Strejc method¶

Identify an aperiodic process from step-response timing by fitting a restricted inertial transfer-function model.

Version: v1 · 2026-10-17 · History
Domain-specific #: 14318
Domain group: Applied Sciences & Engineering
Origin domain: Engineering & Design (beyond software)
Subdomains: Control Engineering, Process Identification → Engineering & Design (beyond software)
Aliases: Strejc identification method

Core Idea¶

The Strejc method identifies a compact dynamic model of a process from its response to a step change. The analyst reads characteristic times from an aperiodic input-output response, uses a prescribed comparison to choose an inertial transfer-function form, and estimates its parameters. The result is an approximation of the observed process at the tested operating conditions, not a discovery of every physical lag inside it.^[1]^[2]

The name covers more than one documented construction. A graphical account uses the tangent near a response's inflection point and a characteristic-time table to fit repeated equal lags with any needed compensating delay.^[1] A later research account explicitly distinguishes that equal-lag family from a double-inertia variant with two separately estimated lag constants and a time read at 70 percent of the response's final change.^[2] Both remain step-response identification procedures. Neither the tangent construction nor equal lag constants are necessary across every use of the unqualified name.

Structural Signature¶

Step-excited aperiodic process — an input step and corresponding output trace provide the transient to be modeled. If no usable response is available, these characteristic-time constructions cannot be applied; a model obtained another way does not establish use of this method.^[1]^[2]
Characteristic-time readout — times are taken from the response geometry. One documented construction uses an inflection-region tangent; another reads a specified fraction of the final response. The exact marker belongs to the chosen variant, not to an all-variant formula.^[1]^[2]
Restricted inertial-model comparison — the graphical variant compares a response-derived time ratio with theoretical ratios to choose an order; the double-inertia variant reads the normalized response height at one quarter of the time when the response reaches 70 percent of its final change, then matches that height to a table to infer the two-lag ratio. This distinguishes the operation from fitting any transfer function to the same data by an arbitrary optimizer.^[1]^[2]
Parameterized approximation — gain and lag parameters, plus a delay where the chosen form calls for one, make a specific model whose step response can be compared with the observed trace. A nonzero delay is not required of the double-inertia variant or of every equal-lag fit.^[1]^[2]

What It Is Not¶

Not the Strejc model by itself. An inertial transfer function is an output or candidate form. A formula on a page does not show that a process was identified from a step response.
Not solely the inflection-tangent, equal-lag construction. That is one documented version. A separate Strejc-named double-inertia version uses a different response marker and can return unequal lags.^[2]
Not unrestricted system identification. Least-squares estimation, frequency-response fitting and other procedures can produce dynamic models without this response-to-restricted-model comparison.
Not an exact account of the plant's internals. Equal fitted time constants need not be separately measured physical time constants. A heat-exchanger study reports asymmetric dynamics and parameter intervals across several input steps.^[1]^[3]
Not a license to fit oscillatory behavior as a simple aperiodic response. The cited variant account limits the original approach to aperiodic step responses and describes complex or oscillatory transients as troublesome for its simple fits.^[2]

Scope of Application¶

The method belongs to control-engineering identification where an input can be stepped, an output trajectory can be observed, and a compact inertial model serves analysis or controller design. Published applications include heat-exchanger temperature modeling from heating-water-flow steps, coupled-tank level modeling, and a pilot reactor's temperature response.^[3]^[4]^[5] These are different processes, yet the input-step, response-time and fitted-dynamics roles are literal in each.

The application papers do not all print their intermediate calculations. The heat-exchanger article reports that it used Strejc identification from step responses and gives a third-order plus-delay fit; it does not publish each tangent or table choice.^[3] The tank article defines a tangent method and reports a fitted model under that heading but does not show the case-specific tangent measurements; its open-loop calculation also uses earlier researchers' data.^[4] The reactor article labels a graphical Strejc construction and reports two fitted, unequal lag constants, but does not print each case's intermediate Strejc time readout.^[5] These are attested uses, not independently reconstructed step-by-step demonstrations.

