1,720,996 research outputs found
ADAPTIVE TUNING OF MODEL PREDICTIVE CONTROL PARAMETERS VIA REINFORCEMENT LEARNING
No full textThis thesis presents a reinforcement learning (RL)-assisted model predictive control (MPC) framework for multivariable chemical processes that experience external time varying disturbances. MPC is widely used in industry for its ability to predict future behaviour and enforce operating constraints. However, its performance depends on tuning parameters, in this case, the prediction and control horizons. They are usually selected offline and kept fixed during operation. Even with offset-free formulations, the chosen horizons remain fixed, which can result in degraded plant performance. The thesis presents a control framework in which a Deep Q-Network agent dynamically updates the MPC prediction and control horizons in real time. The resulting improvement in set-point tracking is demonstrated by comparing it with the industrial MPC with fixed horizons. A variation of the formulation is also introduced with the reward function explicitly penalizing changes in the agents actions, resulting in more stable agent behavior under the closed-loop MPC.ThesisMaster of Applied Science (MASc
NOISE AWARE BAYESIAN PARAMETER ESTIMATION IN BIOPROCESSES
No full textThis thesis demonstrates a parameter estimation technique for bioprocesses that utilizes
measurement noise in experimental data to determine credible intervals on parameter
estimates, with this information of potential use in prediction, robust control,
and optimization. To determine these estimates, the work implements Bayesian inference
using nested sampling, presenting an approach to develop neural network (NN)
based surrogate models. To address challenges associated with non-uniform sampling
of experimental measurements, an NN structure is proposed. The resultant surrogate
model is utilized within a Nested Sampling Algorithm that samples possible parameter
values from the parameter space and uses the NN to calculate model output
for use in the likelihood function based on the joint probability distribution of the
noise of output variables. This method is illustrated against simulated data, then
with experimental data from a Sartorius fed-batch bioprocess. Results demonstrate
the feasibility of the proposed technique to enable rapid parameter estimation for
bioprocesses.ThesisMaster of Applied Science (MASc)Bioprocesses require models that can be developed quickly for rapid production of desired
pharmaceuticals. Parameter estimation is necessary for these models, especially
first principles models. Generating parameter estimates with confidence intervals is
important for model based control. Challenges with parameter estimation that must
be addressed are the presence of non-uniform sampling and measurement noise in
experimental data. This thesis demonstrates a method of parameter estimation that
generates parameter estimates with credible intervals by incorporating measurement
noise in experimental data, while also employing a dynamic neural network surrogate
model that can process non-uniformly sampled data. The proposed technique
implements Bayesian inference using nested sampling and was tested against both
simulated and real experimental fed-batch data
Data-Driven Modelling and Control of Batch Processes
No full textIt is important to note that the techniques proposed in this thesis are applicable to any batch process with a similar input-output-quality variable structure. The thesis demonstrates the effectiveness of these approaches through experiments on example batch processes. The primary focus of this thesis is to facilitate the modelling of batch processes considering quality variables in the context of model-based quality control and provide direct quality control formulations as opposed to existing techniques which are primarily focused on trajectory control of measured variables. These quality variables, which are unique to batch processes, cannot be measured during a batch run and can only be assessed at the end of the batch.
The ability to implement good quality control is dependent on the ability to capture the complex dynamic behavior of batch processes. The first challenge is that of handling process non-linearities and/or multiple phases while being cognizant of the fact that a relatively modest amount of informative data is available, making the recently developed deep-learning-based machine-learning techniques not directly implementable. Yet another challenge/opportunity that exists is in situations where traditional sensors such as thermocouples may not be possible to implement in practice, but feedback may be available through non-traditional sensors such as thermal images.
