A Spectral-Based Tuning Criterion for PI Controllers in IPDT Systems With Unified Tracking and Disturbance Rejection Performance
Dhamdhawach Horsuwan · 2025
Read A Spectral-Based Tuning Criterion for PI Controllers in IPDT Systems With Unified Tracking and Disturbance Rejection Performance by Dhamdhawach Horsuwan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-01197.txt.
Introduction
Dhamdhawach Horsuwan tunes a common industrial controller by its spectrum in a 2025 paper. The controller is the proportional-integral type, a two-knob workhorse, and the plant is the awkward kind that integrates its input and answers late, called IPDT. The goal is one tuning that makes both following a setpoint and shrugging off disturbances decay at the same clean exponential rate.
His method minimizes the spectral abscissa, the slowest decay rate among the closed loop's poles. A semi-discrete model captures the integrator and the delay while letting the dominant poles be pulled out cheaply. Those poles are mapped back to continuous time and polished against the exact delay equation. The result is a single gain set with no hand-set trade-off weights. Simulations show faster settling and better safety margins than the classic tuning rules.
Catalog record
| CREATOR | Dhamdhawach Horsuwan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-01197
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Spectral-Based Tuning Criterion for PI Controllers in IPDT Systems With Unified Tracking and Disturbance Rejection Performance Dhamdhawach Horsuwan Abstract This paper proposes a spectral-based tuning method for proportional–integral (PI) controllers in integrating-plus-dead-time (IPDT) systems. The design objective is to achieve unified exponential decay for both reference tracking and disturbance rejection by minimizing the spectral abscissa of the closed-loop system. A second-order semi-discrete model accurately captures the integrator and delay dynamics while enabling efficient dominant pole extraction. These discrete-time poles are mapped to continuous time and refined using Newton–Raphson iterations on the exact transcendental characteristic equation. The method produces a unique PI gain set without requiring heuristic trade-offs or weighting parameters. Comparative simulations demonstrate that the proposed tuning achieves faster convergence and improved robustness margins compared to classical rules (Ziegler–Nichols, SIMC) and integral performance criteria (IAE, ITAE). The approach provides a transparent and computationally efficient framework for PI control in delay-dominant systems. I Introduction Proportional–Integral (PI) controllers remain the predominant choice in industrial process control due to their simplicity, ease of implementation, and ability to achieve both reference tracking and disturbance rejection. As the minimal feedback configuration capable of eliminating steady-state error in regulating integrating or type-1 systems, PI controllers are widely employed across diverse industrial applications. Many processes can be modeled as either First-Order Plus Dead Time (FOPDT) or Integrating Plus Dead Time (IPDT) systems. Under conditions of high integral gain or small process time constants, FOPDT dynamics often approximate those of an IPDT model. Consequently, the IPDT framework offers both generality and analytical convenience for analyzing delay-dominant systems. Time delays are pervasive in practical control systems, arising from sensor dynamics, actuator backlash, communication latencies, or transport phenomena such as conveyor systems. These delays can significantly degrade control performance and, if not properly accounted for, may compromise closed-loop stability. Despite their practical importance, delays remain challenging to address rigorously within classical control design frameworks. For IPDT systems under proportional-only control, the ultimate gain K_{u} and oscillation period P_{u} are directly related to the process parameters: K_{u}=\frac{\pi}{2LK},\quad P_{u}=4L, where K is the process gain and L is the time delay. These expressions form the basis for ultimate sensitivity tuning and process identification. Multiple tuning strategies have been proposed for the IPDT configuration. Classical methods, such as Ziegler–Nichols (ZN) [ 1 ] , Cohen–Coon [ 2 ] , and Chien–Hrones–Reswick (CHR) [ 3 ] , offer heuristic rules derived from step response characteristics or sustained oscillations. While straightforward to implement, these approaches often compromise performance optimality. More systematic approaches, such as the SIMC method [ 4 ] , yield explicit tuning formulas based on reduced-order models. However, these formulations typically rely on first-order dynamics and crude delay approximations, which may oversimplify the underlying system. Frequency-domain techniques, including phase margin, gain margin, and bandwidth specifications, offer useful robustness metrics but do not directly quantify time-domain behaviors such as reference tracking and disturbance rejection. Likewise, graphical methods such as root-locus and pole-placement provide valuable design insights, but when applied to systems with transport