ABAbdelmajid Benahmed

Abdelmajid Benahmed

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Latest Publications

  • Open Access

    Hierarchical Bayesian inference of β-lactoglobulin denaturation and lactic acidification kinetics for a stirred-yoghurt line: a meta-analysis of published kinetics and its consequences for process design.

    by Abdelmajid Benahmed

    In an earlier preprint I optimised a stirred-yoghurt line with a deterministic model, and the optimiser pushed pasteurisation to the upper bounds of temperature and holding time. That result rests on single values of the kinetic constants. Published work on β-lactoglobulin (β-Lg) denaturation in milk does not agree on those constants: reported orders run from 1 to 2, and the Arrhenius break sits anywhere between 83 and 95◦C. Here I treat the disagreement as uncertainty and follow it through to the design. The model part gives a normalised two-regime Arrhenius law with a smooth break, a rational lifting for symbolic identifiability analysis, the gradient of the stopping time at pH 4.60, and a chance-constrained version of the optimisation. The data part is a hierarchical Bayesian meta-analysis of four studies, seven entries in all. The data are of three kinds: 107 points of the Dannenberg and Kessler curves, digitised from a published re-plot; nine isothermal rate constants from Claeys et al.; and the Arrhenius regressions of Manji and Kakuda and of Oldfield. The population activation energy is 280 kJmol−1 (95% interval 261–299) below the break and 51 kJmol−1 (44–62) above it. The break lies at 89.6 ◦C (86.6–92.4), and the mean order is 1.56 (1.21–1.90). At 90◦C the rate constants of the seven entries span a factor of four. For a milk that has not been characterised, the predicted denaturation after 300 s at 90 ◦C has a 95% interval from 0.54 to 1.00. The value used in the earlier preprint, 0.80, falls at the 18th percentile. StructuralIdentifiability.jl shows that one non-isothermal curve identifies Ea/R globally. From pH data alone, the growth rate, the lag variable and the product of the buffering peak with its shape factor are locally identifiable. Uncertainty does not move the optimum away from 95◦C and 600 s. If the plant also sets a ceiling on denaturation, no operating point meets both chance constraints until the plant’s own milk has been measured. With two (simulated) measurements on that milk, a ceiling of 0.92 is met at 92.5 ◦C and 420 s. Batch variability means that four tanks sized deterministically are too small for about 62% of simulated batches.

  • Open Access

    A coupled mechanistic and reservoir-computing model of a stirred-yoghurt line built from public data

    by Abdelmajid Benahmed

    This paper couples four standard models of a stirred-yoghurt line (denaturation of βlactoglobulin during heat treatment, growth and acidification by the starter, texture and syneresis of the stirred gel, specific energy of the line) and uses the result to pick an operating point and size the main equipment. An echo state network corrects the residual between model and plant pH. All parameters come from the literature. The plant context is a large dairy site in Akbou, Algeria, known to me only through public sources: I had no access to company data, and the signals used to train the correction are synthetic. Under these assumptions the batch reaches pH 4.6 in 4.09 h at 42◦C. The optimizer drives the heat treatment to 95◦C and 600 s and the smoothing shear down to 60 s−1, which lowers predicted syneresis from 12.4% to 8.7% and specific energy by 1.8%. Every starting point ends in this corner, mostly because fouling, cooked flavour and the cost of the holding tube are missing from the objective. The texture and syneresis laws have not been fitted to measurements. I state which parts of the model are established and which are assumed, and give a calibration protocol; the notebook and scripts are attached.

