1 Problem Statement
Total Submissions : 3

Many times the plant control strategy is complex and not possible to be implemented into commercial SCADA  hardware such as PLCs due unavailability of a defined control model. The requirement is then for a machine learning platform which will use the data parameters from plant operator’s control room alarm logger, recorder, operations database and generate a pattern to distinguish the Normal and OFF normal conditions using multi variable analysis.  One such requirement is for automation of plant operations which presently heavily depend on the design fixed parameters and or operator’s experience of operation under the safe conditions. The machine learning platform may accept multivariable data in suitable form such as 1D,2D,3D…matrices per parameter and generate abstract relations from them to indicate OFF normal and normal conditions of plant, which are defined by operators on the go. The generic machine learning platform can be put to use in any process plant to read from a set of variables and generate abstracted unique patterns for operators to distinguish visually the normal conditions from off normal conditions quickly.
One such example of machine learning is auto pilot and driverless car. If the machine learning platform is applied to the variables of these control applications then it will generate abstract of the conditions for drivers to distinguish the normal and off normal conditions of fleet. The final decision of pattern being off normal or normal is dependent on the control room operator in the case of this problem statement.
It is applied to the process applications where the mathematical models for control are not easily available and or the process is highly depended on fixed operating parameters such as IMPUREX reprocessing control system. With lab scale model of such process one can generate enough data to feed the required machine learning platform and then abstract the control strategy for automation of process control.

Sample data required: Yes; example_dataset_hackathon2017 , hackathon2017