An Efficient Control Formulation for Cyclic Process with Application to an Industrial System

Many industrial processes are cyclic, i.e., they exhibit periodic repetition of certain dynamics. In general, start or stop of individual cycles are subjected to some discrete events such as time event, control event, state events. These events are characterized by opening/closing of solenoid valves, turning on/off the heater, cycle time, etc. Examples of such cyclic processes are found in number of industrial applications such tyre manufacturing [1], upstream oil and gas [2], material processing,etc. In tyre manufacturing, the mixture consisting of latex, cross-linker, filler and other chemicals is set in a mold, which is then heated by a steam jacket to vulcanization temperature and then cooled to achieve desired properties. The discrete events include introduction of the mold in the processing chamber and turning the heating on/off [1]. Similarly, periodic shut-in in upstream oil and gas [3] can also be classified as cyclic processes where system dynamics repeats based on the binary decision of shutting or opening of the production valve.  In this contribution, an efficient control methodology will be developed, which is suitable to be implemented in low power controllers such as RTUs/PLCs.

First principles modeling of such process is computationally complex and at times it becomes intractable when used for model based control. On the other hand, available data based modeling, in its current form, is unable to generate control relevant model for such processes. Mainly, because in cyclic process, objective is to control the key performance indices (KPIs) instead of measured variables. These KPIs vary with cycle-to-cycle rather than in time-series fashion. Thus, use of time-series input output data is not required to build control relevant models for such processes.  For example, the aim of rubber manufacturing is to ensure tyre with uniform properties, whereas that of periodic well operation is to ensure maximum oil removal from the well. In either case, the key performance indices (KPIs) are not measured and the measured variables e.g. temperature of steam jacket in tyre manufacturing process cannot be used directly to build control relevant model.

In this work, we will develop an efficient data based modeling approach that refrains from using time-series input output data. In particular, time-series input-output data will be mapped into cycle-wise KPIs and all discrete events will be converted into continuous function evaluation in a systematic manner. These batch-wise data will then be used to develop a KPIs-control relevant model and its corresponding control law. Efficacy of the proposed modeling and control scheme will be demonstrated on an industrial relevant application where at least one of the inputs is binary, and at least one of the KPIs need to be calculated from measured data.

References:

[1]   A.P. Methal, “Modelling and simulation of degree of crosslinking in tyre manufacturing”, MS Thesis, Indian Institute of Technology - Madras (2010)

[2]   N.S. Kaisare, A. Gupta, V. Kariwala, N.N. Nandola, J.W. Green, G. Seikel and P.C. Somdecerff, “Control and Optimization Challenges in Liquid-Loaded Shale Gas Wells”, in Proceedings of Dynamics and Control of Process Systems – DYCOPS (2013), Mumbai, India.

[3]   B.R. Knudsen, “Production optimization in shale gas reservoirs”, MS Thesis, Norwegian University of Science and Technology (2010).