Operating-point, step-response and reduced-order model data for a multivariable roof-tile tunnel kiln

Context and methodology

This dataset supports research on reduced-order modeling and state-space predictive functional control (SSMPFC) of a multivariable tunnel kiln for industrial roof-tile production. It was prepared within the FurnAIce project and accompanies the manuscript Data-Driven Reduced-Order Modeling and State-Space Predictive Functional Control for a Multivariable Tunnel Kiln.

The data were generated using a proprietary high-fidelity thermodynamic kiln simulator informed by industrial plant data. Fourteen separate experiments apply a small natural-gas flow increment of 1 Nm³/h to one burner input at a time. Product-temperature responses are recorded at 14 fixed axial locations, yielding 196 input–output response channels. The released responses are deviations obtained by subtracting the corresponding nominal simulation trajectory. Sampling is synchronized with the kiln-car pushing cycle at 10-minute intervals; it is not temporal averaging.

The dataset enables reproduction of the paper’s identification results and provides a basis for comparing local modeling and multivariable control methods.

Technical details

The package contains:

• Operating-point temperature profiles and supplied burner-flow reference values.
• Output locations and spatial interpolation weights.
• Fourteen step-response files, each containing 600 samples for all 14 outputs.
• The identified 14 × 14 steady-state gain matrix, 14 dominant time constants, and associated discrete-time state-space matrices.
• Channel-fit statistics, descriptive metadata, source provenance, checksums, and a verification utility.

Numerical data are provided in both MATLAB MAT and plain-text CSV formats. Files named input_01.csv through input_14.csv correspond to the separate input-step experiments. Channel indices are consistent across response files, model parameters, and operating-point tables.

The MAT file can be read using MATLAB, GNU Octave, or Python with SciPy. CSV files require no specialist software. The optional verification script requires Python 3.10 or later, NumPy, and SciPy. A comprehensive README documents variable definitions, array dimensions, units, preprocessing, model equations, and loading examples.

Further details

These are simulator-generated data, not industrial measurements. The proprietary simulator, its implementation and parameter files, and industrial calibration measurements are not included.

The response records retain their original timestamps, starting at 60 seconds with 600-second spacing and ending at 99.85 hours. Temperatures are expressed in degrees Celsius, positions in metres, and burner flows in Nm³/h, referenced to dry gas at 273.15 K and 101.325 kPa absolute.

The data describe local behavior around the investigated operating regime. Reported fitting statistics quantify in-sample approximation rather than independent validation. The dataset does not establish performance for other operating conditions, products, or plants, nor does it demonstrate actuator feasibility or high-fidelity closed-loop control performance.

The dataset is available under the Creative Commons Attribution-ShareAlike 4.0 International license (CC BY-SA 4.0). Please cite the dataset using DOI 10.48436/vc2bj-bdf34.