance, that shows the effect of design choices on the security and performance of the implementation. Next we present work that identifies the problems that arise while designing software for a particular type of SCA resistant architecture - the Virtual Secure Circuit. We provide a solution in terms of a methodology that can be used for developing software for such a system - and also demonstrate that this methodology can be conveniently automated - leading to swifter and easier software development for side channel resistant designs. / Master of Science

[EN] In the first part of the current thesis, two fundamental numerical models (Fin2D-W and Fin1D-MB) for analyzing the air-side performance of minichannel evaporators were developed and verified. The Fin2D-W model applies a comprehensive two-dimensional scheme to discretize the evaporator. On the other hand, the Fin1D-MB model is based on the one-dimensional fin theory in conjunction with the moving boundaries technique along the fin height. The first objective of the two presented models is to identify and quantify the most influential phenomena encountered in the process of cooling and dehumidification. The second objective is to study the impact of the classical modeling assumptions on the air-side performance of minichannel evaporators. Different comparative studies between the traditional Effectiveness-NTU approach and the proposed numerical models were implemented to achieve the mentioned goals. The results revealed that the modeling assumptions which have the most significant impacts on the heat and mass transfer rates are: the uniform air properties along the fin height, adiabatic-fin-tip at half the height, and negligence of partial dehumidification scenarios. These widely used assumptions resulted in substantial deviations in total heat transfer rate, up to 52%, between the Effectiveness-NTU approach and Fin2D-W model. In the second part of the thesis, the Fin1D-MB model was integrated into the IMST-ART® simulation tool to evaluate the global performance of minichannel evaporators (air- and refrigerant-side). The Fin1D-MB model was selected because of its simplicity, calculation speed, and reasonable solution accuracy relative to the Fin2D-W model. The validation of the complete Fin1D-MB model was conducted against many experimental data and numerical models available in the literature. The validation process was achieved for different heat exchanger geometries, refrigerants, and operating conditions. The results showed that for the R134a minichannel evaporators studied, the Fin1D-MB model successfully predicted the Inlet refrigerant and outlet air temperatures, cooling capacity, and refrigerant-side pressure drop within error bands of ±0.5 ºC, ±5%, and ±20%, respectively. For the CO2 (R744) minichannel evaporator studied, the presented model estimated the cooling capacity and outlet air temperature within error bands of ±10% and ±1.0 ºC, respectively. Regarding the CO2 pressure drop, the Fin1D-MB model generally underpredicted the pressure drop values compared to the experimental data, with a maximum deviation of 11 kPa. / [ES] En la primera parte de la tesis actual, dos modelos numéricos fundamentales (Fin2D-W y Fin1D-MB) para analizar el lado del aire de los evaporadores de minicanales se han desarrollado y verificado. El modelo Fin2D-W aplica un esquema detallado de dos dimensiones para discretizar el evaporador mientras que el modelo Fin1D-MB se basa en la teoría de la aleta unidimensional junto con la técnica de fronteras móviles para el lado del aire. El primer objetivo de los dos modelos presentados es identificar y cuantificar los fenómenos más influyentes encontrados en el proceso de enfriamiento y deshumidificación. El segundo objetivo es estudiar el impacto de las hipótesis comúnmente usadas en el modelado de la transmisión de calor del aire de los evaporadores de minicanales. Se implementaron diferentes estudios comparativos entre el enfoque tradicional Effectiveness-NTU y los modelos numéricos propuestos para alcanzar los objetivos mencionados. Los resultados muestran que las hipótesis que provocan una mayor desviación con respecto a la solución detallada en la transferencia de calor y masa son: propiedades de aire uniforme a lo largo de la altura de la aleta, extremo adiabático de aleta a mitad de su longitud, y no contemplar el supuesto de deshumidificación parcial en la aleta. Estas hipótesis ampliamente utilizadas han resultado en errores importantes en la transferencia de calor total, hasta un 52%, entre el enfoque Effectiveness-NTU y el modelo Fin2D-W. En la segunda parte de la tesis, el modelo Fin1D-MB se integró en la herramienta de simulación IMST-ART® para evaluar el rendimiento global de los evaporadores de minicanales (en el lado del aire y del refrigerante). El modelo Fin1D-MB se seleccionó