skip to main content


Title: Supply Air Temperature Prediction in an Air-Handling Unit Using Artificial Neural Network
Continuous provision of quality supply air to data center’s IT pod room is a key parameter in ensuring effective data center operation without any down time. Due to number of possible operating conditions and non-linear relations between operating parameters make the working mechanism of data center difficult to optimize energy use. At present industries are using computational fluid dynamics (CFD) to simulate thermal behaviour for all types of operating conditions. The focus of this study is to predict Supply Air Temperature using Artificial Neural Network (ANN) which can overcome limitations of CFD such as high cost, need of an expertise and large computation time. For developing ANN, input parameters, number of neurons and hidden layers, activation function and the period of training data set were studied. A commercial CFD software package 6sigma room is used to develop a modular data center consisting of an IT pod room and an air-handling unit. CFD analysis is carried out for different outside air conditions. Historical weather data of 1 year was considered as an input for CFD analysis. The ANN model is “trained” using data generated from these CFD results. The predictions of ANN model and the results of CFD analysis for a set of example scenarios were compared to measure the agreement between the two. The results show that the prediction of ANN model is much faster than full computational fluid dynamics simulations with good prediction accuracy. This demonstrates that ANN is an effective way for predicting the performance of an air handling unit.  more » « less
Award ID(s):
1738811
NSF-PAR ID:
10101162
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
ASME 2018 International Mechanical Engineering Congress and Exposition
Volume:
8B
Issue:
Heat Transfer and Thermal Engineering
Page Range / eLocation ID:
V08BT10A064
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. In this study, pressure drop ( ) across air-cooled heat sinks (HSs) are predicted using an artificial neural network (ANN). A multilayer feed-forward ANN architecture with two hidden layers is developed. Backpropagation algorithm is used for training the network, and the accuracy of the network is evaluated by the root mean square error. The input data for training the neural network is prepared through three-dimensional simulation of air inside the channels of heat sinks using a computational fluid dynamics (CFD) approach. The developed ANN-based model in this study predicts with a high accuracy and within of the CFD-based data. The present study suggests that developing an ANN-based model with a high level of accuracy overcomes the limitations of physics-based correlations that their accuracy strongly depends on identifying and implementing key variables that affect the physics of a thermo-fluid phenomenon. 
    more » « less
  2. Most of the thermal management technologies concentrate on managing airflow to achieve the desired server inlet temperature (supply air operating set point) and not to manage/improve the amount of cool air (CFM) that each computer rack (i.e. IT servers) should receive in order to remove the produced heat. However, airflow is equally important for quantifying adequate cooling to IT equipment, but it is more challenging to obtain a uniform airflow distribution at the inlet of computer racks. Therefore, as a potential option for improving airflow distribution is to eliminate the sources of non-uniformities such as maldistribution of under-floor plenum pressure field caused by vortices. Numerous researchers focus on the adverse effects of under-floor blockages. This study focused to numerically investigate the positive impact of selectively placed obstructions (on-purpose air-directors); referred as partitions; Quantitative and qualitative analysis of underfloor plenum pressure field, perforated tiles airflow rate and racks inlet temperature with and without partitions using two Computational Fluid Dynamics (CFD) models, which were built using Future Facilities 6SigmaRoom CFD tool. First, a simple data center model was used to quantify the partitions benefits for two different systems; Hot Aisle Containment (HAC) compared to an open configuration. Second, the investigation was expanded using a physics-based experimentally validated CFD model of medium size data center (more complicated data center geometry) to compare different types of proposed partitions. Both models results showed that partition type I (partitions height of $\frac{2}{3}$ of plenum depth measured from the subfloor) eliminates the presence of vortices in the under-floor plenum and hence, more uniform pressure differential across the perforated tiles that drives more uniform airflow rates. In addition, the influence of proposed partitions on the rack inlet temperature was reported through a comparison between open versus hot aisle containment. The results showed that the partitions have a minor effect on the rack inlet temperature for the hot aisle containment system. However, the partitions significantly improve the tiles flowrate. On the other hand, for the open system, the presence of partitions has improved the tiles airflow rate, rack inlet temperature and hence eliminate the hot spots formation at computer rack inlet 
    more » « less