    • 20 November 2026
    • Published by MDPI AG in Preprints.org
      • https://doi.org/10.20944/preprints202611.2162.v1
    Open Access

    A Coupled Mechanistic and Reservoir-Computing Model of a Stirred-Yoghurt Line Built from Public Data

    by Abdelmajid Benahmed

    This paper couples four standard models of a stirred-yoghurt line (denaturation of β-lactoglobulin during heat treatment, growth and acidification by the starter, texture and syneresis of the stirred gel, specific energy of the line) and uses the result to pick an operating point and size the main equipment. An echo state network corrects the residual between model and plant pH. All parameters come from the literature. The plant context is a large dairy site in Akbou, Algeria, known to me only through public sources: I had no access to company data, and the signals used to train the correction are synthetic. Under these assumptions the batch reaches pH 4.6 in 4.09 h at 42 ◦C. The optimizer drives the heat treatment to 95 ◦C and 600 s and the smoothing shear down to 60 s−1, which lowers predicted syneresis from 12.4% to 8.7% and specific energy by 1.8%. Every starting point ends in this corner, mostly because fouling, cooked flavour and the cost of the holding tube are missing from the objective. The texture and syneresis laws have not been fitted to measurements. I state which parts of the model are established and which are assumed, and give a calibration protocol; the notebook and scripts are attached.

    • 20 November 2026
    • Published by MDPI AG in Preprints.org
      • https://doi.org/10.20944/preprints202611.2131.v1
    Open Access

    Hierarchical Bayesian Inference of β-Lactoglobulin Denaturation and Lactic Acidification Kinetics for a Stirred-Yoghurt Line: A Meta-Analysis of Published Kinetics and Its Consequences for Process Design

    by Abdelmajid Benahmed

    In an earlier preprint I optimised a stirred-yoghurt line with a deterministic model, and the optimiser pushed pasteurisation to the upper bounds of temperature and holding time. That result rests on single values of the kinetic constants. Published work on β-lactoglobulin (β-Lg) denaturation in milk does not agree on those constants: reported orders run from 1 to 2, and the Arrhenius break sits anywhere between 83 and 95 ◦C. Here I treat the disagreement as uncertainty and follow it through to the design. The model part gives a normalised two-regime Arrhenius law with a smooth break, a rational lifting for symbolic identifiability analysis, the gradient of the stopping time at pH 4.60, and a chance-constrained version of the optimisation. The data part is a hierarchical Bayesian meta-analysis of four studies, seven entries in all. The data are of three kinds: 107 points of the Dannenberg and Kessler curves, digitised from a published re-plot; nine isothermal rate constants from Claeys et al.; and the Arrhenius regressions of Manji and Kakuda and of Oldfield. The population activation energy is 280 kJ mol−1 (95 % interval 261–299) below the break and 51 kJ mol−1 (44–62) above it. The break lies at 89.6 ◦C (86.6–92.4), and the mean order is 1.56 (1.21–1.90). At 90 ◦C the rate constants of the seven entries span a factor of four. For a milk that has not been characterised, the predicted denaturation after 300 s at 90 ◦C has a 95 % interval from 0.54 to 1.00. The value used in the earlier preprint, 0.80, falls at the 18th percentile. StructuralIdentifiability.jl shows that one non-isothermal curve identifies Ea/R globally. From pH data alone, the growth rate, the lag variable and the product of the buffering peak with its shape factor are locally identifiable. Uncertainty does not move the optimum away from 95 ◦C and 600 s. If the plant also sets a ceiling on denaturation, no operating point meets both chance constraints until the plant’s own milk has been measured. With two (simulated) measurements on that milk, a ceiling of 0.92 is met at 92.5 ◦C and 420 s. Batch variability means that four tanks sized deterministically are too small for about 62 % of simulated batches.

  • Open Access

    A coupled mechanistic and reservoir-computing model of a stirred-yoghurt line built from public data

    by Abdelmajid Benahmed

    This paper couples four standard models of a stirred-yoghurt line (denaturation of βlactoglobulin during heat treatment, growth and acidification by the starter, texture and syneresis of the stirred gel, specific energy of the line) and uses the result to pick an operating point and size the main equipment. An echo state network corrects the residual between model and plant pH. All parameters come from the literature. The plant context is a large dairy site in Akbou, Algeria, known to me only through public sources: I had no access to company data, and the signals used to train the correction are synthetic. Under these assumptions the batch reaches pH 4.6 in 4.09 h at 42◦C. The optimizer drives the heat treatment to 95◦C and 600 s and the smoothing shear down to 60 s−1, which lowers predicted syneresis from 12.4% to 8.7% and specific energy by 1.8%. Every starting point ends in this corner, mostly because fouling, cooked flavour and the cost of the holding tube are missing from the objective. The texture and syneresis laws have not been fitted to measurements. I state which parts of the model are established and which are assumed, and give a calibration protocol; the notebook and scripts are attached.

