Gas identification/classification through pattern recognition techniques based on gas sensor arrays often requires the equilibrium responses or the full traces of time-series data of the sensor array. Leveraging upon the diverse gas sensing kinetics behaviors measured via the sensor array, a computational intelligence- based meta-model is proposed to automatically conduct the feature extraction and subsequent gas identification using time-series data during the transitional phase before reaching equilibrium. The time-series data contains implicit temporal dependency/correlation that is worth being characterized to enhance the gas identification performance and reliability. In this context, a tailored approach so-called convolutional long short-term memory (CLSTM) neural network is developed to perform the identification task incorporating temporal characteristics within time-series data. This novel approach shows the enhanced accuracy and robustness as compared to the baseline models, i.e., multilayer perceptron (MLP) and support vector machine (SVM) through the comprehensive statistical examination. Specifically, the classification accuracy of CLSTM reaches as high as 96%, regardless of the operating condition specified. More importantly, the excellent gas identification performance of CLSTM at early stages of gas exposure indicates its practical significance in future real-time applications. The promise of the proposed method has been clearly illustrated through both the internal and external validations in the systematic case investigation.

Fault diagnosis of rolling bearings becomes an important research subject, where the data-driven deep learning-based techniques have been extensively exploited. While the state-of-the-art research has shown the substantial progresses in bearing fault diagnosis, they mostly were implemented upon the hypothesis that the location of bearing prone to failure already is known. Nevertheless, in actual practice many rolling bearings are installed in a complex machinery system, any of which is likely subject to fault. As such, fault diagnosis essentially is a process to achieve both fault localization and identification, which results in many fault scenarios to be handled. This will significantly degrade the fault diagnosis performance using conventional deep learning analysis. In this research, we aim to develop a new deep learning framework to address abovementioned challenge. We particularly design a hierarchical deep learning framework consisting of multiple sequentially deployed deep learning models built upon the transfer learning. This can improve the learning adequacy for a high-dimensional problem with many fault scenarios involved even under limited dataset, thereby enhancing the fault diagnosis performance. Without the prior knowledge regarding the fault location, this methodology is greatly favored by the sensor/data fusion which takes full advantage of the enriched pivot fault-related features in the measurements acquired from different accelerometers. Systematic case studies using the publicly accessible experimental rolling bearing dataset are carried out to validate this new methodology.