Ari Ingimundarson

Robust tuning procedures of dead-time conpensating controllers

Abstract This paper describes tuning procedures for dead-time compensating controller (DTC). Both stable and integrating processes are considered. Simple experiments are performed to obtain process models as well as bounds on the allowable bandwidth for stability. The DTCs used have few parameters with clear physical interpretation so that manual tuning is possible. Furthermore, it is shown how the DTCs can be made robust towards dead-time variations.

Performance comparison between PID and dead-time compensating controllers

This paper is intended to answer the question: “When is a simple dead-time compensator expected to perform better than a PID?”. The performance criterion used is the integrated absolute error (IAE). It is compared for PI and PID controllers and a simple dead-time compensator (DTC) when a step load disturbance is applied at the plant input. Both stable and integrating processes are considered. For a fair comparison the controllers should provide equal robustness in some sense. Here, as a measure of robustness, the maximum of the absolute value of the sensitivity function is used. Since this is not a sufficient measure of robustness when errors in the dead time are considered, performance of the DTCs is given also as a function of dead-time margin (DM) where appropriate. (Less)

Objective Prioritization Using Lexicographic Minimizers for MPC of Sewer Networks

In this brief, objective prioritization of multiobjective cost functions using the lexicographic approach is applied in the model predictive control (MPC) framework of sewer networks. Using the lexicographic approach, the control problem solution can be obtained by solving a sequence of single objective, constrained, convex programming problems. This brief demonstrates with an elaborated case study treating a portion of the Barcelona sewer network, that important improvements can be achieved in performance using lexicographic optimization. At the same time, costly commissioning and implementation efforts related to the traditional weight based approach for implementation of priorities are avoided.

Model-Based Detection of Hydrogen Leaks in a Fuel Cell Stack

Hydrogen leaks are potentially dangerous faults in fuel cell systems that are fed with hydrogen-rich gas mixtures. This brief presents an approach to hydrogen leak detection and, thus, complements direct detection using hydrogen sensors. It relies on simple mass balance equations of an anode filling volume after taking into account the natural leak of the stack. A hydrogen mass flow, anode pressure, and relative humidity sensor are employed. Hydrogen leak detection without the use of relative humidity sensors is considered by employing adaptive alarm thresholds to eliminate false alarms. The validity of the method is also discussed in terms of common hydrogen supply system configurations. The detection method is validated using a 1.25-kW polymer electrolyte membrane fuel cell stack in a laboratory facility where leaks could be introduced in a controlled manner.

Using the Unfalsified Control Concept to achieve Fault Tolerance

Abstract The paper investigates the use of the unfalsified control concept in the area of fault tolerant control. No fault diagnosis system is required but rather by a simultaneous on-line performance assessment of multiple controllers in a bank of controllers, the best one for the plant at each time can be selected. A controller does not need to form part of the feedback loop for its performance to be assessed. Strategies to construct the bank of controllers are discussed and a switching strategy for fault tolerant control is presented. No previous models of system or faults are necessary, only real-time input/output data streams. Finally the investigated methodology is put to the test by applying it to a non-linear model of the breathing system of a PEM fuel cell.

Robust Fault Diagnosis using Parallelotope-based Set-membership Consistency Tests

The paper proposes robust fault detection methods that take advantage of a recently proposed set-membership identification procedure based on parallelotopes for systems linear in the parameters. It is shown that consistency checks indicating faults can be performed in a natural manner with a parallelotope description of the feasible parameter set. Fault detection algorithms are presented for systems with invariant parameters, with parameter variation bounded between samples and with unbounded variation. Finally, an example is given which demonstrates how the algorithms work on a multivariable process.

Robust automatic tuning of an industrial PI controller for dead-time systems

This paper describes an automatic tuning procedure for dead-time compensating controllers. A first order plus dead time model is identified from two step response experiments. The identification is based on the methods of moments. The model uncertainty is estimated automatically and used in the design to obtain a robust controller.

Fault Tolerant Model Predictive Control applied on the Barcelona Sewer Network

Fault-tolerant model predictive control strategies of sewer networks are investigated in this article. A general linear model of sewer networks is presented and the corresponding expression into MPC formalism are discussed. From FTC and MPC, two strategies are proposed and compared: Passive Fault Tolerant MPC, that takes advantage of natural tolerance (robustness) of MPC and Active Fault Tolerant MPC, that uses active fault tolerance techniques in combination with MPC. Finally, some results are obtained applying these strategies over a portion of the Barcelona sewage network.

On Hybrid Model Predictive Control of Sewer Networks

Real-time control (RTC) of sewer-network systems plays an important role in meeting increasingly restrictive environmental regulations to reduce release of untreated wastewater to the environment. This chapter presents the application of hybrid model predictive control (HMPC) on sewer systems. It is known from the literature that HMPC has a computational complexity growing exponentially with the size of the system to be controlled. However, the average solution time of modern mixed integer program (MIP) solvers is often much better than the predicted worst-case-solution time. The problem is to know when the worst-case computational complexity appears. In addition to presenting the application, a secondary aim of the chapter is to discuss the limits of applicability due to real-time constraints on computation time when HMPC is applied on large-scale systems such as sewer networks. By using a case study of a portion of the Barcelona sewer system, it is demonstrated how the computational complexity of HMPC appears for certain state and disturbance combinations.

Model based detection of hydrogen leaks in a fuel cell stack

Hydrogen leaks are potentially dangerous faults in fuel cell systems. The paper presents an approach to detect hydrogen leaks. The method is applicable during startup and shutdown as well as normal operating conditions. The method relies on simple mass balance equations but takes into account the natural leak of the stack and humidity. Hydrogen leak detection without using relative humidity sensors is specially studied. In that case, adaptive alarm thresholds are given so that false alarms due to the lack of humidity sensors are eliminated. The validity of the method is also discussed in terms of common hydrogen supply system configurations. The detection method is validated on an real fuel cell laboratory rig where leaks could be introduced in a controlled manner.