gracias a su simplicidad, velocidad de cálculo, y solución de una precisión razonable relativa al modelo Fin2D-W. Se realizó una validación del modelo completo Fin1D-MB con la ayuda de datos experimentales y modelos numéricos ya disponibles en la literatura. El modelo se ha validado para diferentes geometrías de intercambiadores de calor, refrigerantes y condiciones de funcionamiento. Los resultados han mostrado que para los evaporadores de minicanales funcionando con el refrigerante R134a, el modelo Fin1D-MB predice de manera correcta las temperaturas de entrada del refrigerante y de salida del aire, la capacidad de enfriamiento, y la caída de presión del lado de refrigerante dentro de las bandas de error de ±0.5 ºC, ±5%, y ±20%, respectivamente. Para el evaporador de minicanales con CO2 (R744) estudiado, el modelo estima la capacidad de refrigeración y la temperatura de salida del aire dentro de las bandas de error de ±10% y ±1.0 ºC, respectivamente. En cuanto a la caída de presión de CO2, el modelo Fin1D-MB generalmente predice a la baja los valores de la caída de presión en comparación con los datos experimentales, con una desviación máxima de 11 kPa. / [CA] A la primera part de la tesi actual, dos models numèrics fonamentals (Fin2D-W i Fin1D-MB) per analitzar el costat de l'aire dels evaporadors de minicanals s'han desenvolupat i verificat. Al model Fin2D-W s'aplica un esquema detallat de dues dimensions per discretitzar l'evaporador mentre que al model Fin1D-MB es basa en la teoria d'aleta unidimensional juntament amb la tècnica de frontera mòbil per al costat de l'aire. El primer objectiu dels dos models presentats és identificar i quantificar els fenòmens més influents trobats en el procés de refredament i deshumidificació. El segon objectiu és estudiar l'impacte de les hipòtesis comunament utilitzades en el modelatge de la transmissió de calor de l'aire dels evaporadors de minicanals. Es van implementar diferents estudis comparatius entre l'enfocament tradicional Effectiveness-NTU i els models numèrics proposats per assolir els objectius esmentats. Els resultats mostren que les hipòtesis que provoquen una major desviació respecte a la solució detallada a la transferència de calor i massa són: propietats d'aire uniforme al llarg de l'altura de l'aleta, extrem adiabàtic d'aleta a la meitat de la seua longitud, i no contemplar el supòsit de deshumidificació parcial en l'aleta. Aquestes hipòtesis àmpliament utilitzades donen errors importants en la transferència de calor total, fins a un 52%, entre l'enfocament Effectiveness-NTU i el model Fin2D-W. A la segona part de la tesi, el model Fin1D-MB es va integrar en l'eina de simulació IMST-ART® per avaluar el rendiment global dels evaporadors de minicanals (al costat de l'aire i del refrigerant). El model Fin1D-MB es va seleccionar gràcies a la seva simplicitat, velocitat de càlcul, i solució d'una precisió raonable relativa al model Fin2D-W. Es va realitzar una validació del model complet Fin1D-MB amb l'ajuda de dades experimentals i models numèrics ja disponibles a la literatura. El model s'ha validat per a diferents geometries d'intercanviadors de calor, refrigerants i condicions de funcionament. Els resultats mostren que per als evaporadors de minicanals funcionant amb el refrigerant R134a, el model Fin1D-MB prediu de manera correcta les temperatures d'entrada del refrigerant i de sortida de l'aire, la capacitat de refreda-ment, i la caiguda de pressió del costat de refrigerant dins de les bandes d'error de ±0.5 ºC, ±5%, i ±20%, respectivament. Per a l'evaporador de minicanals amb CO2 (R744) estudiat, el model estima la capacitat de refrigeració i la temperatura de sortida de l'aire dins de les bandes d'error de ±10% i ±1.0 ºC, respectivament. Pel que fa a la caiguda de pressió de CO2, el model Fin1D-MB generalment prediu a la baixa els valors de la caiguda de pressió en comparació amb les dades experimentals, amb una desviació màxima d'11 kPa. / Hassan, AHA. (2016). Development and Validation of a Minichannel Evaporator Model under Dehumidification [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/71357