The present thesis addresses the above-mentioned challenges and demonstrates the approaches on a pilot-scale experimental setup. In the first contribution, a data-driven model-based economic control formulation was initially developed and implemented on a batch Rotational Molding process to achieve product specifications through constraints on the predicted quality variables, while either minimizing the total input consumption or further maximizing the product quality. A Linear Time-Invariant (LTI) State Space (SS) model is used in conjunction with a Partial Least Squares (PLS) quality model to model the process and quality variables. The next contribution presents one way of handling process nonlinearity. An adaptive modelling strategy unique to batch processes was proposed to alleviate this issue and implemented on the Rotational molding process to continuously adapt the LTI SS model during a batch run. The next contribution leveraged the nonlinearity-capturing capabilities of modelling techniques like Neural Networks (NN) while handling the overfitting problem. The key idea in this approach was to use a subspace identification approach to first determine the state trajectory evolution followed by a Recurrent Neural Network (RNN) to model the non-linear process dynamics. This approach, along with the previous PLS quality model, also exhibited superior dynamic and quality predictions compared to a standard RNN case. In a departure from existing approaches, where the dynamic model and quality model are identified separately, an adaptation of the prediction error minimization framework was proposed where the dynamic model and the quality model are identified simultaneously, resulting in an improved unified model. Finally, the opportunity/challenge of non-traditional sensors was addressed. A framework was proposed where first, the high dimensional output, a thermal image in the specific example batch process, is reduced using a suitable dimensionality reduction technique to a set of latent features, which then would be used as the outputs of the dynamic model in any of the previously discussed approaches. The proposed modelling framework was implemented as a part of a Model Predictive Controller implementation using the thermal images as feedback to produce the desired product.DissertationDoctor of Philosophy (PhD
The Right Tools for the Job: Design Choices of Parallel First Principle and Data-Driven Hybrid Modelling for Prediction and Control of Batch and Fed-Batch Reactors
Full text linkThird submission of my master thesis, first one didn't finish for some reason, the second I think I forgot to include my name.This thesis focuses on the creation of new parallel hybrid model designs for prediction and control in batch and fed-batch reactors within Model Predictive Control (MPC) frameworks. In the hybrid model, the first principle (FP) explains the dynamics and the residual Subspace Identification (SID) model explains the error between the FP and the process. Modifications to the structure of the hybrid model are motivated by limitations of MPC frameworks. MPCs need accurate models to explain the system dynamics to make informed control decisions, and mechanistic models can be difficult to implement due to challenges of solving the optimization problem in real time. Two tools are demonstrated to help solve these problems. The first tool, Residual First Principle 0 Hybrid (RFP0H) model, helps to deal with the intractability of a mechanistic model in a hybrid modelling framework. The input for the FP model is kept constant and the SID predicts the error between the first principle and the process. Allowing for the desired output to be subtracted by the predicted FP to create a desired error value. Thus, MPC control only needs to be solved using the linear SID model in a linear or quadratic framework. Making a potentially intractable problem, tractable in MPC. This is demonstrated using a simulated fed-batch crystallization process. The second tool, Scaling Factor First Principle 0 Hybrid (SFFP0H) model, modifies the hybrid model structure to multiple the sub-models’ outputs together. The SID data driven model predicts a factor to scale the FP output for the process prediction. The results demonstrate that the SFFP0H model has increased predictive ability and has smaller variability in control compared to the RFP0H model. Helping to solve the problem of needing accurate models within an MPC formulation. This is demonstrated by using a laboratory scale batch polymerization process.ThesisMaster of Applied Science (MASc
ADVANCES IN MODEL PREDICTIVE CONTROL
Full text linkIn this thesis I propose methods and strategies for the design of advanced model predictive control designs. The contributions are in the areas of data-driven model based MPC, model monitoring and explicit incorporation of closed-loop response considerations in the MPC, while handling issues such as plant-model mismatch, constraints and uncertainty.
In the initial phase of this research, I address the problem of handling plant-model mismatch by designing a subspace identification based MPC framework that includes model monitoring and closed-loop identification components.
In contrast to performance monitoring based approaches, the validity of the underlying model is monitored by proposing two indexes that compare model predictions with measured past output. In the event that the model monitoring threshold is breached, a new model is identified using an adapted closed-loop subspace identification method. To retain the knowledge of the nominal system dynamics, the proposed approach uses the past training data and current input, output and set-point as the training data for re-identification.
A model validity mechanism then checks if the new model predictions are better than the existing model, and if they are, then the new model is utilized within the MPC.
Next, the proposed MPC with re-identification method is extended to batch processes. To this end, I first utilize a subspace-based model identification approach for batch processes to be used in model predictive control. A model performance index is developed for batch process, then in the case of poor prediction, re-identification is triggered to identify a new model. In order to emphasize on the recent batch data, the identification is developed in order to increase the contribution of the current data.
In another direction, the stability of data driven predictive control is addressed. To this end, first, a data-driven Lyapunov-based MPC is designed, and shown to be capable of stabilizing a system at an unstable equilibrium point. The data driven Lyapunov-based MPC utilizes a linear time invariant (LTI) model cognizant of the fact that the training data, owing to the unstable nature of the equilibrium point, has to be obtained from closed-loop operation or experiments. Simulation results are first presented demonstrating closed-loop stability under the proposed data-driven Lyapunov-based MPC. The underlying data-driven model is then utilized as the basis to design an economic MPC.