  3. Abstract An increasingly common power saving practice in data center thermal management is to swap out air cooling unit blower fans with electronically commutated plug fans, Although, both are centrifugal blowers. The blade design changes: forward versus backward curved with peak static efficiencies of 60% and 75%, respectively, which results in operation power savings. The side effects of which are not fully understood. Therefore, it has become necessary to develop an overall understanding of backward curved blowers and compare the resulting flow, pressure, and temperature fields with forwarding curved ones in which the induced fields are characterized, compared, and visualized in a reference data center which may aid data center planning and operation when making the decisions of which computer room air handler (CRAH) technology to be used. In this study, experimental and numerical characterization of backward curved blowers is introduced. Then, a physics-based computational fluid dynamics model is built using the 6sigmaroom tool to predict/simulate the measured fields. Five different scenarios were applied at the room level for the experimental characterization of the cooling units and another two scenarios were applied for comparison and illustration of the interaction between different CRAH technologies. Four scenarios were used to characterize a CRAH with backward curved blowers, during which a CRAH with forwarding curved was powered off. An alternate arrangement was examined to quantify the effect of possible flow constraints on the backward curved blower's performance. Then parametric and sensitivity of the baseline modeling are investigated and considered. Different operating conditions are applied at the room level for experimental characterization, comparison, and illustration of the interaction between different CRAH technologies. The measured data is plotted and compared with the computational fluid dynamics (CFD) model assessment to visualize the fields of interest. The results show that the fields are highly dependent on CRAH technology. The tile to CRAH airflow ratios for the flow constraints of scenarios 1, 2, 3, and 4 are 85.5%, 83.9%, 61%, and 59%, respectively. The corresponding leakage ratios are 14.5%, 16%, 38.9%, and 41%, respectively. Furthermore, the validated CFD model was used to investigate and compare the airflow pattern and plenum pressure distribution. Lastly, it is notable that a potential side effect of backward curved technology is the creation of an airflow dead zone. 
    more » « less
  4. Abstract Continuous rise in cloud computing and other web-based services propelled the data center proliferation seen over the past decade. Traditional data centers use vapor-compression-based cooling units that not only reduce energy efficiency but also increase operational and initial investment costs due to involved redundancies. Free air cooling and airside economization can substantially reduce the information technology equipment (ITE) cooling power consumption, which accounts for approximately 40% of energy consumption for a typical air-cooled data center. However, this cooling approach entails an inherent risk of exposing the ITE to harmful ultrafine particulate contaminants, thus, potentially reducing the equipment and component reliability. The present investigation attempts to quantify the effects of particulate contamination inside the data center equipment and ITE room using computational fluid dynamics (CFD). An analysis of the boundary conditions to be used was done by detailed modeling of ITE and the data center white space. Both two-dimensional and three-dimensional simulations were done for detailed analysis of particle transport within the server enclosure. An analysis of the effect of the primary pressure loss obstructions like heat sinks and dual inline memory modules inside the server was done to visualize the localized particle concentrations within the server. A room-level simulation was then conducted to identify the most vulnerable locations of particle concentration within the data center space. The results show that parameters such as higher velocities, heat sink cutouts, and higher aspect ratio features within the server tend to increase the particle concentration inside the servers. 
    more » « less
  5. The objective of this work is to introduce the application of an artificial neural network (ANN) to assist in the evaporative cooling in data centers. To achieve this task, we employ the neural network algorithms to predict weather conditions outside the data center for direct evaporative cooling (DEC) operations. The predictive analysis helps optimize the cooling control strategy for maximizing the usage of evaporative cooling thereby improving the efficiency of the overall data center cooling system. A typical artificial neural network architecture is dynamic in nature and can perform adaptive learning in minimal computation time. A neural network model of a data center was created using operational historical data collected from a data center cooling control system. The neural network model allows the control of the modular data center (MDC) cooling at optimum configuration in two ways. First way is that the network model minimizes time delay for switching the cooling from one mode to the other. Second way, it improves the reaction behavior of the cooling equipment if an unexpected ambient condition change should come. The data center in consideration is a test bed modular data center that comprises of information Technology (IT) racks, Direct Evaporative cooling (DEC) and Indirect Evaporative Cooling (IEC) modules; the DEC/IEC are used together or in alternative mode to cool the data center room. The facility essentially utilizes outside ambient temperature and humidity conditions that are further conditioned by the DEC and IEC to cool the electronics, a concept know as air-side economization. Various parameters are related to the cooling system operation such as outside air temperature, IT heat load, cold aisle temperature, cold aisle humidity etc. are considered. Some of these parameters are fed into the artificial neural network as inputs and some are set as targets to train the neural network system. After the training the process is completed, certain bucket of data is tested and further used to validate the outputs for various other weather conditions. To make sure the analysis represents real world scenario, the operational data used are from real time data logged on the MDC cooling control unit. Overall, the neural network model is trained and is used to successfully predict the weather conditions and cooling control parameters. The prediction models have been demonstrated for the outputs that are static in nature (Levenberg Marquardt method) as well as the outputs that are dynamic in nature i.e., step-ahead & multistep ahead techniques. 
    more » « less