Clarity¶

“Identified a Strejc model” can refer to an observed procedure, a reported fitted model, or a later optimization of that model. Those claims have different evidence requirements. A step-response trace plus a named fit supports an application claim; it does not prove which intermediate times were read unless the study prints them. Likewise, an equal-lag approximation and an unequal two-lag approximation can both bear the Strejc name. Naming the variant resolves what the reported time constants mean and prevents a particular formula from becoming the whole identity.^[2]^[5]

Manages Complexity¶

A process response can reflect many unobserved time scales. The graphical equal-lag construction replaces individually inaccessible constants with a repeated-lag approximation and a possible residual delay; gain scales the response magnitude.^[1] The double-inertia construction instead represents the response with two estimated lag constants.^[2] Both reduce a full transient trace to a few parameters that can be compared, simulated and used for control design. This compression sacrifices detail: the heat-exchanger study reports different parameter intervals when the flow step changes size, and the reactor reports different fitted constants at different temperature ranges.^[3]^[5]

Abstract Reasoning¶

First decide whether the observed response is sufficiently aperiodic and well resolved for the chosen construction. Then identify the input change, output baseline and final change, read the variant's characteristic time or times, and derive the quantity that its table compares: a characteristic-time ratio in the graphical variant or a normalized response height at one quarter of the 70-percent time in the double-inertia variant. Use that prescribed comparison to select an inertial form and lag parameters. Estimate gain from the relative input and output changes; include a delay only when the chosen form and observed timing warrant it. Finally compare the model's response with the process response over the operating range of interest.^[1]^[2]

This chain lets an engineer ask more precise questions than “does the curve look similar?” If a fitted model works for one step size and shifts for another, the discrepancy may reflect nonlinear operating behavior rather than an error in arithmetic. If a reported paper gives only its final model, the reader can discuss its reported use of Strejc but cannot audit the particular graphical readout. If the response oscillates, a simple aperiodic Strejc fit is outside the supported scope.^[3]^[2]

Knowledge Transfer¶

The procedure can transfer literally from a heating-water/kerosene exchanger model to tank-level or reactor-temperature dynamics because each has a process input, a time-domain output response and a control-oriented model target.^[3]^[4]^[5] The chosen variant, operating range and provenance of the response must travel with the claim. The tank paper's previously collected open-loop data and simulated controller comparison cannot be relabeled as a new physical-plant validation.^[4]

Outside process dynamics, “take a tangent and fit a curve” is an analogy. It does not retain the named method's step-response, inertial-model and time-constant semantics. A general idea of approximation or comparison may recur elsewhere, but the Strejc name does not follow that abstract resemblance.

Examples¶

Canonical: pilot-reactor temperature identification¶

Diallo and colleagues report 10 and 15 percent excitation steps on a pilot chemical reactor and temperature responses used to identify the process. They discuss both Broïda and Strejc models, illustrate a Strejc graphical construction in Figure 4, and report two unequal Strejc time constants across several temperature ranges in Table I. The paper's \(t_1\) and \(t_2\) table columns belong to the Broïda calculation; they are not evidence of case-specific Strejc time readings. The individual Strejc readout and comparison calculations remain unreported.^[5]

Mapped back: reactor excitation → temperature response → Strejc graphical construction identified in the paper, with case-specific markers unreported → restricted two-lag model named by the authors → unequal fitted lag constants in the results table. This is a physical-process example, but the table is not a complete audit trail for the intermediate Strejc calculation.

Applied: heat-exchanger model and coupled-tank level¶

For a co-current tubular heat-exchanger model, Vasičkaninová and Bakošová generate several heating-water-flow steps and show outlet kerosene-temperature responses. They say the transfer-function approximation was identified by the Strejc method and report order three with gain, common-time-constant and delay intervals. Their paper does not give the characteristic-time readings for each fit, and the process/controller comparison is model-based and simulated.^[3]

Mapped back: heating-water-flow steps → model temperature traces → Strejc identification reported, with the intermediate time comparison unprinted → equal-lag order-three plus-delay form → interval-valued fitted parameters. The multiple step sizes test how far one nominal approximation transfers.

