-
Structural Health Monitoring of Cantilever Beam, a Case Study -- Using Bayesian Neural Network AND Deep Learning
Authors:
Rahul Vashisht,
H. Viji,
T. Sundararajan,
D. Mohankumar,
S. Sumitra
Abstract:
The advancement of machine learning algorithms has opened a wide scope for vibration-based SHM (Structural Health Monitoring). Vibration-based SHM is based on the fact that damage will alter the dynamic properties viz., structural response, frequencies, mode shapes, etc of the structure. The responses measured using sensors, which are high dimensional in nature, can be intelligently analyzed using…
▽ More
The advancement of machine learning algorithms has opened a wide scope for vibration-based SHM (Structural Health Monitoring). Vibration-based SHM is based on the fact that damage will alter the dynamic properties viz., structural response, frequencies, mode shapes, etc of the structure. The responses measured using sensors, which are high dimensional in nature, can be intelligently analyzed using machine learning techniques for damage assessment. Neural networks employing multilayer architectures are expressive models capable of capturing complex relationships between input-output pairs but do not account for uncertainty in network outputs. A BNN (Bayesian Neural Network) refers to extending standard networks with posterior inference. It is a neural network with a prior distribution on its weights. Deep learning architectures like CNN (Convolutional neural network) and LSTM(Long Short Term Memory) are good candidates for representation learning from high dimensional data. The advantage of using CNN over multi-layer neural networks is that they are good feature extractors as well as classifiers, which eliminates the need for generating hand-engineered features. LSTM networks are mainly used for sequence modeling. This paper presents both a Bayesian multi-layer perceptron and deep learning-based approach for damage detection and location identification in beam-like structures. Raw frequency response data simulated using finite element analysis is fed as the input of the network. As part of this, frequency response was generated for a series of simulations in the cantilever beam involving different damage scenarios. This case study shows the effectiveness of the above approaches to predict bending rigidity with an acceptable error rate.
△ Less
Submitted 17 August, 2019;
originally announced August 2019.
-
Heat spreader with parallel microchannel configurations employing nanofluids for near active cooling of MEMS
Authors:
Lakshmi Sirisha Maganti,
Purbarun Dhar,
T Sundararajan,
Sarit Kumar Das
Abstract:
While parallel microchannel based cooling systems have been around for quite a period of time, employing the same and incorporating them for near active cooling of microelectronic devices is yet to be implemented and the implications of the same on thermal mitigation to be understood. The present article focusses on a specific design of the PMCS such that it can be implemented at ease on the heat…
▽ More
While parallel microchannel based cooling systems have been around for quite a period of time, employing the same and incorporating them for near active cooling of microelectronic devices is yet to be implemented and the implications of the same on thermal mitigation to be understood. The present article focusses on a specific design of the PMCS such that it can be implemented at ease on the heat spreader of a modern microprocessor to obtain near active cooling. Extensive experimental and numerical studies have been carried out to comprehend the same and three different flow configurations of PMCS have been adopted for the present investigations. Additional to focussing on the thermofluidics due to flow configuration, nanofluids have also been employed to achieve the desired essentials of mitigation of overshoot temperatures and improving uniformity of cooling. Two modelling methods, Discrete Phase Modelling and Effective Property Modelling have been employed for numerical study to model nanofluids as working fluid in micro flow paths and the DPM predictions have been observed to match accurately with experiments. To quantify the thermal performance of PMCS, an appropriate Figure of Merit has been proposed. From the FoM It has been perceived that the Z configuration employing nanofluid is the best suitable solutions for uniform thermal loads to achieve uniform cooling as well as reducing maximum temperature produced with in the device. The present results are very promising and viable approach for futuristic thermal mitigation of microprocessor systems.
△ Less
Submitted 27 February, 2017;
originally announced April 2017.
-
Selecting optimal parallel microchannel configurations for active hot spot mitigation of multicore microprocessors in real time
Authors:
Lakshmi Sirisha Maganti,
Purbarun Dhar,
T Sundararajan,
Sarit K Das
Abstract:
Design of effective micro cooling systems to address the challenges of ever increasing heat flux from microdevices requires deep examination of real time problems and has been tackled in depth. The most common and apparently misleading assumption while designing micro cooling systems is that the heat flux generated by the device is uniform, but the reality is far from this. Detailed simulations ha…
▽ More
Design of effective micro cooling systems to address the challenges of ever increasing heat flux from microdevices requires deep examination of real time problems and has been tackled in depth. The most common and apparently misleading assumption while designing micro cooling systems is that the heat flux generated by the device is uniform, but the reality is far from this. Detailed simulations have been performed by considering non uniform heat load employing the configurations U, I, Z for parallel microchannel systems with water and nanofluids as the coolants. An Intel Core i7 4770 3.40 GHz quad core processor has been mimicked using heat load data retrieved from a real microprocessor with non-uniform core activity. The study clearly demonstrates that there is a non-uniform thermal load induced temperature maldistribution along with the already existent flow maldistribution induced temperature maldistribution. The suitable configuration(s) for maximum possible overall heat removal for a hot zone while maximizing the uniformity of cooling have been tabulated. An Eulerian Lagrangian model of the nanofluids show that such smart coolants not only reduce the hot spot core temperature, but also the hot spot core region and thermal slip mechanisms of Brownian diffusion and thermophoresis are at the crux of this. The present work conclusively shows that high flow maldistribution leads to high thermal maldistribution, as the common prevalent notion, is no longer valid and existing maldistribution can be effectively utilized to tackle specific hot spot location, making the present study important to the field.