  • Open Access

    Hierarchical Bayesian inference of β-lactoglobulin denaturation and lactic acidification kinetics for a stirred-yoghurt line: a meta-analysis of published kinetics and its consequences for process design

    by Abdelmajid Benahmed

    In an earlier preprint I optimised a stirred-yoghurt line with a deterministic model, and the optimiser pushed pasteurisation to the upper bounds of temperature and holding time. That result rests on single values of the kinetic constants. Published work on β-lactoglobulin (β-Lg) denaturation in milk does not agree on those constants: reported orders run from 1 to 2, and the Arrhenius break sits anywhere between 83 and 95◦C. Here I treat the disagreement as uncertainty and follow it through to the design. The model part gives a normalised two-regime Arrhenius law with a smooth break, a rational lifting for symbolic identifiability analysis, the gradient of the stopping time at pH 4.60, and a chance-constrained version of the optimisation. The data part is a hierarchical Bayesian meta-analysis of four studies, seven entries in all. The data are of three kinds: 107 points of the Dannenberg and Kessler curves, digitised from a published re-plot; nine isothermal rate constants from Claeys et al.; and the Arrhenius regressions of Manji and Kakuda and of Oldfield. The population activation energy is 280 kJmol−1 (95% interval 261–299) below the break and 51 kJmol−1 (44–62) above it. The break lies at 89.6 ◦C (86.6–92.4), and the mean order is 1.56 (1.21–1.90). At 90◦C the rate constants of the seven entries span a factor of four. For a milk that has not been characterised, the predicted denaturation after 300 s at 90 ◦C has a 95% interval from 0.54 to 1.00. The value used in the earlier preprint, 0.80, falls at the 18th percentile. StructuralIdentifiability.jl shows that one non-isothermal curve identifies Ea/R globally. From pH data alone, the growth rate, the lag variable and the product of the buffering peak with its shape factor are locally identifiable. Uncertainty does not move the optimum away from 95◦C and 600 s. If the plant also sets a ceiling on denaturation, no operating point meets both chance constraints until the plant’s own milk has been measured. With two (simulated) measurements on that milk, a ceiling of 0.92 is met at 92.5 ◦C and 420 s. Batch variability means that four tanks sized deterministically are too small for about 62% of simulated batches.

  • Open Access

    Rank-Adaptive IMEX Tensor-Train BUG Integrators for High-Dimensional Fractional Fokker–Planck Equations

