This study presents a machine learning-based framework for detecting critical events in railway infrastructure by analyzing vibration signals from trackside accelerometers. Traditional maintenance is often reactive and labor-intensive, but this approach uses continuous sensing and data analytics to enable proactive, real-time monitoring. The research leverages a comprehensive pipeline that includes data preprocessing, segmentation of time-series data into one-second intervals labeled as "event" or "no-event", and the extraction of statistical, temporal, and spectral features like crest factor and kurtosis. Key contribution of this work is the systematic evaluation of 72 algorithm-feature selection configurations. Twelve diverse classification algorithms were compared, including tree-based, linear, and neural network models. Extensive hyperparameter optimization was performed to benchmark performance using metrics such as accuracy, precision, recall, and F1-score. The Multi-Layer Perceptron (MLPClassifier) achieved a peak cross-validation accuracy of 98.89% with the full feature set. The study also found that comparable accuracy (98.67%) could be achieved with a 47% dimensionality reduction using Recursive Feature Elimination (RFE) with only eight features, demonstrating a balance between efficiency and performance. The findings provide actionable insights for developing scalable, high-performance event detection systems.

Continuum robots offer high dexterity and compliance, which makes them attractive for tasks in confined, hazardous, and hard-to-reach environments. Despite this potential, inverse kinematics (IK) for multi-section continuum robots remains challenging due to strong nonlinearities and redundancy, and the problem becomes more demanding when each section can actively change its backbone length. This paper addresses obstacle-aware IK for a cable-driven variable-length continuum robot by formulating IK as a constrained optimization problem built on a constant-curvature forward kinematic model. A teaching–learning-based optimization (TLBO) algorithm is adopted to search for section bending angles, orientation angles, and section lengths that minimize end-effector tracking error while avoiding static obstacles through a capsule-based penalty constraint handling strategy that accounts for the robot’s physical radial dimension. The approach is evaluated through multiple threedimensional MATLAB simulations, including linear and circular trajectory tracking with and without obstacle avoidance, and is benchmarked against particle swarm optimization (PSO), a real-coded genetic algorithm (GA), and differential evolution (DE) over 30 independent runs. Statistical analysis shows that TLBO achieves the best or near-best tracking accuracy (mean error 4.95×10−7.84×10−mm, best mm) while requiring no algorithm-specific tuning parameters. The method is further validated experimentally on a Continuum Bionic Handling Assistant (CBHA) platform by comparing the IK-derived cable-length profiles with potentiometer-based measurements. The results demonstrate accurate trajectory tracking in simulation and good agreement with experimental cable-length measurements, supporting the feasibility of TLBO for constrained IK of variable-length continuum robots.

Hydrogen gas has gained significant attention as a cleaner alternative to fossil fuels offering a sustainable energy solution. This study explores the production efficiency of oxyhydrogen (HHO) gas using a flat plate electrolyser with potassium hydroxide (KOH) as the electrolyte. Machine learning regression models were employed to estimate hydrogen generation rates and system efficiency based on key operational parameters that includes voltage, current and electrolyte concentration. A set of gradient-boosting algorithms was evaluated utilizing raw experimental data to predict (i) hydrogen output in liters per minute (LPM) and (ii) system efficiency. The results indicate that Categorical Boosting (CatBoost) excelled in forecasting system efficiency (R2 = 0.9748, RMSE = 1.6567 on testing data) and predicting HHO gas generation rate (R2 = 0.9936, RMSE = 0.0090). The experimental results show that with the increase in KOH concentration there is increase in production of Hydrogen. Maximum efficiency was noted with 0.5 N of KOH with the peak efficiency of 99.8 % because of its optimal conductivity and power consumption. It can also be absorbed that higher concentration such 0.75 N and 1 N have shown significant improvement in hydrogen production. Experimental findings further revealed that moderate operating conditions maximize hydrogen production with efficiency varying as a function of applied current and electrolyte concentration. This study highlights the advantages of integrating machine learning models with electrolysis-based hydrogen production offering a scalable and data-driven approach to optimizing energy efficiency. The results underscore the potential of KOH-based electrolysis for sustainable hydrogen generation and reinforce the role of predictive modeling in enhancing system performance.

The integration of oxyhydrogen (HHO) gas into internal combustion (IC) engines has attracted substantial interest among researchers in improving engine performance and reducing emissions. In the present work, a wet-type flat-plate electrolyser utilizing sodium hydroxide (NaOH) as electrolyte is investigated to determine the interdependent effects of voltage, current, and NaOH concentration on HHO gas generation rate and system efficiency. The results show that moderate current and voltage levels, along with higher NaOH concentrations (e.g., 5.87 V and 1 N) yield a maximum gas production rate of 0.5 L/min while conserving energy efficiency. The experimental analysis also showed that as the current increase the rate of production also increased. The maximum production of 0.5 L/min was achieved with 30 A. The study also extends to use experimental data to train machine learning algorithm to estimate the performance of the HHO gas system. Voltage, current, power consumption, resistance and electrolyte concentration were used as input parameters while efficiency and HHO gas production were the output parameters measured with a total dataset size of 112 observations. To reduce the experimental burden and establish an efficient predictive framework five gradient boosting algorithms namely, categorical boosting (CatBoost), extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), adaptive boosting (AdaBoost) and gradient boosting (GB) are evaluated among which CatBoost achieved maximum accuracy with R2 values of 0.9903 (for hydrogen production) and 0.9583 (for efficiency) on test data. The findings highlight how crucial intermediate operating conditions are for optimizing gas output and efficiency while lowering resource usage.

Industrial cybersecurity has become a critical concern in today's interconnected world, as critical infrastructure systems increasingly rely on digital technologies. This paper explores the unique challenges and opportunities presented by industrial cybersecurity, highlighting the need for enhanced cybersecurity measures. The paper discusses the potential consequences of cyberattacks on industrial systems, including disruptions to critical services, economic losses, and even physical harm. To address these challenges, this paper discusses cybersecurity initiatives, standards, guidelines, directives, and acts that can provide a comprehensive framework for cybersecurity and AI governance. A systematic literature review has been conducted in this paper using Scopus and Google Scholar, which provide the foundation for identifying relevant publications. These publications show key trends and themes in industrial cybersecurity research, including the growing importance of education and training, as well as cybersecurity risk assessment and mitigation.