delays, they almost universally require finite-order Padé approximations [ 5 ] , which obscure the infinite-dimensional nature of delay dynamics and may produce misleading conclusions. To address these limitations, numerical methods such as Chebyshev collocation [ 6 ] and semi-discretization [ 7 ] have been developed to discretize the time domain and approximate the system’s infinite spectrum. These methods enable estimation of the dominant (rightmost) poles, facilitating stability and performance analysis for delay systems under a given controller. Such numerical tools support continuous-time pole placement [ 8 ] and PI tuning based on spectral abscissa optimization, which directly minimizes the real part of the slowest decaying closed-loop pole as a unified performance criterion [ 9 , 10 ] . This formulation naturally links controller design to the system’s exponential decay characteristics. In this paper, we extend this framework by formulating a PI controller design for IPDT systems that explicitly minimizes the spectral abscissa. Since the dominant pole governs the system’s slowest convergence mode, the proposed method ensures that both reference tracking and disturbance rejection exhibit identical exponential decay rates. Importantly, this approach eliminates the need for manually selected performance weights as required in IAE- or ITAE-based multi-objective designs. As shown in [ 11 ] , assigning weights between tracking and disturbance rejection often involves heuristic trade-offs. In contrast, the proposed spectral criterion provides a unified, dynamics-based performance measure that simultaneously addresses both objectives. II System Model and Problem Formulation We consider an Integrating Plus Dead Time (IPDT) process, which serves as a representative model for many slow industrial systems, including temperature regulation, liquid-level control, and other transport-dominated processes. The process is described by the transfer function: G(s)=\frac{K}{s}e^{-Ls}, (1) where K>0 is the process gain and L>0 is the time delay. The presence of the exponential delay term e^{-Ls} introduces an infinite set of characteristic roots, which complicates both stability analysis and controller design [ 12 ] . The controller under consideration is a standard proportional–integral (PI) compensator: C(s)=K_{P}+\frac{K_{I}}{s}, (2) where K_{P} and K_{I} are the proportional and integral gains, respectively. Applying unity feedback yields the closed-loop characteristic equation: 1+C(s)G(s)=1+\left(K_{P}+\frac{K_{I}}{s}\right)\frac{K}{s}e^{-Ls}=0. (3) Due to the presence of the delay term, ( 3 ) is transcendental and admits infinitely many roots in the complex plane. The closed-loop dynamics are primarily governed by the dominant roots, i.e., those with the largest real parts, which determine asymptotic stability and convergence rates. Identifying these dominant poles generally requires numerical methods. Two prominent approaches are the Chebyshev collocation method [ 9 , 10 ] , implemented for example in MATLAB’s BIFTOOL, and the semi-discretization method [ 7 ] . In this work, we adopt the semi-discretization approach with second-order accuracy to approximate the delay system dynamics. The resulting finite-dimensional model provides initial estimates of the dominant poles. These estimates are subsequently refined using a Newton-based root-finding algorithm applied directly to the exact transcendental characteristic equation ( 3 ), yielding high-precision localization of the dominant poles and enabling smooth continuation with respect to controller parameters. III Discrete-Time Approximation and Spectral Analysis To efficiently explore admissible PI gain configurations and accurately predict dominant poles, we construct a reduced-order discrete-time approximation of the closed-loop IPDT system. In contrast to Padé approximations or finite-dimensional lifted representations of delay systems [ 12 ] , this formulation retains the input delay explicitly and leverages the structure of the PI controller to minimize model dimensionality. The input delay L is discretized into M equal segments of duration h=L/M , which we normalize as h=1/M . The model tracks the current error e[k] together with a memory buffer of the most recent M+1 control inputs u[k],u[k-1],\ldots,u[k-M] . The integral state s[k] is not stored explicitly but is computed recursively using a fourth-order accurate numerical integration [ 13 ] . This yields a compact state vector of dimension M+2 : x[k]=\begin{bmatrix}e[k]&u[k]&u[k-1]&\cdots&u[k-M]\end{bmatrix}^{\top}. (4) The system dynamics follow from the IPDT relation e^{\prime}[k]=-Ku[k-M] , combined with the PI control law u[k]=K_{P}e[k]+K_{I}s[k] . The error is updated via a second-order finite-difference approximation: e[k+1]=e[k]-\frac{hK}{2}\left(u[k-M]+u[k-M+1]\right). (5) The integral state is updated using fourth-order numerical integration: \displaystyle s[k+1]=\; \displaystyle s[k]+\tfrac{h}{2}(e[k]+e[k+1]) \displaystyle-\tfrac{Kh^{2}}{12}(u[k-M+1]-u[k-M]). (6) Substituting into the control law yields: u[k+1]=K_{P}e[k+1]+K_{I}s[k+1], (7) which enables full state propagation from known quantities at time step k