Finally, I address the problem of control of nonlinear systems to deliver a prescribed closed-loop behavior. In particular, the framework allows for the practitioner to first specify the nature and specifics of the desired closed-loop behavior (e.g., first order with smallest time constant, second order with no more than a certain percentage overshoot, etc.). An optimization based formulation then computes the control action to deliver the best attainable closed loop behavior. To decouple the problems of determining the best attainable behavior and tracking it as closely as possible, the optimization problem is posed and solved in two tiers. In the first tier, the focus is on determining the best closed-loop behavior attainable, subject to stability and tracking constraints. In the second tier, the inputs are tweaked to possibly improve the tracking of the optimal output trajectories given by the first tier. The effectiveness of all of the proposed methods are illustrated through simulations on nonlinear systems.DissertationDoctor of Philosophy (PhD
Energy Efficient Model Predictive Building Temperature Control
Full text linkSustainability considerations have placed increasing emphasis on the energy efficient operation and control of temperature control systems. It is estimated that the use of advanced control structures could lead to valuable savings in energy expenditure (up to 15-20 %) . This work considers the problem of developing a model predictive control (MPC) algorithm for temperature control in buildings. To this end, a cascade control structure was designed to regulate the room temperature subject to heat load disturbances, such as outdoor conditions or changes in the internal gains (i .e., number of people in a room). The inner loop of the cascade control structure involved controlling key variables of a vapor compression cycle (VCC), namely the superheat and supply air temperature (from the evaporator), by manipulating the compressor speed and valve opening (components in the VCC). Linear inputoutput models were appropriately identified for the VCC using a detailed first-principles model (adapted from Thermosys) for eventual utilization in a predictive control design. Then, closed loop simulations were performed by interfacing the VCC model with EnergyPlus (developed by the U.S . Department of Energy) , which was used to model realistic room temperature behavior. The control performance using a predictive controller (in the inner loop) was then evaluated against PI control.Master of Applied Science (MASc
A safe-parking framework to handle faults in nonlinear process systems
Full text link This thesis considers the problem of control of nonlinear process systems subject to
input constraints and faults in the control actuators and process equipments. Faults
are considered that preclude the possibility of continued operating at the nominal
equilibrium point and a framework (which we call the safe-parking framework) is
developed to enable efficient resumption of nominal operation upon fault-recovery.
First, Lyapunov-based model predictive controllers, that allow for an explicit characterization
of the stability region subject to constraints on the manipulated input,
are designed. The stability region characterization is utilized in selecting 'safe-park'
points from the safe-park candidates (equilibrium points subject to failed actuators).
This safe-park point is chosen as a temporary operating point where process is to
be operated during fault rectification. This ensures that process can be safely operated
during fault rectification and the nominal operation can be resumed upon fault
recovery. When multiple candidate safe-park points are available, performance considerations,
such as ease of transition from and to the safe-park point and cost of
running the process at the safe-park point, are quantified and utilized in choosing the
optimal safe-park point. Next, we extend the safe-parking framework to handle practical issues such as plant-model mismatch, disturbances and unavailability of all process state measurements.
\i\Te first consider the presence of constraints and uncertainty and develop
a robust Lyapunov-based model predictive controller. This controller is utilized to
characterize robust stability region which, subsequently, is utilized to select 'safepark'
points. Then we consider the problem of availability of limited measurements.
