Haslinda Mohamed Kamar

Application of Multilayer Perceptron and Radial Basis Function Neural Network in steady state modeling of automotive air conditioning system

In this paper, steady-state models of an automotive air conditioning (ACC) are identified based on two different artificial neural networks (ANN) architectures: Multilayer Perceptron Neural Networks (MLPNN) and Radial Basis Function Neural Networks (RBFNN). The ANN models are developed with a four-in three-out configuration to simulate the outlet evaporating air temperature, cooling capacity, and compressor power under different combination of input compressor speeds, evaporating air speeds, air temperature upstream of the condenser and evaporator. The required data for the system identification are collected from an experimental bench made up of the original components of an AAC system. Investigations signify the advantage of a RBFNN model over MLPNN in modeling the AAC system.

Dynamic modeling of an automotive air conditioning system and an auto tuned PID controller using extremum seeking algorithm

This paper presents the application of Extremum Seeking (ES) algorithm in auto tuning of an incremental proportional-integral-derivative (PID) controller in order to achieve optimal cabin temperature control of an automotive air conditioning (AAC) model. A mathematical nonlinear model of an AAC system with variable speed compressor is first developed. The AAC model consists of a condenser, evaporator, thermal expansion valve, variable speed compressor driven by a direct current (DC) motor and a vehicle cabin. The ES algorithm is introduced and used to tune the parameters of the PID controller via iterative step function simulation. Finally, step response tests show that PID controller delivers better performance after being tuned by ES algorithm.

Assessment of thermal comfort in a naturally ventilated residential terrace house

In hot and humid climates thermal discomfort is a major problem to the occupants of many residential terrace houses especially when they are not equipped with an air-conditioning system. This paper presents a study on an assessment of the level of thermal comfort in a naturally ventilated residential terrace house in Malaysia using computational fluid dynamics (CFD) method. Actual measurements were made to obtain the average air temperature, relative humidity and air flow pattern in various sections of the house. CFD simulations were conducted on a simplified model of the house to predict and visualize the temperature distribution and air flow pattern and its velocity in the house. The level of thermal comfort in the house was found to be well outside the comfort limits as specified by ASHRAE standards.

CFD Simulation of Air Temperature inside a Bus Passenger Compartment

Good ventilation is important for passenger for sufficient supply of fresh air during commuting in a bus. Insufficient fresh air causes feeling of uncomfortable to passenger and affects passenger’s health. Airborne transmission disease, headache and respiratory allergies are the usual health symptoms. This paper presents the CFD study of air flow inside a bus passenger compartment. The objective is to estimate the temperature level at the diffuser, seat and floor locations of the bus passenger compartment. Two conditions of airflow velocity at the supply diffuser were examined, namely 2.7 m/s and 3.1 m/s. A CFD Fluent software was employed to develop and meshed a simplified 3D model of a quarter section of a bus passenger compartment. Air velocity and temperature boundary conditions were prescribed on the model based on the actual data obtained from field measurement. Turbulent flow analyses were carried out using standard k-ε model to visualize the air flow distribution inside the compartment. The results show that the velocity distribution is uniform when the diffuser air velocity is 3.1 m/s. When the diffuser air velocity is 3.1 m/s the air temperature of the seat area was decreased by 0.3°C. The air temperature inside the cabin can be maintained uniform at 23°C when diffuser air velocity was fixed at 3.1 m/s.

Multi objective optimization of an evolutionary feedforward neural network for the automotive air conditioning system performance prediction

In this paper, a novel multi-objective evolutionary artificial neural network approach is proposed to predict the performance of an automotive air conditioning (AAC) system. A Feedforward Neural Network (FNN) was used to simulate the cooling capacity and compressor power under different combination of input compressor speeds, evaporator inlet air speeds, air temperature upstream of the condenser and evaporator. Differential Evolution (DE) algorithm was employed to automatically optimize the FNNs parameters, involving the number of hidden layers and the number of neurons in each hidden layer. The training of connection weights and biases is carried out using the basic backpropagation algorithm with Levenberg Marquardt nonlinear optimization method. For the purpose of multi-objective optimization, the DE algorithm is incorporated with two key elements of the NSGA-II (Non-dominated Sorting Genetic Algorithm II), namely the non-dominated sorting method and the crowding distance metric. A parametric study was performed on the proposed algorithm and the best DE base variant was determined. The experimental results show that the proposed algorithm with DE based variant ‘DE/Best/1’ exhibited its superiority in term of prediction performance. The best neural network obtained is FNN with 4×18×2 network configuration and its network complexity is equivalent to 108 connection weights. It yields an average relative error of 0.60% for the prediction of cooling power and one of 3.0% for the prediction of compressor power.