In a different fluid-level setting, Faizal's coupled-tank article defines Strejc identification through a tangent at a step-response inflection and provides a parameter table, then reports a nominal fitted tank model under a Strejc heading. It cites earlier rig calculations and uses earlier researchers' data in a separate open-loop calculation; the controller results are simulations. The specific nominal-fit tangent measurements and table selection are not printed.^[4]

Mapped back: coupled-tank rig context and previously available response information → generic tangent/table rule stated in the paper → particular intermediate choices unreported → reported tank-level model used in a simulated controller comparison. The generic method description is not substituted for a missing case-specific observation.

Structural Tensions¶

T1: compact fit versus distinct process lags. A repeated-lag approximation with any needed delay reduces the response to a simple model, but a process with distinct lag scales can be better described by a two-lag form. Both forms are still approximations and must be compared with observations; the reactor's unequal constants and the heat exchanger's changing parameter intervals show why one convenient nominal fit need not be universal.^[1]^[5]^[3] Diagnostic: Which variant represents the observed response adequately at the operating conditions that matter?

T2: transferable method versus operating-range evidence. The input-step and response-time logic travels among control processes, but its fitted parameters can change when a nonlinear process is tested at different step sizes or regimes. Reusing a nominal model makes design simpler; checking more responses can expose limits that require intervals, multiple local fits or another model.^[3]^[5] Diagnostic: What new operating range has evidence that the same fitted parameters remain adequate?

Structural–Framed Character¶

Strejc identification is strongly structural within control engineering: the same process-response-to-inertial-model operation survives changes of plant, measured quantity and model parameters. Its evaluative weight is limited; the method names a fitting operation, not a guarantee of accuracy. Human-practice dependence enters through experiment design, reading a response curve, choosing a documented variant and judging adequacy. Its institutional origin is process and control engineering, while its operation does not depend on one manufacturer or laboratory. Its vocabulary travels only where step responses, transfer functions and time constants retain their technical meanings. Under the import-versus-recognition test, an engineer recognizes the method when those roles are performed; a generic curve fit elsewhere imports a metaphor. Its character: a structurally reusable identification rule within control engineering, still framed by that field's step-response and inertial-model practice rather than a substrate-independent rule for every curve.

Structural Core vs. Domain Accent¶

The structural core is a constrained inference from observed transient behavior to a compact model. Comparison is an internal constituent: a response-derived time ratio or normalized response height is aligned with a candidate inertial-model table value under the chosen variant's rule, yielding an order or lag relation. The resulting transfer function approximates the process in the ordinary engineering sense. The live Prime Approximation has additional quantified-error and tolerance roles that are not established for every reported Strejc use, so it is not asserted as a typed parent here. The domain accent is constitutive, not decorative: a controlled step, an aperiodic dynamic response, inertial transfer functions, time constants and variant-specific parameter rules determine whether this is a Strejc identification. Removing those operations leaves only generic curve fitting. The two attested variants share that narrower control-engineering core while differing in their exact time marker and lag structure.^[1]^[2]

The named Strejc method does not clear the Prime bar: its constitutive decision rule remains tied to control-process dynamics. A separate, more general response-to-restricted-model Prime would need evidence from at least two materially unlike non-control settings that use the same time-readout and model-choice operation, with a comparable failure when those roles are removed. A clinical response curve and an ecological perturbation curve would count only if their actual methods met that test; sharing a plot and fitted parameters would show analogy, not an established portable mechanism. No such broader identity is asserted here.

This entry is part of Comparison.

The method contains Comparison as the response-to-model alignment used to select its order or lag relation. A later comparison of the fitted and observed responses can assess adequacy. The fitted model is an approximation in ordinary engineering language, but the live Approximation Prime additionally requires a named quantitative error and usable tolerance; those are not universal documented steps of this method. Measurement is involved when characteristic times and gain are read from a trace, but the method can use previously collected or simulated responses and does not require a new calibrated measurement operation. These related ideas do not supply Strejc's step-response and variant-specific model-choice rule. The catalog's Prime Identification concerns recognition through self-concept and is not the control-engineering sense of “identification.”

Relationships to Other Abstractions¶

Current abstraction Strejc method Domain-specific

Parents (1) — more general patterns this builds on

Strejc method is part of Comparison Prime

A response-derived ratio or normalized response height must be compared with a restricted model table to select the Strejc form and lags.

