Herman Wahid

Neural network-based meta-modelling approach for estimating spatial distribution of air pollutant levels

Continuous measurements of the air pollutant concentrations at monitoring stations serve as a reliable basis for air quality regulations. Their availability is however limited only at locations of interest. In most situations, the spatial distribution beyond these locations still remains uncertain as it is highly influenced by other factors such as emission sources, meteorological effects, dispersion and topographical conditions. To overcome this issue, a larger number of monitoring stations could be installed, but it would involve a high investment cost. An alternative solution is via the use of a deterministic air quality model (DAQM), which is mostly adopted by regulatory authorities for prediction in the temporal and spatial domain as well as for policy scenario development. Nevertheless, the results obtained from a model are subject to some uncertainties and it requires, in general, a significant computation time. In this work, a meta-modelling approach based on neural network evaluation is proposed to improve the estimated spatial distribution of the pollutant concentrations. From a dispersion model, it is suggested that the spatially-distributed pollutant levels (i.e. ozone, in this study) across a region under consideration is a function of the grid coordinates, topographical information, solar radiation and the pollutants precursor emission. Initially, for training the model, the input-output relationship is extracted from a photochemical dispersion model called The Air Pollution Model and Chemical Transport Model (TAPM-CTM), and some of those input-output data are correlated with the ambient measurements collected at monitoring stations. Here, improved radial basis function networks, incorporating a proposed technique for selection of the network centres, will be developed and trained by using the data obtained and the forward selection approach. The methodology is then applied to estimate the ozone concentrations in the Sydney basin, Australia. Once executed, apart from the advantage of inexpensive computation, it provides more reliable results of the estimation and offers better predictions of ozone concentrations than those obtained by using the TAPM-CTM model only, when compared to the measurement data collected at monitoring stations.

Enhanced radial basis function neural networks for ozone level estimation

Assessment of air pollutant profiles by using measurements involves some limitations in the implementation. For this, deterministic air quality models are often used. However, its simulation usually needs high computational requirements due to complex chemical reactions involved. In this paper, a neural network-based metamodel approach is used in conjunction with a deterministic model and some measured data to approximate the non-linear ozone concentration relationship. For this, algorithms for performance enhancement of a radial basis function neural network (RBFNN) are developed. The proposed method is then applied to estimate the spatial distribution of ozone concentrations in the Sydney basin. The experimental comparison between the proposed RBFNN algorithm and the conventional RBFNN algorithm demonstrates the effectiveness and efficiency in estimating the spatial distribution of ozone level.

A metamodel for background ozone level using radial basis function neural networks

In air quality modelling, determination of the background ozone level is essential as it highly affects the accuracy of the photochemical air quality model. It is known that the background ozone level, especially in urban areas, has been changing over the years. Unfortunately, the reasons of that alteration were not clear and the background ozone itself was not easily derived in practice. In this paper, a new background ozone model will be developed by using the ozone ambient quality data and the meteorological data at the several stations in the Sydney basin. To accomplish the modelling process, an adaptively-tuned radial basis function neural network metamodel is proposed and utilised in the simulation. Different input parameters are considered to evaluate their influence on the constructed background ozone model. The proposed model, subject to some statistical criteria, demonstrates its capability of estimating the background ozone level with a reasonably good accuracy.

A charge pump-based power conditioning circuit for low powered thermoelectric generator (TEG)

Thermoelectric generator (TEG) is a sensor that utilizes thermal gradients between cold plate and hot plate of the sensor and convert it into electricity. By having a concise and non-moving structure, the sensor attracts interest in studies to implement an autonomous self sustain system. Power generation of TEG is directly proportional to temperature gradient. Therefore, the device is dismissed when the gradient source of heat is small. This limits the aim of autonomous self sustain system when it uses human temperature as source of conversion. This paper focused on the power conditioning system for a low powered thermoelectric generator. At the same time, it also determines the viability of TEG application in human portable device by adapting body temperature as source of energy conversion. This paper reveals the feasibility of signal conditioning method for a regular TEG module using charge pump.

Computational intelligence estimation of natural background ozone level and its distribution for air quality modelling and emission control

Background ozone, known as the ozone that occurs in the troposphere as a result of biogenic emissions without photochemical influences, has a close relationship with human health risk. The prediction of the background ozone level by an air quality model could cover a wider region, whereas a measurement method can only record at monitoring sites. The problem is that simulation with deterministic models is quite tedious because of the nonlinear nature of some particular chemical reactions involved in the pollutant formulation. In this work, we present a reliable method for determination of the background ozone using the ambient measurement data. Our proposed definition can be used to determine the background level at any part of the globe and in any seasons without relying on data obtained at remote sites. A statistical model approach will be used for the estimation of the background ozone concentration, and a method for extrapolating the site data will be utilised to approximate the spatial distribution on the region. The proposed method will be applied in the Sydney basin to evaluate its effectiveness in background ozone determination. The results show the advantage of the proposed approach as a globally generic and computationally efficient way for the background ozone estimation with a reasonable accuracy.