    by Abdelmajid Benahmed

    We consider nonlinear fractional Fokker--Planck equations of McKean--Vlasov type in dimension \(D \gg 10\). Three difficulties occur at the same time. A tensor-product grid needs \(O(n^D)\) storage. The discrete Riesz operators of order \(\alpha\in(0,2)\) have condition numbers that grow like \(n^\alpha\). Finally, the curvature of the manifold of tensor trains (TT) of fixed rank behaves like \(\sigma_{\min}^{-1}\) and becomes large as soon as one singular value is small. Explicit projected integrators are then limited to steps \(\Delta t=O(\min(\sigma_{\min},n^{-\alpha}))\). The projector-splitting integrator is insensitive to small singular values, but its \(S\)-step runs backward in time, which is unstable when the operator is dissipative. We propose a rank-adaptive implicit--explicit integrator of basis-update & Galerkin (BUG) type, written directly in TT format (IMEX-TT-BUG). In each step, the separable fractional drift--diffusion part is treated implicitly in a left-to-right sweep of \(K\)-steps, and the convolution and reaction terms are treated explicitly. Each updated core enlarges a nested left basis of bond dimension at most \(2r\); a Galerkin problem for the last core is then solved in the linear space these bases span, and the result is truncated by TT-SVD. All substeps go forward in time, and the \(O((2r)^D)\) core of the Tucker version of BUG never appears. We prove a Frobenius-norm error bound in which the amplification constants depend neither on \(\sigma_{\min}^{-1}\) nor on \(\kappa_2(L)\). The bound also contains a projection residual that depends on the trajectory; one of its terms involves \(L\), and we state precisely when it stays bounded under mesh refinement. The linear part of a step costs \(O(Dnr^3+Dn\log n\,r^2N_{\mathrm{kry}})\) operations; evaluating the nonlinear term in TT format adds \(O(DnM_W^3r^6)\) with direct rounding, where \(M_W\) is the separation rank of the interaction kernel. We also describe a Julia implementation, based on circulant embedding of the Toeplitz matrices and a preallocated memory arena, designed so that the time loop performs no heap allocation. Experiments with a reference implementation in dimension up to \(20\) show first-order convergence with errors that do not depend on the mesh, stability for steps several hundred times the explicit limit, and correct damping of a mode whose singular value is \(10^{-12}\).

  • Open Access

    Numerical Analysis of Prabhakar Fractional Differential Equations: Error Bounds, Stability, and Green's Function Methods

    by Abdelmajid Benahmed

    Prabhakar fractional PDEs require discretization of singular Mittag-Leffler kernels. This paper establishes three results: (1) Lemma proving singularity extraction preserves O(∆x 4) convergence via analytical log-integral and Simpson quadrature. (2) Theorem with composite error bound O(∆t 2 + ∆x 4 + exp(-ρn max)). (3) Stability validation across 125 parameter combinations (α × β × γ grid). Tests include grid refinement (Zeng comparison < 0.1%), discontinuous data (2nd-order), and 20 manufactured solutions across parameter space. Independent verification of Karimov et al. (2025) Green's function via classical heat equation limit (agreement 4.0 × 10-10) confirms construction.

  • Open Access

    Algorithmic Resilience Under Resource Constraints: The Novosibirsk School and the Method of Fractional Steps (1955–1975)

    by Abdelmajid Benahmed

    This article examines the development of operator splitting methods in Soviet numerical analysis during 1955–1975, with particular focus on N.N. Yanenko’s formalization of the Method of Fractional Steps at the Siberian Branch of the USSR Academy of Sciences. While similar techniques were independently developed in the West (Peaceman-Rachford 1955, Douglas-Rachford 1956), the Soviet school pursued a distinct trajectory shaped by acute hardware constraints and deep epistemological commitments to operator theory. Through analysis of technical publications, archival materials, and comparative historiography, this study argues that material scarcity catalyzed a systematic research program emphasizing computational economy, while a pre-existing mathematical culture valorizing theoretical elegance reinforced this trajectory. The case illuminates how geopolitical constraints and intellectual traditions jointly shaped algorithmic innovation, contributing to methods that ironically became foundational for modern massively parallel computing. Significant archival gaps limit definitive claims about industrial applications, highlighting the need for further primary source research.

  • Open Access

    Numerical Analysis of Prabhakar Fractional Differential Equations: Error Bounds, Stability, and Green’s Function Methods

    by Abdelmajid Benahmed

    Prabhakar fractional PDEs require discretization of singular Mittag-Leffler kernels. This paper establishes three results: (1) Lemma proving singularity extraction preserves O(∆x 4 ) convergence via analytical log-integral and Simpson quadrature. (2) Theorem with composite error bound O(∆t 2 + ∆x 4 + exp(−ρnmax)). (3) Stability validation across 125 parameter combinations (α × β × γ grid). Tests include grid refinement (Zeng comparison < 0.1%), discontinuous data (2nd-order), and 20 manufactured solutions across parameter space. Independent verification of Karimov et al. (2025) Green’s function via classical heat equation limit (agreement 4.0 × 10−10) confirms construction.