The success of Internet of Things solutions allowed the establishment of new applications such as smart hydroponic agriculture. One typical problem in such an application is the rapid degradation of the deployed sensors. Traditionally, this problem is resolved by frequent manual maintenance, which is considered to be ineffective and may harm the crops in the long run. The main purpose of this thesis was to propose a machine learning approach for automating the detection of sensor fault drifts. In addition, the solution’s operability was investigated in a cloud computing environment in terms of the response time. This thesis proposes a detection algorithm that utilizes RNN in predicting sensor drifts from time-series data streams. The detection algorithm was later named; Predictive Sliding Detection Window (PSDW) and consisted of both forecasting and classification models. Three different RNN algorithms, i.e., LSTM, CNN-LSTM, and GRU, were designed to predict sensor drifts using forecasting and classification techniques. The algorithms were compared against each other in terms of relevant accuracy metrics for forecasting and classification. The operability of the solution was investigated by developing a web server that hosted the PSDW algorithm on an AWS computing instance. The resulting forecasting and classification algorithms were able to make reasonably accurate predictions for this particular scenario. More specifically, the forecasting algorithms acquired relatively low RMSE values as ~0.6, while the classification algorithms obtained an average F1-score and accuracy of ~80% but with a high standard deviation. However, the response time was ~5700% slower during the simulation of the HTTP requests. The obtained results suggest the need for future investigations to improve the accuracy of the models and experiment with other computing paradigms for more reliable deployments.

A sensor is a technological device or biological organ that detects, or senses, a signal or physical condition and chemical compounds. Technological developments in the recent decades have brought along with it several environmental problems and human safety issues to the fore. In today's world, therefore, sensors, which detect toxic and inflammable chemicals quickly, are necessary. Gas sensors which form a subclass of chemical sensors have found extensive applications in process control industries and environmental monitoring. The present thesis reports the attempt made in development of Zinc oxide thin film based gas sensors. ZnO is sensitive to many gases of interest like hydrocarbons, hydrogen, volatile organic compounds etc. They exhibit high sensitivity, satisfactory stability and rapid response. In the present work the developed sensors have been tested for their sensitivity for a typical volatile organic compound, acetone. An objective analysis of the various substrates namely borosilicate glass, sintered alumina and hard anodized alumina, has been performed as a part of this work. The substrates were evaluated for their electrical insulation and thermal diffusivity. The microstructure of the gas sensitive film on the above mentioned substrates was studied by SEM technique. The gas sensitive Zinc oxide film is deposited by D.C reactive magnetron sputtering technique with substrate bias arrangement. The characterization of the as-deposited film was performed by XRD, SEM and EDAX techniques to determine the variation of microstructure, crystallite size, orientation and chemical composition with substrate bias voltage. The thesis also describes the development of the gas sensor test setup, which has been used to measure the sensing characteristics of the sensor. It was observed that the ZnO sensors developed with higher bias voltages exhibited improved sensitivity to test gas of interest. Gas sensors essentially measure the concentration of gas in its vicinity. In order to determine the distribution of gas concentration in a region, it is necessary to network sensors at remote locations to a host. The host acts as a gateway to the end user to determine the distribution of gas concentration in a region. However, wireless gas sensor networks have not found widespread use because of two inherent limitations: Metal oxide gas sensors suffer from output drift over time; frequent recalibration of a number of sensors is a laborious task. The gas sensors have to be maintained at a high temperature to perform the task of gas sensing. This is power intensive operation and is not well suited for wireless sensor network. This thesis reports an exploratory study carried out on the applicability of gas sensors in wireless gas sensor network. A simple prototype sensing node has been developed using discrete electronic components. A methodology to overcome the problem of frequent calibration of the sensing nodes, to tackle the sensor drift with ageing, is presented. Finally, a preliminary attempt to develop a strategy for using gas sensor network to localize the point of gas leak is given.

Chemical sensor based systems that are used for detection, identification, or quantification of various gases are very complex in nature. Sensor response data collected as a multivariate time series signals encounters gradual change of the sensor characteristics(known as sensor drift) due to several reasons. In this thesis, drift component of a silicon carbide Field-Effect Transistor (SiC-FET) sensor data was analyzed using time series. The data was collected from an experiment measuring output response of the sensor with respect to gases emitted by certain experimental object at different temperatures. Augmented Dickey Fuller Test (ADF) was carried out to analyze the sensor drift which revealed that stochastic trend along with deterministic trend characterized the drift components of the sensor. The drift started to rise in daily measurements which contributed to the total drift. / Traditional Autoregressive Integrated Moving Average (ARIMA) and deep learning based Long Short-Term Memory (LSTM) algorithm were carried out to forecast the sensor drift in reduced set of data. However, reduction of the data size degraded the forecasting accuracy and imposed loss of information. Therefore, careful selection of data using only one temperature from the temperature cycle was chosen instead of all time points. This chosen data from sensor array outperformed forecasting of sensor drift than reduced dataset using both traditional and deep learning methods.