New statistical machine learning methods applied to high dimensional sensory input data from chemistry

Machine learning is the concept where a computer can be trained to recognize data and predict future outcomes based on the trends that exist in the data. This method of analysis has not been used on engine data, specifically in-line oil. Oil is an engine’s lifeblood and a lot of data can be collected and engine health can be predicted based on these measurements. This project aims to deploy machine learning concepts in the area of engine failure prediction. A special sensor equipped with the machine learning algorithm will be able to report all vital signs of an engine in a matter of minutes.

Development of a hybrid seismic data inversion method for determining well-drilling location at complex geophysical area

Due to the current economic downturn, especially the lower crude oil price, the drilling success rate become the most important goal for any oil/gas company. For a start-up company, any failure in drilling will be a disaster. To this end, the Deep Treasure Corp wishes that through the combination of mature hydrocarbon prediction techniques and new research results in seismic inversion, the success rate of hydrocarbon prediction, the theoretical basis for well placement can be provided in Roncott field, which will improve the success rate in drilling.

Numerical modeling of interaction between soil and seabed infrastructure during submarine landslides

Large amounts of crude oil and natural gas are located beneath Canada’s ocean floors. Currently, the offshore oil and gas industry makes a significant contribution to the Canadian economy. Although it is not visible as onshore landslides, numerous underwater landslides occur where huge amount of seabed sediment might displace and could result in tsunami in some cases. The failed soil mass might impact seabed infrastructure and could damage or destroy them as reported from some field observation.

REEs in brachiopods dwelling oxygen deficient habitats as proxies of paleoredox and potential source rocks

Source rocks are one of most important components of a petroleum system (a source rock, a reservoir rock and a trap) since it is economically irrelevant to exploit a hydrocarbon play without a source. The potentiality of rocks to retain hydrocarbons is defined by their organic contents. The environmental conditions prevailed during the deposition of sediments control the amount of the incorporated organic matter. In general, source rocks are precipitated in highly reducing or anoxic environments and enclose moderate to high organic contents.

Optical and Electrochemical Corrosion Detection and Protection

During the proposed internships, adaptive corrosion protection system (ACPS) will be developed as a stand-alone unit to provide optimum corrosion protection by changing the protection power according to the changes in environment or the material properties. This will allow the dynamic adjustments by implementing the feedback loop for the protected system. The proposed ACPS will also use efficiently stored energy from harvesting or charging. The proposed ACPS will significantly reduce and/or eliminate human interaction for an efficient and a cost-effective.

Investigation of Rock Penetration and Fragmentation Problems

Anaconda Mining Inc. is experiencing three rock penetration and fragmentation problems that impact costs, recovery and overall feasibility. Through collaboration between Anaconda Mining and the Drilling Technology Laboratory (DTL) at Memorial University of Newfoundland (MUN) and the MITACS graduate internships described in this proposal, these problems will be investigated and appropriate solutions developed. The investigation of these research questions is applicable in the mining as well as oil & gas fields.

Development of Novel Microparticle and Nanoparticle-Based Controlled Release Formulations for Agriculture

Increasing the productivity of agricultural fields is essential to secure our existing food supplies and provide for the growing world population. In this context, pesticides and fertilizers play an essential role in both increasing crop yields, as well as enhancing crop defenses against environmental stresses such as drought, pests, or diseases. However, the application of existing chemical pesticides and fertilizers can lead to eventual crop resistance as well as potential downstream environmental issues.

Machine learning in fluid composition quantification

A critical issue in the oil and gas industry is to quantify the composition of fluids flowing back from the hydraulic fracturing process. This quantification is usually carried out by a manual process (frequently via a visual test) to estimate the water and oil produced from a well flow back process. A sample of these onsite tests are sent to laboratories for chemical analysis. This process has been the status quo for decades. This approach is manual, prone to error, and does not lend itself to sophisticated real time analysis.

New statistical machine learning methods applied to high dimensional sensory input data from chemistry

Machine learning is the concept where a computer can be trained to recognize data and predict future outcomes based on the trends that exist in the data. This method of analysis has not been used on engine data, specifically in-line oil. Oil is an engine’s lifeblood and a lot of data can be collected and engine health can be predicted based on these measurements. This project aims to deploy machine learning concepts in the area of engine failure prediction.

Behavioral Analysis of H1 Reconstruction in New Software Environments

The precise prediction of fluid behavior is required in many fields of engineering. Fluid flows are governed by a complex system of continuous partial differential equations (PDEs) which rarely have an exact analytical solution. Computational Fluid Dynamics (CFD) has emerged as a leading method of analyzing fluid flows, by numerically solving the respective PDEs. Current methods in finite volume schemes of CFD on unstructured meshes have two major sources of errors: noise in the reconstructed gradients and lack of cancellation during flux integration.

Pages