Mohammad Hossein Karimi Darvanjooghi

   Monitoring the Environmental processes such as biomining and bio-oxidation within the introduction of Artificial Intelligence (AI) is scarce due to the complexity of the bio-oxidation environment and the nature of the microorganisms involved in bioleaching. Therefore, combining advanced models for biooxidation operations and equipment as well as the management of the big data from the laboratory to pilot scale can help to better monitor the bio-oxidation process and to provide experimental validation and a better parameters control for optimal results. Monitoring and optimization of parameters in real-time can help to have decision actions if any variations occur during the oxidation of ores allowing further quick test scenarios to improve the bioleaching operations. The use of Machine Learning (ML) and AI can help to improve complex biological processes in the mining industry including bio-oxidation and bioleaching.

​

   Tools have shown their success recently in Escondida Mine and have led to the establishment of a Decision Support System designed to provide recommendations based on similar historic operations. A synergistic collaboration between scientists, industries, providers of modeling software, plant automation, and control systems is recommended for better process efficiency and management of big data with advanced tools for data analysis, machine learning and digital twins to improve operational performance and scale-up bio-oxidation technology.

 

   It is obvious that ML and AI will be other approaches to deliver the optimization of biotechnological processes. In this method, different mathematical methods will be used to apply powerful interpolation methods on data which is provided by changes in the input parameters. The overall input parameters are divided into two main categories of ore and microorganisms’ characteristics and operational parameters. Then by applying different types of data analysis methods including both hybrid methods or a single approach a mathematical model will be developed for forecasting the output parameters and performing sensitivity analysis and optimization algorithm for efficient utilization in particular in bio-oxidation of refractory gold ores. The only challenge of these methods is entitled to the wide range of input data and the number of input variables that must be considered for the production of highly accurate mathematical relations to predicting the output parameters. However, this method will be a pioneer engineered approach in the data interpretation and optimization in the biotechnological process. It will also influence bio-oxidation of refractory gold ores due to the complexity and the difficulty of mechanisms interpretation which is attributed to bacterial growth rate, mineral properties of ores, operational parameters.

​

    My experience in the field of Machine Learning is related to the optimization and generilization of data interpretation in both biotechnological processes and transport phenomena. The only factor that influence the selection of data-driven models is the complexity of the steps involved in the problem and the production. Therefore, considering the following parts in my research program not only help us to understand the behaviour of the natural phenomena but also leads us to find the optimum oprational condition. The following research program was done by implementation of Machine learning techniques:

 

• Utilization of hybrid AI (Neural Network and Group Method of Data Handling) for effcienct separation of light rare earth elemnts by using superadsorbent magnetic-nanocomposite: in this project various factor were chosen as input and the correspondance data were gathered from experiment for training the model. The input parameters were ion’s electronegativity, molecular weight, initial concentration, and time. The findings presented in this research project postulated that deviations of estimated values from experimental data are mostly less than 12.5% and it is less than 25% for training and testing data. Therefore, it can be mentioned that the proposed correlation can calculate light rare earth elements recovery in a time range of 0–120 min, molecular weight range of 138.9–144.4 g/mol, electronegativity range of 1.10–1.14, and different initial ions’ concentrations of 100, 250, and 500 mg/L.

• Implementation of GMDH-based Neural Network for prediction of transport properties of nanofluids and boiling phenomenon: in this research various parameters such as nanoparticle size, concentration, physical properties as well as the base fluid properties were chosen as input parameter of model. The optimization was performed to obtain the condition where the optimum nanofluid properties was obtained. 

• Co-metabolic biodegradation of recalcitrant aromatic hydrocarbons and optimization method by using machine learning techniques: The degradation of organic maters via microoogranisms is dependent to the complex and un-explored reaction chains. Therefore, the only method that can be used for obtaining optimum condition and do decision making on the biotechnological processes are data-driven model based on machine learning techniques. In this research project we did the modeling part of the expriment and developed it by considering the chemical structure of substrate, to be degraded by using Pseudomonas putida. A single parameter was defined to correlate the chemical structure of substrates (Toluene, Benzene, Ethylbenzene) to the numeric values and the concentration of these substrate as well as time were also used as inputs to the GMDH-based Neural Network model. 

​

    The pltaform that was helpful to develop data-driven model for these research projects were Python, R, and SQL*. 

​

​

​

 

     *For more information please contact.

​

​

​