Ingegneria industrialehttp://elea.unisa.it/xmlui/handle/10556/25942024-03-28T10:46:52Z2024-03-28T10:46:52ZNew green nano-materials for the reduction of friction and wear in lubricants for power transmissionsWaleed Ahmed Abdaglil Mustafahttp://elea.unisa.it/xmlui/handle/10556/66092024-02-07T15:42:28Z2021-10-25T00:00:00ZNew green nano-materials for the reduction of friction and wear in lubricants for power transmissions
Waleed Ahmed Abdaglil Mustafa
Nanomaterials have emerged as future environmentally sustainable lubricant additives to improve conventional lubricants such as car oils, industrial oils, grease and metalworking fluids. In particular, as far as this thesis is concerned, nanoparticle additives' applications in transmissions oils are based on the concepts of solid lubrication and are often used as anti-wear, anti-friction, and high-pressure additives. Their multiple advantages include small enough scale, thermal stability, unique particle chemistries, mechanical properties, and a high surface reaction rate. These benefits translate into the prolonged running of the equipment, improved fuel efficiency and extended maintenance cycles. The main drawback of solid lubricant additives is the inherent low stability of liquid base lubricant systems, which has greatly limited them from industrial applications. Thus, the current PhD thesis is aimed to design novel techniques to improve the dispersibility of nanoadditives in base oils, while encompassing both the theoretical and the industrial point of view. The concluded results from this industrial research have shown promising results in simultaneous stability and tribological improvement. Also, future use of nanoadditives in electric vehicle applications is critically discussed, and key findings are achieved. [edited by Author]
2019 - 2020
2021-10-25T00:00:00ZDrying of fruits: improvement of quality and process modellingÖnal, Begümhttp://elea.unisa.it/xmlui/handle/10556/65912024-02-07T15:37:19Z2020-06-26T00:00:00ZDrying of fruits: improvement of quality and process modelling
Önal, Begüm
In the recent years, dehydrated food industry has gained prominence in
the world. In concern with increasing demand of high quality and healthy
products and changing customer behaviour, the food market needs to
maintain at a high level nutritional and sensory properties of the initial fresh
products. Drying of fruits and vegetables enables the availability of these
products on the global markets during all seasons.
The aim of this research wasto investigate the effects of different pretreatments (i.e natural - innovative dipping solutions, microwave and ultrasound applications) and hot air drying process conditions on the drying characteristics and quality propertiesof selected fruits in terms of colour,
shrinkage, total phenolics, antioxidant activity, volatile aroma compounds,
microstructure, texture, preliminary sensorial evaluation, rehydration
behaviour. .. [edited by the Author]
2018 - 2019
2020-06-26T00:00:00ZUltra-low power HW accelerator for the integration of Binary Neural NetworksDe Vita, Antoniohttp://elea.unisa.it/xmlui/handle/10556/64902024-02-07T15:35:54Z2021-06-14T00:00:00ZUltra-low power HW accelerator for the integration of Binary Neural Networks
De Vita, Antonio
The research activity described in this thesis aims to demonstrate the possibility to embed Artificial
Intelligence (AI) capabilities in wearable and portable devices by deploying and executing Neural Network
(NN) models close to the sensing element. Among AI models, Deep Learning (DL) and Deep Neural Networks
can achieve high performance in many tasks, e.g. image classification, activity recognition, and so on.
However, DL models usually require a huge amount of memory resources and high-performance digital
architecture to be executed. These specifications are hardly met by wearable and portable devices, which have
to be as small as possible and guarantee a satisfactory battery lifetime. For this reason, the cloud computing
strategy is often used. However, higher latencies occur in this case, which can be unacceptable in many latencysensitive
applications, such as autonomous vehicles or assisted microsurgery. Moreover, the data transfer
consumes network bandwidth and energy. In this context, moving the computation close to the device is highly
demanded, and it is named edge-computing. However, deploying DL models on edge devices is still a
challenge. General-purpose platforms (i.e. CPUs, GPUs) are not the best solution in terms of energy efficiency,
especially for wearable and battery-powered devices, where the device lifetime is a major concern. Thus, a lot
of research is being made about the design of custom HW accelerators for DL and to move the circuitry needed
to implement the computation closer to the sensing element, thus obtaining a smart sensor. In this thesis, a
novel Hybrid Binary Neural Network (HBN) model is proposed, which exploits the advantages of Binarized
Neural Networks (BNNs). Human Activity Recognition (HAR) based on inertial sensors has been selected as
a case study. Also, a pre-processing algorithm has been developed to solve the device-orientation problem for
3-axis accelerometers. The pre-processing operations can improve the accuracy of the proposed system in
some conditions when it is used in conjunction with the HBN model. The results show an accuracy of up to
99% in recognizing 5 human activities. After having developed the model, a custom ultra-low power HW
accelerator has been designed and implemented with both FPGA and CMOS standard cells. Due to the very
low operating frequency required by HAR applications, power consumption has been reduced by reducing the
number of resources. The design can implement both the pre-processing operation and the HBN model. The
results show that the HW accelerator has a power consumption of 6.3 μW and an area occupation of 0.20 mm2
when synthesized with CMOS 65 nm Low-Power (LP) High Voltage Threshold (HVT) standard cells. The
proposed design has at least 7.3 times lower power consumption than the state-of-the-art solution. Also, a
FPGA-based demo board has been developed to demonstrate the real-time operation of the system. [edited by Author]
2019 - 2020
2021-06-14T00:00:00ZUtilization of high pressure processing (HPP) for the production of starch-based hydrogels for innovative applicationsLarrea-Wachtendorff, Dominiquehttp://elea.unisa.it/xmlui/handle/10556/64652024-02-07T15:37:01Z2020-07-07T00:00:00ZUtilization of high pressure processing (HPP) for the production of starch-based hydrogels for innovative applications
Larrea-Wachtendorff, Dominique
Nowadays, the development of plant-based systems to replace or reduce the utilization of synthetic
materials has been receiving significant attention, in view to fulfil consumers demand natural
products in all the industrial areas.
Hydrogels represent a group of polymeric materials, composed of three-dimensional crosslinked
polymeric networks, capable to absorb and retain a significant amount of water. They have been listed
as “smart structures” whose tailor-made design confers them different functional attributes allowing
their use in biomedical, cosmeceutical, pharmaceutical and food applications. Hydrogels can be
produced either from natural or synthetic sources. However, those produced from natural sources
have gained a great interest in research for the development of novel biomaterials for which a wide
range of applications could be envisaged due to their safety, biocompatibility, and biodegradability.
In the last decade, among the natural sources to produce hydrogels, starches have been receiving
increasing attention as one of the most promising natural biopolymers. Hydrogels are traditionally
produced by chemical or physical methods. However, long processing time, high energy
consumption, and safety issues related to the synthesis of these products have been identified as
important limitations of these methods. .. [edited by the Author]
2018 - 2019
2020-07-07T00:00:00Z