There is a significant current need to elucidate the molecular mechanisms of the side-effects caused by widely-used pharmaceuticals. Examples include the acute nephrotoxicity and irreversible ototoxicity promoted by the cationic drugs gentamicin and cisplatin. Gentamicin is an aminoglycoside antibiotic used for the prevention and treatment of life-threatening gram-negative bacterial infections, such as tuberculosis and meningitis. Cisplatin is used to treat a broad spectrum of cancers including head and neck, ovarian, cervical, stomach, bladder, sarcoma, lymphoma, testicular cancer and others. The objective of this study is to design and synthesize rhodamine derivatives that can be used for the construction of geometrically well-defined cationic drug conjugates. The long-term goal is to use the conjugates as tools to aid in elucidating the properties and identities of ion channels involved in the uptake of cationic pharmaceuticals into kidney and cochlear hair cells. This will shed light on the origin and potential prevention of unwanted side effects such as nephrotoxicity and ototoxicity associated with specific cationic drugs. A series of extended rhodamine analogs with reactive groups for biomolecule conjugation has been synthesized. These fluorophores show similar spectral properties to their prototype, Texas Red succinimidyl ester (TR-SE). However, they contain rigid linkers between the fluorophore and amine-reactive moiety. The resultant gentamicin conjugates of these materials are rigidified enabling one to assess channel pore dimensions without the confounding issue of conjugate folding. Preliminary cell studies are promising, as one observes reduced gentamicin uptake in both kidney and sensory hair cell upon systematically increasing the dimension of the fluorophore. This work has enabled us to tentatively assign the maximum dilated MET channel pore size as between 1.44 nm to 1.56 nm. However, this preliminary finding, though encouraging, needs further validation via ongoing studies with larger diameter fluorophore conjugates, A cisplatin-Texas Red conjugate has also been synthesized to enable studies of cellular uptake mechanisms. This conjugate preserves not only the spectral properties of Texas Red after conjugation, but also the cytotoxicity of cisplatin. This has been validated in zebrafish. The series of rhodamine probes that have been conjugated to gentamicin should be similarly useful for cisplatin studies. These studies are planned. Additional future work includes the synthesis of semi-flexible (glycol) and flexible (alkyl) linkers to evaluate structure-activity relationships.

South Africa’s electricity supply is under pressure because of inadequate capacity expansion in the early 2000s. One of the initiatives funded by Eskom to alleviate the pressure on the national electricity grid was an aggressive demand-side management (DSM) programme that commenced in 2004. A positive outcome of the DSM programme was that the industrial sector in South Africa benefited from the implementation of a relatively large number of DSM projects. These DSM projects reduced the electricity costs of industrial clients and reduced the demand on the national electricity grid. Unfortunately, the performance of industrial DSM projects deteriorates without proper maintenance. This results in wasted savings opportunities that are costly to industrial clients and Eskom. The purpose of this study was therefore to develop a maintenance strategy that could be applied, firstly, to reverse the deterioration of DSM project performance and, secondly, to sustain and to improve DSM project performance. The focus of the maintenance strategy was to obtain maximum project performance that translated to maximum electricity cost savings for the client. A new performance-centered maintenance (PCM) strategy was developed and proven through practical experience in maintaining industrial DSM projects over a period of more than 60 months. The first part of the PCM strategy consisted of developing a new strategy for the outsourcing of DSM project maintenance to energy services companies (ESCOs) on the company group level of the client. The strategy served as a guideline for both ESCOs and industrial clients to implement and manage a group-level DSM maintenance agreement successfully. The second part of the PCM strategy consisted of a simplified method that was developed to identify DSM projects where applying a PCM strategy would increase or sustain electricity cost savings. The third part of the PCM strategy consisted of practical maintenance guidelines that were developed to ensure maximum project performance. It was based on the plan-do-check-act cycle for continuous improvement with an emphasis on the monitoring of DSM project performance. The last part of the PCM strategy consisted of various alternative key performance indicators that should be monitored to ensure maximum sustainable DSM project performance. The PCM strategy was evaluated by implementing it on ten different DSM projects. The results showed that applying a PCM strategy resulted in an average increase of 64.4% in the electricity cost savings generated by these projects. The average implementation cost of the PCM strategy was 6% of the total benefit generated through it. This indicated that implementing the PCM strategy was a cost-effective manner to ensure that maximum performance of DSM projects was maintained sustainably. / PhD (Computer and Electronic Engineering), North-West University, Potchefstroom Campus, 2015