Identification for automotive air-conditioning system using Particle Swarm Optimization

This paper present the representation of the dynamic model of the temperature an automotive air conditioning system (AAC) as the speed of the air conditioning compressor is varied. The performance of system identification of an AAC system using Recursive Least Squares (RLS) and Particle Swarm Optimization (PSO) techniques measured and discussed. The input - output data are collected through an experimental study using an AAC system integrated with air duct system experimental rig complete with data acquisition and instrumentation system. The single input single output dynamic model was established by using Autoregressive with exogenous input (ARX) model. Recursive Least Squares and Genetic Algorithms were validated using one step-ahead prediction (OSA), mean squared error (MSE) and correlation tests. The comparison results between these parameter estimation optimization techniques were highlighted. It was found that the estimated models using these two methods proposed are comparable, acceptable and possible to be used as a platform of new controller development and evaluation the performance of AAC system in the future work. Amongst all, it was found that the Particle Swarm optimization method produce the best ARX model with the lowest prediction MSE value of 8.5472×10-5 as compared to the Recursive Least Squares performance.

Genetic algorithm identification for automotive air-conditioning system

In this study, system identification of an automotive air conditioning (AAC) system with different parameter estimation methods were conducted. The AAC experiment rig which comes complete with an air duct system to simulate the actual behaviour of AAC system was used to acquire the input and output datasets for the identification of the system. The single input single output dynamic model was established by using Autoregressive with exogenous input (ARX) model. Recursive Least Square (RLS) and Genetic Algorithm (GA) were used to optimize the ARX model and hence to obtain the dynamic model of AAC system based on one-step-ahead (OSA) prediction. The performances of the models were validated using statistical analysis based on the mean squares of error (MSE) between the actual and predicted output responses of the models. The comparison results between these parameter estimation optimization techniques were highlighted. The GA optimization method produce the best ARX model with the lowest prediction MSE value of 0.0015059 and it was proposed to be used to represent the AAC system for further development of the controller strategy.

Numerical analysis of air-flow and temperature field in a passenger car compartment

This paper presents a numerical study on the temperature field inside a passengers compartment of a Proton Wira saloon car using computational fluid dynamics (CFD) method. The main goal is to investigate the effects of different glazing types applied onto the front and rear windscreens of the car on the distribution of air-temperature inside the passenger compartment in the steady-state conditions. The air-flow condition in the passengers compartment is also investigated. Fluent CFD software was used to develop a three-dimensional symmetrical model of the passengers compartment. Simplified representations of the driver and one rear passenger were incorporated into the CFD model of the passengers compartment. Two types of glazing were considered namely clear insulated laminated tint (CIL) with a shading coefficient of 0.78 and green insulated laminate tint (GIL) with a shading coefficient of 0.5. Results of the CFD analysis were compared with those obtained when the windscreens are made up of clear glass having a shading coefficient of 0.86. Results of the CFD analysis show that for a given glazing material, the temperature of the air around the driver is slightly lower than the air around the rear passenger. Also, the use of GIL glazing material on both the front and rear windscreens significantly reduces the air temperature inside the passengers compartment of the car. This contributes to a better thermal comfort condition to the occupants. Swirling air flow condition occurs in the passenger compartment. The air-flow intensity and velocity are higher along the side wall of the passengers compartment compared to that along the middle section of the compartment. It was also found that the use of glazing materials on both the front and rear windscreen has no significant effects on the air-flow condition inside the passengers compartment of the car.

Development of vapor pressure in FR4-copper composite material during solder reflow process

This paper presents a finite element (FE) methodology for predicting the distribution of vapor pressure in a simple FR4-copper composite material when it is heated up to 215°C. A general purpose finite element software was used to develop a two-dimensional plane strain model of the composite material. FE simulation of transient moisture absorption was performed to predict the distribution of wetness fraction in the material after pre-conditioning at an 85°C/85%RH environment for 15 days. FE simulation of transient moisture desorption was carried out at the peak solder reflow temperature of 215°C to predict new distribution of wetness fraction in the material. The results of the moisture desorption analysis were used to compute the magnitude of vapor pressure in the material and its distribution at 215°C. It was found that the moisture in the material redistributes itself during solder reflow. The moisture concentration in the area close to the FR4-copper interface below the longer copper trace increases during the solder reflow. The magnitude of the vapor pressure in 70% of the FR4 and near the FR4-copper interface below the lower copper trace is closed to the saturation pressure of water vapor at 215°C. The distribution of the vapor pressure in the material is in similar fashion as the new distribution of wetness fraction after the moisture desorption analysis.