pith. machine review for the scientific record. sign in

arxiv: 2605.08438 · v1 · submitted 2026-05-08 · 🧮 math.OC · cs.SY· eess.SY

Recognition: 2 theorem links

· Lean Theorem

Generalized Global Self-Optimizing Control for Chemical Processes: Part II Objective-Guided Controlled Variable Learning Approach

Chenchen Zhou, Hongxin Su, Lingjian Ye, Shuang-hua Yang, Xinhui Tang, Yi Cao

Pith reviewed 2026-05-12 01:20 UTC · model grok-4.3

classification 🧮 math.OC cs.SYeess.SY
keywords self-optimizing controlcontrolled variablesglobal SOCchemical processesnonlinear CV designsymbolic-numerical integrationOGCVLscalable design
0
0 comments X

The pith

The OGCVL algorithm designs scalable nonlinear controlled variables for global self-optimizing control of chemical processes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper proposes the objective-guided controlled variable learning (OGCVL) method as the second part of generalized global self-optimizing control. It seeks to select general nonlinear functions as controlled variables that keep process operation near optimal over the entire operating space. The approach merges the performance strengths of optimization-based designs with the scalability of regression-based ones by integrating symbolic and numerical computation. A reader would care because it addresses the practical barrier of computational cost in large systems while avoiding the limitations of linear combinations traditionally used in self-optimizing control.

Core claim

The OGCVL method guides the learning of controlled variables directly by the process objective, allowing seamless symbolic-numerical integration to produce high-quality nonlinear CVs. This overcomes the infeasibility of pure optimization-based g2SOC for large-scale problems while retaining its advantages over regression-based alternatives, as demonstrated by good results and maintained efficiency in two numerical examples.

What carries the argument

The OGCVL algorithm, an objective-guided procedure that learns general nonlinear controlled variables by blending symbolic manipulation with numerical optimization to achieve scalable self-optimizing design.

Load-bearing premise

The assumption that perfect self-optimizing controlled variables exist in theory and that their practical approximation via seamless symbolic-numerical integration will produce high-quality, scalable results in real nonlinear chemical systems without hidden accuracy or computational trade-offs.

What would settle it

A large-scale nonlinear process example in which OGCVL either produces controlled variables with significantly higher economic loss than the optimization-based method or requires computation time comparable to or exceeding that method.

Figures

Figures reproduced from arXiv: 2605.08438 by Chenchen Zhou, Hongxin Su, Lingjian Ye, Shuang-hua Yang, Xinhui Tang, Yi Cao.

Figure 1
Figure 1. Figure 1: Illustration of the three approaches (a) Regression-based approach (b) [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Flow chart of the Objective-guided Controlled Variables Learning framework [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Variation of active constraints Under the assumed disturbance range of FA ∈ [1, 2]kg/s, the purity constraint on xG is always active. As shown in [PITH_FULL_IMAGE:figures/full_fig_p030_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Close loop performance As shown in [PITH_FULL_IMAGE:figures/full_fig_p031_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Dynamic simulation for the SOC scheme (for every 100 h, the system is operated [PITH_FULL_IMAGE:figures/full_fig_p039_5.png] view at source ↗
read the original abstract

Self-optimizing control (SOC) aims to maintain near-optimal process operation by judiciously selecting controlled variables (CVs). In this series of work, the generalized global SOC (g2SOC) approach is proposed, which extends the concept of SOC to the whole operation space and uses general nonlinear functions to design CVs instead of linear combinations. In the first part of this series work, two numerical approaches for g2SOC are proposed: the optimization-based approach and the regression-based approach, based on a theoretical analysis of the existence of perfect self-optimizing CVs. The CVs designed by the former perform better, but are usually infeasible for large-scale problems. In this paper, we propose an algorithm called objective-guided controlled variable learning (OGCVL) that combines the advantages of both and has a better scalability. OGCVL is proposed for efficient CV design that seamlessly integrates symbolic and numerical computation techniques. Finally, the effectiveness of the OGCVL method is verified in two numerical examples. Both examples illustrate show that the OGCVL method is able to achieve good results while maintaining computational efficiency and is also feasible in large-scale problems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper proposes the objective-guided controlled variable learning (OGCVL) algorithm for generalized global self-optimizing control (g2SOC) of chemical processes. Building on the existence result for perfect self-optimizing controlled variables from Part I, OGCVL integrates symbolic and numerical computation techniques to design nonlinear CVs. It is positioned as combining the performance advantages of the optimization-based approach with the scalability of the regression-based approach. Effectiveness is verified on two numerical examples, which are stated to demonstrate good results, maintained computational efficiency, and feasibility for large-scale problems.

