Continuous casting of steel involving different grades in the same casting sequence remains a challenge to billet caster operators. The intermixed composition obtained during the grade change does not meet the specifi...Continuous casting of steel involving different grades in the same casting sequence remains a challenge to billet caster operators. The intermixed composition obtained during the grade change does not meet the specification of either grade and must be downgraded. Incorrect identification of this intermixed region may result in non-conforming products reaching the customer. In this study, a numerical model based on CFD (computational fluid dynamics approach) has been developed which predicts the start and end of the intermixed composition and the tonnage to be downgraded under different casting conditions. This model was validated and the results were in good agreement with the actual plant data for a 6-strand billet caster at LD-1 of TATA Steel, India. This model is used to calculate transition tonnage for different scenarios, e.g. when one of the outermost strands is not functional or some combinations are not functional and varying casting speed during operation. Furthermore, impact of different design of baffles on performance of Tundish has been evaluated to find a way to reduce transition or intermixed composition.展开更多
In Computer-Aided Detection(CAD)brain disease classification is a vital issue.Alzheimer’s Disease(AD)and brain tumors are the primary reasons of death.The studies of these diseases are carried out by Magnetic Resonan...In Computer-Aided Detection(CAD)brain disease classification is a vital issue.Alzheimer’s Disease(AD)and brain tumors are the primary reasons of death.The studies of these diseases are carried out by Magnetic Resonance Imaging(MRI),Positron Emission Tomography(PET),and Computed Tomography(CT)scans which require expertise to understand the modality.The disease is the most prevalent in the elderly and can be fatal in its later stages.The result can be determined by calculating the mini-mental state exam score,following which the MRI scan of the brain is successful.Apart from that,various classification algorithms,such as machine learning and deep learning,are useful for diagnosing MRI scans.However,they do have some limitations in terms of accuracy.This paper proposes some insightful pre-processing methods that significantly improve the classification performance of these MRI images.Additionally,it reduced the time it took to train the model of various pre-existing learning algorithms.A dataset was obtained from Alzheimer’s Disease Neurological Initiative(ADNI)and converted from a 4D format to a 2D format.Selective clipping,grayscale image conversion,and histogram equalization techniques were used to pre-process the images.After pre-processing,we proposed three learning algorithms for AD classification,that is random forest,XGBoost,and Convolution Neural Networks(CNN).Results are computed on dataset and show that it outperformed with exiting work in terms of accuracy is 97.57%and sensitivity is 97.60%.展开更多
文摘Continuous casting of steel involving different grades in the same casting sequence remains a challenge to billet caster operators. The intermixed composition obtained during the grade change does not meet the specification of either grade and must be downgraded. Incorrect identification of this intermixed region may result in non-conforming products reaching the customer. In this study, a numerical model based on CFD (computational fluid dynamics approach) has been developed which predicts the start and end of the intermixed composition and the tonnage to be downgraded under different casting conditions. This model was validated and the results were in good agreement with the actual plant data for a 6-strand billet caster at LD-1 of TATA Steel, India. This model is used to calculate transition tonnage for different scenarios, e.g. when one of the outermost strands is not functional or some combinations are not functional and varying casting speed during operation. Furthermore, impact of different design of baffles on performance of Tundish has been evaluated to find a way to reduce transition or intermixed composition.
文摘In Computer-Aided Detection(CAD)brain disease classification is a vital issue.Alzheimer’s Disease(AD)and brain tumors are the primary reasons of death.The studies of these diseases are carried out by Magnetic Resonance Imaging(MRI),Positron Emission Tomography(PET),and Computed Tomography(CT)scans which require expertise to understand the modality.The disease is the most prevalent in the elderly and can be fatal in its later stages.The result can be determined by calculating the mini-mental state exam score,following which the MRI scan of the brain is successful.Apart from that,various classification algorithms,such as machine learning and deep learning,are useful for diagnosing MRI scans.However,they do have some limitations in terms of accuracy.This paper proposes some insightful pre-processing methods that significantly improve the classification performance of these MRI images.Additionally,it reduced the time it took to train the model of various pre-existing learning algorithms.A dataset was obtained from Alzheimer’s Disease Neurological Initiative(ADNI)and converted from a 4D format to a 2D format.Selective clipping,grayscale image conversion,and histogram equalization techniques were used to pre-process the images.After pre-processing,we proposed three learning algorithms for AD classification,that is random forest,XGBoost,and Convolution Neural Networks(CNN).Results are computed on dataset and show that it outperformed with exiting work in terms of accuracy is 97.57%and sensitivity is 97.60%.