Costs saving measures are important for South African gold producers due to increasing energy costs and decreasing production volumes. Demand Side Management (DSM) is an effective strategy to reduce electricity consumption and costs. DSM projects have been implemented widely on South African mining systems such as pumping, refrigeration, rock transport and compressed air. Implementations have, however, been limited on gold processing plants despite the significant amounts of energy that this section consumes. The main objective of gold processing plants is production orientated and energy management is not a primary focus. This rationale is re-evaluated owing to high electricity price inflation and availability of DSM incentives. This study investigated the cost saving potential of DSM interventions on gold plants. Electrical load management was identified as a key opportunity that can deliver substantial cost savings. These savings were shown to be feasible in respect of the required capital expenditure, effort of implementation and maintenance of operational targets. Investigation procedures were compiled to identify feasible load management opportunities. The most potential for electricity cost savings was identified on comminution equipment. Consequently, a methodology was developed to implement electrical load management on the identified sections. The methodology proposed simulation techniques that enabled load management and subsequent electricity cost optimisation through production planning. Two electrical load management case studies were successfully implemented on comminution equipment at two gold processing plants. Peak period load shift of 3.6 MW and 0.6 MW, respectively, was achieved on average for a period of three months. The annual cost savings of these applications could amount to R1.4-million and R 660 000. This results in specific electricity cost reductions of 3% and 7% for the two respective case studies. Results from the two case studies are an indication of potential for electrical load management on South African gold processing plants. If an average electricity cost saving of 5% is extrapolated across the South African gold processing industry, the potential cost savings amount to R 25-million per annum. Although the costs saving opportunities are feasible, it is influenced by the reliability of the equipment and the dynamics of ore supply. This insight plays a decisive role in determining the feasibility of DSM on gold processing plants. / MIng (Mechanical Engineering), North-West University, Potchefstroom Campus, 2015

MIng (Electrical and Electronic Engineering), North-West University, Potchefstroom Campus, 2015 / Demand side management (DSM) initiatives have been introduced by Eskom to reduce the deficit between the electricity generation capacity and the electricity usage within the country. DSM projects enable Eskom to reduce electricity demand instead of increasing generation capacity. DSM projects are more economical and can be implemented much faster than constructing a new power station. One particular industry where DSM projects can be implemented is on mines. Mines consume about 14.5% of South Africa’s electricity. Producing compressed air, in particular, is one of the largest electricity users on mines. It consumes 17% of the electricity used on mines. The opportunity, therefore, arises to implement DSM projects on the compressed air system of mines. Not only do these projects reduce Eskom’s high electricity demand, but they also induce financial and energy savings for the mine itself. However, during the implementation of a compressed air energy savings project, various challenges arise. These include, among others, operational changes, control limitations, industrial actions and installation delays. All of these can lead to a project not being delivered on time, within budget or with quality results. The purpose of this study is to investigate and address various problems that occur during the implementation of such a compressed air energy savings project. The study shows that although these problems have an impact on the results achievable with the project, significant savings are still possible. Project savings are achieved by reducing the amount of compressed air that is supplied, thereby delivering sufficient compressed air while minimising the amount of compressed air being wasted. During this study, a gold mine’s compressed air network was optimised. The optimisation resulted in an evening peak-clip saving of 2.61 MW. This saving was achieved daily between 18:00 and 20:00 when Eskom’s electricity demand was at its highest. It is equivalent to an annual cost saving of R1.46 million based on Eskom’s 2014/2015 tariffs. When savings from all periods throughout the day are taken into account, the project will produce an annual cost saving of R1.91 million.