Multivariate video analysis and Gaussian process regression model based soft sensor for online estimation and prediction of nickel pellet size distributions

Machine learning based soft sensor model for BOD estimation using intelligence at edge

Real-time water quality monitoring is a complex system as it involves many quality parameters to be monitored, the nature of these parameters, and non-linear interdependence between themselves. Intelligent algorithms crucial in building intelligent systems are good candidates for building a reliable and convenient monitoring system. To analyze water quality, we need to understand, model, and monitor the water pollution in real time using different online water quality sensors through an Internet of things framework. However, many water quality parameters cannot be easily measured online due to several reasons such as high-cost sensors, low sampling rate, multiple processing stages by few heterogeneous sensors, the requirement of frequent cleaning and calibration, and spatial and application dependency among different water bodies. A soft sensor is an efficient and convenient alternative approach for water quality monitoring. In this paper, we propose a machine learning-based soft sensor model to estimate biological oxygen demand (BOD), a time-consuming and challenging process to measure. We also propose a system architecture for implementing the soft sensor both on the cloud and edge layers, so that the edge device can make adaptive decisions in real time by monitoring the quality of water. A comparative study between the computational performance of edge and cloud nodes in terms of prediction accuracy, learning time, and decision time for different machine learning (ML) algorithms is also presented. This paper establishes that BOD soft sensors are efficient, less costly, and reasonably accurate with an example of a real-life application. Here, the IBK ML technique proves to be the most efficient in predicting BOD. The experimental setup uses 100 test readings of STP water samples to evaluate the performance of the IBK technique, and the statistical measures are reported as correlation coefficient = 0.9273, MAE = 0.082, RMSE = 0.1994, RAE = 17.20%, RRSE = 37.62%, and edge response time = 0.15 s only.

Model Calibration of Stochastic Process and Computer Experiment for MVO Analysis of Multi-Low-Frequency Electromagnetic Data

An electromagnetic (EM) technique is employed in seabed logging (SBL) to detect offshore hydrocarbon-saturated reservoirs. In risk analysis for hydrocarbon exploration, computer simulation for subsurface modelling is a crucial task. It can be expensive and time-consuming due to its complicated mathematical equations, and only a few realizations of input-output pairs can be generated after a very lengthy computational time. Understanding the unknown functions without any uncertainty measurement could be very challenging as well. We proposed model calibration between a stochastic process and computer experiment for magnitude versus offset (MVO) analysis. Two-dimensional (2D) Gaussian process (GP) models were developed for low-frequencies of 0.0625–0.5 Hz at different hydrocarbon depths to estimate EM responses at untried observations with less time consumption. The calculated error measurements revealed that the estimates were well-matched with the computer simulation technology (CST) outputs. Then, GP was fitted in the MVO plots to provide uncertainty quantification. Based on the confidence intervals, hydrocarbons were difficult to determine especially when their depth was 3000 m from the seabed. The normalized magnitudes for other frequencies also agreed with the resulting predictive variance. Thus, the model resolution for EM data decreases as the hydrocarbon depth increases even though multi-low frequencies were exercised in the SBL application.

Optimal Soybean (Glycine max L.) Land Suitability Using GIS-Based Multicriteria Analysis and Sentinel-2 Multitemporal Images

Soybean is regarded as one of the most produced crops in the world, presenting a source of high-quality protein for human and animal diets. The general objective of the study was to determine the optimal soybean land suitability and conduct its mapping based on the multicriteria analysis. The multicriteria analysis was based on Geographic Information System (GIS) and Analytic Hierarchy Process (AHP) integration, using Sentinel-2 multitemporal images for suitability validation. The study area covered Osijek-Baranja County, a 4155 km2 area located in eastern Croatia. Three criteria standardization methods (fuzzy, stepwise and linear) were evaluated for soybean land suitability calculation. The delineation of soybean land suitability classes was performed by k-means unsupervised classification. An independent accuracy assessment of calculated suitability values was performed by a novel approach with peak Normalized Difference Vegetation Index (NDVI) values, derived from four Sentinel-2 multispectral satellite images. Fuzzy standardization with the combination of soil and climate criteria produced the most accurate suitability values, having the top coefficient of determination of 0.8438. A total of 14.5% of the study area (602 km2) was determined as the most suitable class for soybean cultivation based on k-means classification results, while 64.3% resulted in some degree of suitability.