Hierarchy path (1) — routes to 1 parentless root

Strejc method → Comparison

Neighborhood in Abstraction Space¶

Strejc method sits in a sparse region of the domain-specific corpus (95^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Structural Vibration & Seismic Analysis (13 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-17

Not to Be Confused With¶

System Identification: the broader construction and testing of dynamic models from data; it need not use a Strejc step-response comparison.
Open-Circuit Time-Constant Method: a circuit-specific capacitor-and-resistance procedure for a frequency-corner estimate, not a process step-response Strejc fit.
Broïda identification: a separate model and timing calculation used alongside Strejc in the reactor paper; its \(t_1/t_2\) columns cannot be reassigned to Strejc.^[5]
A fitted Strejc transfer function: a representation of an identified process, not evidence on its own of how the fit was obtained.
A genetic algorithm that later tunes the model: an optional subsequent optimization, not the response-time procedure that supplied its initial parameters in the cited 2024 study.^[2]

References¶

[1] Michel Verbeken, “Cours d'Automatique, les asservissements continus,” §3.7, “Modèle de Strecj (systèmes apériodiques),” educational course page. https://www.iutenligne.net/rsc/rsc-public/automatique-et-automatismes-industriels/verbeken/cours_au_mv/chapitre3/chap37.html . The displayed full course title separates “Automatique” and “les asservissements continus” with a colon. Full chapter page and course home page inspected. It gives the inflection tangent, characteristic-time comparison, repeated-lag approximation, possible delay and gain rule for one graphical variant; it is a teaching account rather than the original 1959 paper. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k

[2] Dawid Ostaszewicz and Krzysztof Rogowski, “Application of Genetic Algorithms for Strejc Model Parameter Tuning,” Electronics 13, no. 18 (2024), article 3652, §2, DOI 10.3390/electronics13183652. https://www.mdpi.com/2079-9292/13/18/3652 . Official publisher-indexed full §2 inspected through search; direct page opening returned HTTP 429. The section distinguishes equal-lag/delay and unequal double-inertia variants; exact table cells and figures were not independently inspected. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n

[3] Anna Vasičkaninová and Monika Bakošová, “Application of H2 and H∞ Approaches to the Robust Controller Design for a Heat Exchanger,” Chemical Engineering Transactions 35 (2013): 463–468, DOI 10.3303/CET1335077, especially p. 464, Figure 1, Eq. (1), Table 1. https://www.aidic.it/cet/13/35/077.pdf . Full publisher PDF text inspected; transfer-function glyphs were corrupt in extraction, so no exact equation is reproduced. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i

[4] Ahmad Faizal, “Pengendalian Level Coupled Tank Menggunakan Metode Sliding Mode Control (SMC) Hybrid Proportional Integral Derivative (PID) di Simulink Matlab,” Jurnal Sains, Teknologi dan Industri 13, no. 1 (issue dated December 2015): 115–122, especially pp. 116–117. https://ejournal.uin-suska.ac.id/index.php/sitekin/article/download/1334/1266 . Full journal PDF text inspected. The corresponding-author block spells the name Ahmad Faisal; the PDF text corrupts equations and one table value. The paper uses earlier data for its open-loop calculation and reports simulated controller results. registry ↩a ↩b ↩c ↩d ↩e

[5] M. B. Diallo, L. Kara, C. Porte and A. Delacroix, “Pilote chimique polyvalent modulaire et automatisé, Étude de sécurité et mise en œuvre d'une architecture de commande,” Cahiers de notes documentaires – Hygiène et sécurité du travail 171 (second quarter 1998), especially printed pp. 160 and 162, §1.3.1, Figure 4 and Table I. https://www.inrs.fr/dam/inrs/CataloguePapier/HST/TI-ND-2078.pdf . The printed title separates “automatisé” and “Étude” with a colon. Full original PDF text inspected. Table I's \(t_1/t_2\) columns belong to Broïda and its \(\tau_1/\tau_2\) columns to Strejc; exact diagram glyphs and transfer-function formula were not visually certified. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i