Case Studies & Resources

Machine Learning can help in process monitoring in various ways. In a typical process plant there are many process variable are to be monitored. Identifying the onset of deviation by monitoring each of these variables and parameters can very overwhelming and even infeasible at times. However all the variables are not equally responsible for system's behavior, so it can be really helpful if the total number of variables to be monitored to a few important ones. There are ML tec

In AI/ML domain, model development approaches can be segregated as "supervised" or "unsupervised". In supervised learning approach the...

In process and manufacturing industries it is not always feasible to measure every variables of interest. In those scenarios soft sensors can be really helpful. So, Soft sensors, also known as virtual sensors, are models used to estimate the values of unknown process variables using the available measurements. Soft sensors are also used when physical sensors are very costly or real-time measurements are not possible. Soft sensor models can be based on first principles utilizi

There are techniques in AI/ML to solve all of the above challenges. For example Recurrent Neural Network (RNN) can be used to process the...

In process & manufacturing applications identifying the onset of system abnormality / failure at the earliest is very crucial. Models which help in determining the occurrences of such abnormalities as "fault detection" models. The techniques to isolate the variables which are causing the abnormality is generally known as "fault identification". Along with these there are scenarios where the type of fault needs to be determined, this generally known as "Fault Classification or

Unplanned down times and shut-downs, one of the major challenges industries face when disruptions happen due to an unanticipated break-down event. So plants always rely on preventive maintenance to avoid such occurrences. However , in order to optimize the productivity it is always desirable to have minimal number maintenance schedules. Data driven "predictive analytics" is one of the better options to optimize between maintenance stoppages and risk of unplanned disruptions.