Process Systems and Control Engineering

Main Page

From psacelab
Revision as of 05:51, 11 December 2019 by Jinfeng (talk | contribs) (→‎Research)

Jinfeng Liu, PhD, PEng
Associate Professor

Department of Chemical & Materials Engineering
University of Alberta
13-269 Donadeo Innovation Center for Engineering
9211-116 Street, Edmonton, AB, CANADA T6G 1H9

Phone: +1-780-492-1317


We work in the general area of process systems and control engineering. Our primary research objective is to develop computing technologies for smart and sustainable manufacturing and production using systems, computing and engineering principles. We are actively working on the following projects:

Smart closed-loop agriculture irrigation

Water is essential for our daily life and is at the core of sustainable development. While about 70% of the Earth’s surface is covered by water, less than 2.5% of it is freshwater. The global water needs of the fast growing world population are already exceeding the currently available fresh water supply. The largest use of fresh water is in agriculture, which accounts for about 70% of the global fresh water consumption. As population growth continues, 60% more food will be needed to satisfy the demand of more than 9 billion people worldwide by 2050. However, in many regions (even the ones in water-rich countries such as USA and Canada), water allocated to irrigation is largely capped or will be capped in the near future and any further increase in irrigated acres needs to come essentially from improvements in water-use efficiency. The irrigation water-use efficiency worldwide is currently very low (around 50% to 60%). It is clear that policies related to water conservation and new technologies for more efficient water consumption need to be developed.

In the current practice, the amount of water to be irrigated and the time to apply the irrigation are determined in advance based on the irrigator’s (typically the farmer’s) knowledge. The actual conditions in the field are generally not considered in determining the irrigation amount and time. Thus from a process systems engineering (PSE) perspective, the current irrigation practice is an open-loop decision making process. It is well recognized in process control that open-loop control is quite imprecise. It is indeed one of the main causes of the low irrigation water-use efficiency.

To meet the challenges of water sustainability, the key is to close the decision-support loop to form a closed-loop system for precision irrigation. In the closed-loop system, sensing instruments (e.g., soil moisture and nutrient sensors, evapotranspiration (ET) gauge, thermal cameras on drones or satellite imagery data) can be used to collect various real-time field information regularly. The various field information can then be fused together to get estimates of the entire field’s conditions. The estimated field conditions could then be fed back to an adaptive or self-learning control system. The self-learning control system would then calculate the best irrigation commands for the next few hours or day(s) based on a field model, the estimated field conditions, local weather forecast as well as other pre-specified irrigation requirements. Due to significant nonlinearities, uncertainties, and large scale of agricultural fields, there are great challenges in modeling, sensing, estimation and control algorithm development that need to be addressed. We have been actively developing agro- hydrological models, remote sensing and sensor fusing algorithms, optimal control algorithms to realize this revolutionary closed-loop smart irrigation vision in collaboration with different partners (including sensing instruments provider, sprinkler manufacturer, farmers, and government agency).

Advanced process control framework for smart manufacturing

The process industry is in the transition from the traditional process operations paradigm to the smart manufacturing paradigm. Smart manufacturing is the dramatically intensified and pervasive applica- tion of manufacturing intelligence throughout the manufacturing and supply chain enterprise. In smart manufacturing, different function modules are tightly integrated together through real-time commu- nication and respond as a coordinated and performance-oriented enterprise, accounting for economic performance, environmental sustainability, health and safety. This transition to smart manufacturing poses significant challenges for process control and operations. We have been working on advanced process control techniques that enable the transition to smart manufacturing.

Cooperative predictive control framework. Model predictive control (MPC) is the most widely used advanced process control technique and is in the category of core manufacturing intelligence. Traditionally, independently designed MPCs are used to control the important operating units in a plant in a decentralized fashion and there is no or limited coordination between these MPCs. In our work, we have developed cooperative distributed MPC framework in which individual MPCs commu- nicate with each other and coordinate their actions to achieve significantly improved plant-wide control performance. We have also developed economic MPC algorithms that can handle very general control objective (e.g., maximize economic performance) directly. Process nonlinearity, stability, plant-wide performance as well as potential communication issues are all carefully taken into account in our de- signs. Our developed cooperative distributed MPC and economic MPC framework is well fitted to smart manufacturing and can be used for coordinated process control or act as a distributed advanced decision support system.

Distributed estimation and process monitoring. Process monitoring has played a critical role in the traditional process operations paradigm. It will continue to be one critical component in smart manufacturing. In our work, we have developed distributed state estimation algorithms that are suitable for process monitoring in smart manufacturing. In our developed algorithms, there are more than one estimators that can communicate to exchange information and work collaboratively to estimate the state of the entire process network. We have developed a very general distributed state estimation framework for process monitoring purpose. The developed distributed estimation framework is unique in that it does not require the local subsystem estimators to be of the same type. It may be used to transform existing decentralised estimators to a distributed state estimator network, which could be very useful as we transforming existing facilities to smart plants.

Process networks decomposition. As we move towards smart manufacturing, complex and tightly integrated process networks are becoming more and more common in manufacturing industries due to their economic efficiency. A typical process network consists of several operating units, which are connected with each other via material, energy and information flows. When considering cooperative distributed predictive control of these process networks, how to decompose the process network for distributed control system design is very important and is a fundamental problem. Improper subsystem decomposition may lead to the increase of communication and computation burdens, reduced control performance, or even instability of the entire system. We have investigated this fundamental problem from different perspectives including time-scale multiplicity, physical closeness and sensitivity between different variables, and community structure detection based on representing a system as a graph.

Applications to energy systems, wastewater treatment, and anemia management

We have applied our developed advanced process control techniques to optimize the operations of various energy systems including wind-solar systems, deep cone thickener and froth flotation process in coal beneficiation, primary separation vessel in oilsands separation, steam assisted gravity drainage oil recovery processes, post-combustion CO2 capture plants, and coal-fired boiler-turbine systems. We have also developed model reduction, decomposition, and optimal control algorithms for wastewater treatment plants. Features of these systems are explicitly taken into account in our algorithms.

We have also applied advanced process control techniques to develop modeling and control algorithms for anemia management in chronic kidney disease. Chronic kidney disease affects millions of people throughout the world today. In this work, we have developed an economic zone MPC algorithm to optimally recommend erythropoietin (EPO) dosing amounts to the physician to maintain a patient’ hemoglobin within a target range. We have also developed a high fidelity patient simulator based on pharmacokinetics and pharmacodynamics model and identified patient parameters for control perfor- mance evaluation. The developed algorithms are currently implemented in the commercial product of our industrial partner.