△ Less
Submitted 10 July, 2016;
originally announced July 2016.
-
Thermally smart characteristics of nanofluids in parallel microchannel systems to mitigate hot spots in MEMS
Authors:
Lakshmi Sirisha Maganti,
Purbarun Dhar,
T Sundararajan,
Sarit K Das
Abstract:
Mitigation of hot spots in MEMS employing in situ microchannel systems requires a comprehensive picture of the maldistribution of the working fluid and uniformity of cooling within the same. In this article, detailed simulations employing parallel micro channel systems with specialized manifold-channel configurations i.e. U, I and Z have been performed. Eulerian Lagrangian Discrete Phase Model and…
▽ More
Mitigation of hot spots in MEMS employing in situ microchannel systems requires a comprehensive picture of the maldistribution of the working fluid and uniformity of cooling within the same. In this article, detailed simulations employing parallel micro channel systems with specialized manifold-channel configurations i.e. U, I and Z have been performed. Eulerian Lagrangian Discrete Phase Model and Effective Property Model with water and alumina water nanofluid as working fluids have been employed. The distributions of the dispersed particulate phase and continuous phase have been observed to be, in general, different from the flow distribution and this has been found to be strongly dependent on the flow configuration. Particle maldistribution has been conclusively shown to be influenced by various migration and diffusive phenomena like Stokesian drag, Brownian motion, thermophoretic drift, etc. To understand the uniformity of cooling within the device, which is of importance in real time scenario, an appropriate figure of merit has been proposed. It has been observed that uniformity of cooling improved using nanofluid as working fluid as well as enhanced relative cooling in hot zones, providing evidence of the smart nature of such dispersions. To further quantify this smart effect, real time mimicking hot-spot scenarios have been computationally probed with nanofluid as the coolant. A silicon-based microchip emitting non-uniform heat flux (gathered from real-time monitoring of an Intel Core i7 4770 3.40 GHz quad core processor) under various processor load conditions has been studied and evidence of enhanced cooling of hot spots has been obtained from DPM analysis.
△ Less
Submitted 1 May, 2016;
originally announced May 2016.
-
Particle and thermo-hydraulic maldistribution of nanofluids in parallel microchannel systems
Authors:
Lakshmi Sirisha Maganti,
Purbarun Dhar,
T. sundararajan,
Sarit K Das
Abstract:
Fluidic maldistribution in microscale multichannel devices requires deep understanding to achieve optimized flow and heat transfer characteristics. A thorough computational study has been performed to understand the concentration and thermohydraulic maldistribution of nanofluids in parallel microchannel systems using an Eulerian Lagrangian twin phase model. The study reveals that nanofluids cannot…
▽ More
Fluidic maldistribution in microscale multichannel devices requires deep understanding to achieve optimized flow and heat transfer characteristics. A thorough computational study has been performed to understand the concentration and thermohydraulic maldistribution of nanofluids in parallel microchannel systems using an Eulerian Lagrangian twin phase model. The study reveals that nanofluids cannot be treated as homogeneous single phase fluids in such complex flow domains and effective property models fail drastically to predict the performance parameters. To comprehend the distribution of the particulate phase, a novel concentration maldistribution factor has been proposed. It has been observed that distribution of particles need not essentially follow the flow pattern, leading to higher thermal performance than expected from homogeneous models. Particle maldistribution has been conclusively shown to be due to various migration and diffusive phenomena like Stokesian drag, Brownian motion, thermophoretic drift, etc. The implications of particle distribution on the cooling performance have been illustrated and smart fluid effects (reduced magnitude of maximum temperature) have been observed and a mathematical model to predict the enhanced cooling performance in such flow geometries has been proposed. The article presents lucidly the effectiveness of discrete phase approach in modelling nanofluid thermohydraulics and sheds insight on behavior of nanofluids in complex flow domains.
△ Less
Submitted 14 December, 2015;
originally announced January 2016.
-
Criterion for sliding / rolling characterization during droplet motion over superhydrophobic surfaces
Authors:
Indrajit P. Wadgaonkar,
T. Sundararajan,
Sarit K. Das
Abstract:
Super hydrophobic surfaces have been the focus of research in the recent years.One of the reasons for this is the self cleaning property of these surfaces which emerges from the ability of the droplets to roll freely over them.However majority of the studies available in literature are on the static wetting behavior of liquid droplets on such surfaces and the physics of the motion of droplets has…
▽ More
Super hydrophobic surfaces have been the focus of research in the recent years.One of the reasons for this is the self cleaning property of these surfaces which emerges from the ability of the droplets to roll freely over them.However majority of the studies available in literature are on the static wetting behavior of liquid droplets on such surfaces and the physics of the motion of droplets has not been studied exhaustively either theoretically or experimentally.In the present study droplet motion on super hydrophobic surfaces has been modeled to analyze the sliding/ rolling characteristics of the droplet motion.A non-dimensional number is proposed to indicate whether a given droplet would tend to roll or slide more on a given super hydrophobic surface.We refer to this number as 'Slip Reynolds' number. Simulations of droplet motion were carried out with different surface and droplet characteristics leading to a unique value of this number corresponding to sliding/rolling behavior.The applicability of this number was tested with available experiments from literature.
△ Less
Submitted 27 November, 2015;
originally announced November 2015.