An output feedback Lyapunov-based model predictive controller, with high-gain observer
to estimate unmeasured states, is formulated and its stability region explicitly
characterized. An algorithm is then presented that accounts for the estimation errors
in the implementation of the safe-parking framework. We then further extend the framework to handle faults in large scale chemical
plants where multiple process units are connected via material, energy and information
streams. In plant-wide setting, the safe-park point for the faulty unit is chosen
such that the safe-parking has no or minimum effect on downstream units, and hence,
the nominal operation in the downstream units can be continued. Next we consider
the scenario where no viable safe-park point for the faulty unit exists such that its
effect can be completely absorbed in the subsequent unit. A methodology is developed
that allows simultaneous safe-parking of the consecutive units. The efficacy of
the proposed framework is illustrated using a chemical reactor example, a styrene
polymerization process and two CSTRs in series example. Finally, we demonstrate the efficacy of proposed Lyapunov based Model Predictive
Controller and Safe-Parking framework on a polymerization reactor model to control
the polymerization reactor and to handle faults that dont allow continuation of the
nominal operation in the reactor. ThesisDoctor of Philosophy (PhD
NONLINEAR MODEL PREDICTIVE CONTROL DESIGN AND APPLICATIONS
Full text linkThis thesis considers the problem of nonlinear predictive control design and applications. A predictive control formulation is presented which expands on the set of initial conditions for which closed-loop stability can be achieved. The key idea in this control design is to utilize control-law independent characterization of the process dynamics subject to constraints via model predicative controllers. An application of this idea is presented to the case of linear process systems for which characterizations of the null controllable region (the set of initial conditions from where closed-loop stability can be achieved subject to input constraints) are available. A predictive controller is designed that achieves closed-loop stability for every initial condition in the null controllable region. For nonlinear process systems, the constraints within the predictive controller are formulated to require the process to evolve within the region from where continued decay of the Lyapunov function value is achievable and incorporated in the predictive control design, thereby expanding on the set of initial conditions from where closed-loop stability can be achieved. The proposed method is illustrated using a chemical reactor example, and the robustness with respect to parametric uncertainty and disturbances demonstrated via application to a styrene polymerization process. In addition, we also consider the application of the predictive control design to the problem of handling actuator faults in nonlinear continuous-time processes and transport-reaction systems. Specifically, we consider faults that preclude the possibility of continued operating at the nominal equilibrium point using the existing robust or reconfiguration-based fault-tolerant control approaches. The key consideration is to operate the plant using the depleted control action at an appropriate safe-park point to prevent onset of hazardous situations as well as enable smooth resumption of nominal operation upon fault-repair. For the case of continuous-time nonlinear process systems we consider the presence of input constraints, uncertainty, and availability of limited measurements. First a Lyapunov-based predictive controller with an explicitly characterized stability region is developed to handle the aforementioned conditions. This control design is then subsequently used to develop a safe-parking framework in the presence of uncertainty, and availability of limited measurements. The proposed framework is illustrated using a chemical reactor example and demonstrated on a styrene polymerization process. Finally, we consider the problem of model predictive control and handling actuator faults in transport-reaction processes described by quasi-linear parabolic partial differential equations (PDEs) subject to input constraints. A Lyapunov-based model predictive controller is designed which accounts for the distributed nature of transport-reaction processes and provides an explicit characterization of the set of initial conditions from where closed-loop stability of the parabolic PDE system is guaranteed. Similar to the continuous time case, this control design is then subsequently used to develop a safe-parking framework for handling actuator faults in transport-reaction processes. The proposed framework is illustrated on a diffusion-reaction processMaster of Applied Science (MASc
THE DESIGN OF A NOVEL LYAPUNOV-BASED OFFSET-FREE MODEL PREDICTIVE CONTROLLER
Full text linkThis thesis considers the problem of control of nonlinear systems subject to limited availability
of measurements and uncertainty in model parameters. To address this problem, first a
linear offset free MPC is designed. Subsequently, a Lyapunov-based offset free MPC design
is presented to handle structured uncertainty subject to constant disturbances. The controller's ability to handle unstructured uncertainty and measurement noise is demonstrated through simulation examples. Next, the problem of handling lack of state measurements as well as uncertainty is considered. To achieve simultaneous state and disturbance parameter estimation, a Lyapunov-based model predictive controller (MPC) is integrated with a moving horizon based mechanism, to achieve (where possible) offset elimination in the unmeasured states as well. A chemical reaction process example is presented to illustrate the key points. Finally its efficacy is demonstrated through a polymerization process example.ThesisDoctor of Philosophy (PhD
ON THE STABILIZATION OF NONLINEAR CONTROL SYSTEMS SUBJECT TO STOCHASTIC DISTURBANCES AND INPUT CONSTRAINTS
Full text linkThis thesis investigates the broad theme of guaranteeing the stability of nonlinear control systems. In the first section, we describe the application of discrete controller for the stabilization of certain nonlinear stochastic control systems subject to unavailability of state measurements. In the second section, we consider input constrained nonlinear systems and characterize the region from which stabilization to the origin is possible. We then use this information to design a controller which stabilizes everywhere in this set.ThesisMaster of Applied Science (MASc
- …
