Robust fault tolerant model predictive control (RFTMPC) for high order systems

Lead Research Organisation: Imperial College London
Department Name: Electrical and Electronic Engineering


EPSRC area: Control Engineering
Industrial partner: Schlumberger Cambridge Research Limited
Case voucher: 17100038

Process control and optimization are becoming more important as automation and robotics are incorporated into complex safety-critical processes; for example, the construction of oil and gas wells.

Model predictive control (MPC) is an online control strategy for problems with hard constrains. It is most suitable for lower order known models, since the resulting optimization is a quadratic program.

Complex models need to be replaced by approximations. Approximation, parametric uncertainties of the original model and disturbances, requires Robust MPC. In safety critical applications, a fault tolerance measure is necessary.

The project aims to integrate model identification and approximation, incorporate uncertainties and fault scenarios into a Robust Fault Tolerant MPC problem. A further aim is to develop feasible relaxation based solution algorithms to the resulting non convex optimization problems, and apply in industrial problems of interest to Schlumberger (e.g.: curvature control).

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/R512540/1 01/10/2017 31/03/2022
2103076 Studentship EP/R512540/1 15/01/2018 31/01/2022 Anastasis Georgiou Georgiou
Description Rotary steerable system RSS is a drilling technique which has been used in hydrocarbon exploration and expected to drill complex curved borehole trajectories accurately and smoothly in the complex stratum or in deep-ocean. The main difficulty of developing control algorithms for the RSS system is the lack of knowledge on the dynamic system and the presence of disturbances in such an unpredictable working environment. An initial contribution of this project has been to propose a dynamic model of the directional drilling system in terms of ordinary differential equations by a closed-form state-space representation, unlike conventional representations in the literature that utilise either less accurate kinematics system models, or with comprehensive dynamic models that are presented in terms of delay differential equations. Very importantly, the suggested model has been validated successfully against a high-fidelity industry grade finite element model developed by our industrial collaborators, Schlumberger. Ongoing work mainly investigates the Model Predictive Control (MPC) scheme, which has found great success and application in the oil and gas industry. New theoretical contributions are sought to take into account uncertainties of the system which arise by design approximations and disturbances to extend the formulation of the MPC algorithm to an applicable form of Robust MPC (RMPC) scheme, so that it can be used for effective RSS control. The output of the proposing control scheme algorithm is applied in industrial directional drilling model, in order to compare the proposing automated method with existing sub-optimal manual methods. 
Exploitation Route The outcomes of this project can directly benefit the oil and gas industry. More specifically the existing control actuator commands on the directional drilling system are operated with major communication delays by professionals, who are located at the surface close to the drilling rig, using complex data sets, such as location of the reservoir, rock layer geometry, etc. Human errors and communication delays could be minimised by automating the steering commands by the suggested closed-loop control scheme that uses real-time data from sensors located in the drill string. Furthermore, with some customisation the investigated control scheme in this project can be potentially used in various sectors such as aerospace, automotive, power systems, finance and marketing, and chemical process.
Sectors Education,Electronics,Energy,Other

Description Tracking control for directional drilling system 
Organisation Schlumberger Limited
Department Schlumberger Cambridge Research
Country United Kingdom 
Sector Academic/University 
PI Contribution Theoretical contribution related to the following engineering problems: 1) Robust Model Predictive control 2) State Estimation 3) Model reduction 4) Invariant Sets 5) Trajectory tracking control for discrete-time uncertain systems affected by disturbances.
Collaborator Contribution The main contribution from our collaborators was with respect to the description of the dynamic model of directional drilling system, by providing us with industrial knowledge, data and parameter values.
Impact An output from our collaboration is a recently accepted paper to the 21st IFAC World Congress in Berlin. A. Georgiou, S.A. Evangelou, I.M. Jaimoukha and G. Downton. "Tracking Control for Directional Drilling Systems Using Robust Feedback Model Predictive Control." 21st IFAC World Congress, 2020.
Start Year 2018
Description Schlumberger Cambridge Research 1st Annual PhD Workshop 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact The workshop was organised by Schlumberger in order to give the opportunity to fellow Phd students to meet each other and find out more about industrial problems that Schlumberger aims to tackle. Oral presentation, poster session and discussion related to various research works were held by the students and the company in an informal environment to engage students with the technical community.
Year(s) Of Engagement Activity 2019