Integration of machine learning and first principles models

Model building and parameter estimation are traditional concepts widely used in chemical, biological, metallurgical, and manufacturing industries. Early modeling methodologies focused on mathematically capturing the process knowledge and domain expertise of the modeler. The models thus developed are termed first principles models (or white-box models). Over time, computational power became cheaper, and massive amounts of data became available for modeling. This led to the development of cutting edge machine learning models (black-box models) and artificial intelligence (AI) techniques. Hybrid models (gray-box models) are a combination of first principles and machine learning models. The development of hybrid models has captured the attention of researchers as this combines the best of both modeling paradigms. Recent attention to this field stems from the interest in explainable AI (XAI), a critical requirement as AI systems become more pervasive. This work aims at identifying and categorizing various hybrid models available in the literature that integrate machine-learning models with different forms of domain knowledge. Benefits such as enhanced predictive power, extrapolation capabilities, and other advantages of combining the two approaches are summarized. The goal of this article is to consolidate the published corpus in the area of hybrid modeling and develop a comprehensive framework to understand the various techniques presented. This framework can further be used as the foundation to explore rational associations between several models.     

Optimal position and rating of DG in distribution networks by ABC–CS from load flow solutions illustrated by fuzzy-PSO

In this work, load flow problems of both radial distribution networks (RDNs) and mesh distribution networks (MDNs) have been solved using hybrid fuzzy-PSO algorithm. A new voltage stability index (VSI) is also indicated. Based on the suggested load flow, distributed generation (DG) is ready to conduct through the requirement; and with the support of inserting the optimal-sized DG unit in an exact way, the distribution system’s stability is also studied. The exact position of each DG unit has been computed using “loss sensitivity analysis,” whereas the optimal sizing of each DG unit has been done with the help of hybrid artificial bee colony and Cuckoo search algorithm. The suggested method is tested in the regular 33-node and 69-node RDNs as well as in 85-node and 119-node MDNs. The transcendence of the proposed operation has been centered with the aid of comparison to the other existing methods. The suggested VSI is also correlated with other two existing VSIs before and after placement of DG unit(s).

     

Highly birefringent large negative dispersion-flattened photonic crystal fibre for broadband residual dispersion compensation

A triangular lattice photonic crystal fibre is presented in this paper for residual dispersion compensation. The fibre exhibits a flattened negative dispersion of ?992.01 ± 6.93 ps/(nm-km) over S+C+L wavelength bands and ?995.83 ± 0.42 ps/(nm-km) over C-band. The birefringence is about 4.4 × 10^?2 at the excitation wavelength of 1550 nm which is also very high. Full vector finite element method (FEM) with a perfectly matched absorbing layer (PML) boundary condition is applied to numerically investigate the guiding properties of this PCF. The fibre operates at fundamental mode only. All these properties endorse this fibre as a suitable candidate for compensating residual dispersion and polarization maintaining applications.     

Intelligent heart disease prediction in cloud environment through ensembling

Cloud computing is the delivery of on‐demand computing resources. Cloud computing has numerous applications in fields of education, social networking, and medicine. But the benefit of cloud for medical purposes is seamless, particularly because of the enormous data generated by the health care industry. This colossal data can be managed through big data analytics, and hidden patterns can be extracted using machine learning procedures. In particular, the latest issue in the medical domain is the prediction of heart diseases, which can be resolved through culmination of machine learning and cloud computing. Hence, an attempt has been made to propose an intelligent decision support model that can aid medical experts in predicting heart disease based on the historical data of patients. Various machine learning algorithms have been implemented on the heart disease dataset to predict accuracy for heart disease. Naïve Bayes has been selected as an effective model because it provides the highest accuracy of 86.42% followed by AdaBoost and boosted tree. Further, these 3 models are being ensembled, which has increased the overall accuracy to 87.91%. The experimental results have also been evaluated using 10,082 instances that clearly validate the maximum accuracy through ensembling and minimum execution time in cloud environment.     