Adaptive Neural Network Metamodel for Short-Term Prediction of Background Ozone Level

Research has been performed recently on spherical wave modelling of the radio propagation channels. In such a modelling the idea is to consider the responses of the radio channel for different spherical wave mode fields. A natural scan surface for the direct spherical data extraction would be spherical. However, many laboratories are equipped not with spherical scanners but with linear scanners. This paper describes how the required spherical mode responses of a channel can be extracted from the measurement data acquired on a cubical scanning surface. Rules for the required number of sampling locations on a cubical surface are provided.Modelling is important in air quality forecasting and control. Before applying an air quality model, it is required to accurately estimate the biogenic emission. The assessment of the background ozone concentration is essential for this estimation. It has been known that the biogenic ozone level in urban areas is changing over the years, and hence information about the temporal trends in air quality data is helpful for the assessment. This paper presents a neural-network metamodel for prediction of the background ozone level in the Sydney basin. Based on measured monitoring data under non-photochemical conditions collected at a number of monitoring stations, the proposed model can reliably provide short-term predictions in the biogenic ozone trends to be used for analysis of ground-level emission impact on air quality.

Determination of Modeling Parameters for a Low Cost Air Pollution Measurement System Using Feedforward Neural Networks

Air pollution model is commonly used to predict the pollutant level in the air for the upcoming days based on the previous data. In this paper, a new model for predicting ozone, nitrogen dioxide and sulphur dioxide will be developed using the previous data of pollutants agents such as carbon monoxide, sulphur dioxide, nitrogen dioxide, ozone, particulate matter and the meteorological data includes wind speed, temperature and humidity. It is developed to improve the estimation values for a low cost setup of air pollution measurement system. The models are constructed using the Levenberg-Marquardt training algorithms in the neural network tool. Different input parameters are investigated to develop better performance model for predicting air pollution. The proposed model is capable to predict the air pollution level with high accuracy and the meteorological data are dominantly influenced the accuracy of the model.

Modeling, Design and Simulation of Systems

Image segmentation is the most important steps in image processing process, especially in detecting and segmenting the main focus object from its background or others unwanted image. The objective of this paper is to develop a segmentation technique for pineapple fruit from crop background at the plantation level. Hue value is used to remove the ground and sky from the image. Then, Adaptive Red and Blue chromatic (ARB) is implemented to segmenting the pineapple fruit from the background such as leaves. In this case, the ARB method is still produced misclassifies error. Further segmentation uses Ellipse Hough Transform (EHT) for results enhancement, so that the fruit’s image is completely filtered from misclassify and background. The results obtained show that the proposed technique manages to identify the fruit from the background with better image output compared to conventional method.

Modelling and Control of a Non-linear Inverted Pendulum Using an Adaptive Neuro-Fuzzy Controller

The inverted pendulum system is a benchmark system for testing the performance of different control algorithms. Since it is a non-stable system, a continuously corrected mechanism is needed to move the cart in a certain way in order to balance the pendulum and prevent it from falling due to gravity. In this paper, a four-input adaptive neuro-fuzzy controller is used to control the system in a short time. The controller is implemented in MATLAB Simulink and its performance was compared with the Sugeno-type fuzzy controller. It is found that the Adaptive Neuro-fuzzy controller provides better performance as it has almost no overshoot compared to the Sugeno-type controller. Moreover, its execution time is much less than the time needed for the Sugeno-type fuzzy controller.

Detecting mechanism of planar electromagnetic sensor in cooking oil discrimination

The methods of cooking oil discerning are important for the sustainability of the cooking oils supply chain. Classical methods to measure cooking oil quality based on gas chromatography (GC) and high-performance liquid chromatography (HPLC) are too expensive for widespread industrial or household uses and require samples to be analysed in dedicated laboratories thus incurring significant time penalty. Therefore, efforts to develop a reliable and highly sensitive system that may provide on-site cooking oil discrimination is an important priority. Presented research work describes a real time non-invasive discriminating technique for cooking oils. A planar electromagnetic sensor consists of the combination of two sensors which are a meander sensor and an interdigital sensor. The meander type of coil is connected in series with the interdigital coil. Due to the ac voltage applied, the combination of meander and interdigital coils produce electromagnetic field which interacts with the material under test. Three types of cooking oil were prepared namely corn oil, refined palm oil, and sunflower oil. The impedance spectra so obtained showed that the sensor was able to detect the difference between different types of cooking oil samples. Both theory formulation and experiment outcomes were used to testify the measurement for cooking oil discrimination.