There has been strong demand for the development of an accurate but simple method to assess the freshness of food. In this study, we demonstrated a system to determine food freshness by analyzing the spectral response from a portable visible/near-infrared (VIS/NIR) spectrometer using the Convolutional Neural Network (CNN)-based machine learning algorithm. Spectral response data from salmon, tuna, and beef incubated at 25 °C were obtained every minute for 30 h and then categorized into three states of “fresh”, “likely spoiled”, and “spoiled” based on time and pH. Using the obtained spectral data, a CNN-based machine learning algorithm was built to evaluate the freshness of experimental objects. In addition, a CNN-based machine learning algorithm with a shift-invariant feature can minimize the effect of the variation caused using multiple devices in a real environment. The accuracy of the obtained machine learning model based on the spectral data in predicting the freshness was approximately 85% for salmon, 88% for tuna, and 92% for beef. Therefore, our study demonstrates the practicality of a portable spectrometer in food freshness assessment.

Gas leaks can cause disasters at process sites, including fires and explosions, and thus, effective gas-leak detection systems are required. This study investigated the limitations of conventional detectors and introduced an innovative ultrasonic sensor-based approach for continuous monitoring. A new configuration for a stationary remote ultrasonic gas-leak monitoring system is proposed. The selected material was 1-Butene. The detection probability was assessed through a simulation based on a gas-leak scenario, detailing the selection criteria for leak sites and simulation conditions. Computational fluid-dynamics (CFD) simulations were used to evaluate the detection capability of the existing system, whereas Monte Carlo simulations were used to compare it with the proposed ultrasonic system. The CFD simulation was performed by setting the lower detection limit of the concentration-measurement-type gas detector to 600 ppm, and the leak-detection time was approximately 8.895 s. A Monte Carlo simulation accounting for wind direction showed that the existing system’s average detection probability was 3.41%. By integrating the existing and new systems, the detection probability increased by 96.58% and 3.92% compared with that using only the concentration-measurement and ultrasonic measurement methods, respectively. These rates provide valuable insights for future applications in oil refining and petrochemical industries.

Leakage of flammable gases and the resulting explosions in petrochemical plants remain latent risks, capable of occurring at any moment. Therefore, to address these worst-case scenarios within a virtual reality framework, we conducted simulations aimed at predicting and effectively responding to potential damages due to gas leakage. This study presents an analysis of the hazards that can lead to leaks and potential explosions in a petrochemical plant using the BREEZE Incident Analyst program. Rapid and accurate recognition of the risk associated with gas leaks, which can cause extensive damage and explosions, is of paramount importance. This study addresses two main aspects: the prediction of the consequences of gas leaks through simulations and the implementation of appropriate detection measures. Better, more efficient risk management and mitigation strategies were implemented by predicting gas leak paths using BREEZE. Using ultrasonic detection technology, detection was demonstrated to be possible in approximately one-third the time required by conventional detectors, and it is weather-insensitive. Simultaneously, considering plant characteristics such as utility configurations, we propose an additional method to prevent leaks from going undetected. This is achieved by integrating gas detection technology that combines both existing and new technologies.