Fuel Cells
Mahdi Keyhanpour; Majid Ghassemi
Abstract
Researchers encounter difficulties in producing clean energy and addressing environmental issues. Solid oxide fuel cells (SOFCs) present a promising prospect to the growing demand for clean and efficient electricity due to their capacity to convert chemically stored energy into electrical energy directly. ...
Read More
Researchers encounter difficulties in producing clean energy and addressing environmental issues. Solid oxide fuel cells (SOFCs) present a promising prospect to the growing demand for clean and efficient electricity due to their capacity to convert chemically stored energy into electrical energy directly. In enhancing this technology, ammonia is employed as a cost-effective and carbon-free fuel with convenient transport capabilities. Efficiently predicting the performance of a system in relation to its operating environment has the potential to expedite the identification of the optimal operating conditions across a broad spectrum of parameters. For this purpose, the performance of intermediate temperature solid oxide fuel cell (IT-SOFC) with inlet ammonia fuel is predicted utilizing machine learning, which is efficient in time and cost. Initially, the system is simulated with computational fluid dynamics finite element code to generate data for training machine learning algorithms (DNN, RFM and LASSO regression), followed by an evaluation of the predictive accuracy of these algorithms. The analysis demonstrates that the three examined algorithms exhibit sufficient accuracy in predicting the performance of the introduced solid oxide fuel cell (SOFC) system, all surpassing a 95 percent threshold. The RFM and DNN exhibit the most accurate predictions for the maximum temperature and power density of fuel cells, respectively.
Vibration
A. W. K. Fahmi; K. Reza Kashyzadeh; S. Ghorbani
Abstract
This research investigates the effectiveness of various vibration data acquisition techniques coupled with different machine learning models for detecting anomalies and classifying them. To this end, synthetic vibration data was generated for techniques such as eddy current proximity transducers (ECPT), ...
Read More
This research investigates the effectiveness of various vibration data acquisition techniques coupled with different machine learning models for detecting anomalies and classifying them. To this end, synthetic vibration data was generated for techniques such as eddy current proximity transducers (ECPT), accelerometer sensor, blade tip timing, laser doppler vibrometer (LDV), and strain gauge. Afterward, the data was pre-processed and used to train gradient boosting machine, support vector machine, and random forest models. Performance evaluation metrics, including accuracy, recall, F1-score, receiver operating characteristic, and area Under curve were employed to assess the models, revealing varying degrees of success across combining techniques and models. Notable achievements were observed for the random forest model coupled with the eddy current proximity transducers technique, underscoring the significance of informed technical selection and model optimization in enhancing vibration anomaly detection systems in combined cycle power plants. The results showed that the LDV technique has a significant increase in accuracy from about 0.49 to approximately 0.52, while the ECPT technique has improved from about 0.9 to close 1.0. These advances highlight the growing accuracy of the methods and enable the development of more efficient and reliable learning machines.
Machining
A. Amith Gadagi; B. Chandrashekar Adake
Abstract
In a turning process, it is essential to predict and choose appropriate process parameters to get a component’s proper surface roughness (Ra). In this paper, the prediction of Ra through the artificial neural network (ANN), multiple regression analysis (MRA), and random forest method (machine learning) ...
Read More
In a turning process, it is essential to predict and choose appropriate process parameters to get a component’s proper surface roughness (Ra). In this paper, the prediction of Ra through the artificial neural network (ANN), multiple regression analysis (MRA), and random forest method (machine learning) are made and compared. Using the process variables such as feed rate, spindle speed, and depth of cut, the turning process of glass fiber-reinforced plastic (GFRP) composite specimens is conducted on a conventional lathe with the help of a single-point HSS turning tool brazed with a carbide tip. The surface roughness of turned GFRP components is measured experimentally using the Talysurf method. By utilizing Taguchi's L27 array, the experiments are carried out and the experimental results are utilized in the development of MRA, ANN, and random forest method models for predicting the Ra. It is observed that the mean absolute error (MAE) of MRA, ANN and random forest for the training cases are found to be 39.33%, 0.56%, and 24.88%, respectively whereas for the test cases MAE is 54.34%, 2.59%, and 24.88% for MRA, ANN, and random forest, respectively.