Intensified and safe ozonolysis of fatty acid methyl esters in liquid CO2 in a continuous reactor

We demonstrate a continuous reactor for performing the ozonolysis of fatty acid methyl esters (FAMEs) using liquid CO₂ as solvent. The fast reaction kinetics allows the use of small‐volume reactors to completely convert the FAMEs, forming secondary ozonides as the primary products. The short residence times also help maximize the yields of the secondary ozonides by minimizing over‐oxidation and the formation of oligomeric products. The liquid CO₂ medium promotes safe reactor operation by providing an essential fraction of overall reactor cooling and by diluting the vapor phase organics. We also demonstrate a continuous stirred reactor for the safe thermal decomposition of the secondary ozonides to their corresponding acids and aldehydes. Using a lumped kinetic model for the thermal decomposition of the ozonolysis products, we estimate activation energy values of 108.6 ± 0.6 kJ mol^?1 for the decomposition of secondary ozonides and 122 ± 3 kJ mol^?1 for the decomposition of the undesired oligomeric species. © 2017 American Institute of Chemical Engineers AIChE J, 63: 2819–2826, 2017     

Corrosion behaviour of a thin section martensitic stainless steel GTA weldment in chloride solutions

A thin section martensitic stainless steel was welded by gas tungsten arc welding and characterized for the microstructure, hardness and corrosion behaviour in chloride solutions. Welds free from defects could be produced by autogenous welding under the optimized welding conditions. The weld metal was over-matched in terms of mechanical properties (hardness, tensile strength). The general corrosion resistance and the passivation behaviour of the weld metal/heat affected zone (HAZ) region were on par with that of the parent material in chloride and sulphuric acid test electrolytes; however, in terms of pitting corrosion resistance, the martensitic-structured weld metal/HAZ region was marginally inferior compared to its parent material.     

Effect of plasma electrolytic oxidation treatment on the corrosion and stress corrosion cracking behaviour of AM50 magnesium alloy

The effect of a silicate-based plasma anodization treatment on the corrosion and stress corrosion cracking behaviour of a cast AM50 magnesium alloy was studied. Electrochemical tests revealed the beneficial effect of the plasma electrolytic oxidation (PEO) in improving the corrosion resistance of the alloy. Although the coating had provided an improved resistance to stress corrosion cracking in this test environment at a nominal strain rate of 10⁻⁶ s⁻¹, it could not completely eliminate the SCC susceptibility of the alloy. Cracking of the coating under conditions of straining was found to be the reason for SCC of PEO-coated alloy.     

PF-ID-2PAKA: Pairing Free Identity-Based Two-Party Authenticated Key Agreement Protocol for Wireless Sensor Networks

Wireless Personal Communications - To ensure secure communication between any two entities, authenticated key agreement protocol is the primary step and current research has a lot of contribution...     

Biosorption of Cr(VI) by immobilized biomass of two indigenous strains of cyanobacteria isolated from metal contaminated soil

Biosorption of Cr(VI) using native strains of cyanobacteria from metal contaminated soil in the premises of textile mill has been reported in this paper. Biosorption was studied as a function of pH (1-5), contact time (5-180 min) and initial chromium ion concentration (5-20mg/l) to find out the maximum biosorption capacity of alginate immobilized Nostoc calcicola HH-12 and Chroococcus sp. HH-11. The optimum conditions for Cr(VI) biosorption are almost same for the two strains (pH 3-4, contact time 30 min and initial chromium concentration of 20mg/l) however, the biomass of Chroococcus sp. HH-11 was found to be more suitable for the development of an efficient biosorbent for the removal of Cr(VI) from wastewater, as it showed higher values of q(m) and K(f), the Langmuir and Freundlich isotherm parameters. Both the isotherm models were suitable for describing the biosorption of Cr(VI) by the cyanobacterial biosorbents.