Significance. If the OGCVL method reliably delivers near-optimal performance with improved scalability and no hidden accuracy or computational trade-offs in high-dimensional nonlinear systems, it would advance practical self-optimizing control by enabling its use in large-scale chemical processes where purely optimization-based methods become intractable. The seamless symbolic-numerical integration, if shown to be robust, represents a useful algorithmic contribution for bridging performance and efficiency in SOC design.

major comments (1)
  1. [§5] §5 (Numerical Examples): the claim that OGCVL 'is also feasible in large-scale problems' rests on verification in two numerical examples, but no problem dimensions, timing or memory scaling data, or head-to-head comparisons against the Part I optimization-based method on larger instances are reported. This makes the large-scale feasibility assertion an extrapolation rather than a demonstrated result.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'Both examples illustrate show that' contains a duplicated word.
  2. [Abstract] The abstract refers to 'this series of work' and 'this paper' without an explicit citation to Part I in the opening paragraph; adding a clear forward reference would improve readability.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address the major comment point by point below and agree that additional quantitative details will strengthen the presentation of the numerical results.

read point-by-point responses

  1. Referee: [§5] §5 (Numerical Examples): the claim that OGCVL 'is also feasible in large-scale problems' rests on verification in two numerical examples, but no problem dimensions, timing or memory scaling data, or head-to-head comparisons against the Part I optimization-based method on larger instances are reported. This makes the large-scale feasibility assertion an extrapolation rather than a demonstrated result.

    Authors: We agree that the manuscript would be improved by providing explicit problem dimensions, timing data, and memory usage for the two examples. In the revised manuscript we will add a table in §5 that reports the number of decision variables, equality/inequality constraints, and measured CPU times and peak memory for OGCVL on both examples. We will also include a brief comparison of OGCVL versus the optimization-based method of Part I on the smaller example where the latter remains tractable. Direct head-to-head timing comparisons on instances larger than those solvable by the Part I method are not possible, because that method becomes computationally intractable precisely in the regime where OGCVL is intended to be used; we will state this limitation explicitly rather than implying a full scaling study has been performed. revision: yes

Circularity Check

1 steps flagged

Minor self-citation to Part I existence result; OGCVL algorithm is a distinct new method with no reduction to prior fits

specific steps
  1. self citation load bearing [Abstract]
    "In the first part of this series work, two numerical approaches for g2SOC are proposed: the optimization-based approach and the regression-based approach, based on a theoretical analysis of the existence of perfect self-optimizing CVs. [...] we propose an algorithm called objective-guided controlled variable learning (OGCVL) that combines the advantages of both and has a better scalability. OGCVL is proposed for efficient CV design that seamlessly integrates symbolic and numerical computation techniques."

    The new OGCVL method is explicitly positioned as building on the existence result from the authors' own prior Part I paper. While this reference is minor and the algorithm itself is presented as distinct, it qualifies as a self-citation that underpins the claimed theoretical grounding without independent verification in the current work.

full rationale

The paper's core contribution is the proposal of the OGCVL algorithm, which integrates symbolic and numerical techniques to design CVs and is verified through two numerical examples. This is independent of the Part I approaches. The only self-reference is to the theoretical existence of perfect self-optimizing CVs from Part I, which serves as background justification rather than a load-bearing reduction of the new algorithm or its performance claims to prior fitted quantities. No self-definitional loops, fitted inputs renamed as predictions, or ansatz smuggling appear in the described derivation. The scalability assertion is an empirical extrapolation from examples and does not constitute circularity in the mathematical sense.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the existence of perfect self-optimizing CVs established in part I and on the unproven assumption that symbolic-numerical integration will deliver both accuracy and scalability for arbitrary nonlinear process models.

axioms (1)
  • domain assumption Existence of perfect self-optimizing controlled variables for the generalized global SOC problem
    Invoked from the theoretical analysis in part I of the series work.

pith-pipeline@v0.9.0 · 5532 in / 1243 out tokens · 46267 ms · 2026-05-12T01:20:36.558012+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

Reference graph

Works this paper leans on

229 extracted references · 229 canonical work pages

  1. [1]

    2001 , month = jan, edition =

    Optimization of Chemical Processes , author =. 2001 , month = jan, edition =

    work page 2001

  2. [2]

    2009 , journal =

    Adaptation Strategies for Real-Time Optimization , author =. 2009 , journal =

    work page 2009

  3. [3]

    2019 , journal =

    110th Anniversary: A Feature-Based Analysis of Static Real-Time Optimization Schemes , author =. 2019 , journal =

    work page 2019

  4. [4]

    Process Decomposition and the Classification of the Control Tasks

    Studies in the Synthesis of Control Structures for Chemical Processes: Part I: Formulation of the Problem. Process Decomposition and the Classification of the Control Tasks. Analysis of the Optimizing Control Structures , author =. 1980 , journal =

    work page 1980

  5. [5]

    2022 , month = may, journal =

    Real-Time Optimization as a Feedback Control Problem -- A Review , author =. 2022 , month = may, journal =. doi:10.1016/J.COMPCHEMENG.2022.107723 , urldate =

    work page doi:10.1016/j.compchemeng.2022.107723 2022

  6. [6]

    2003 , publisher =

    Real Time Optimization by Extremum Seeking Control , author =. 2003 , publisher =

    work page 2003

  7. [7]

    2005 , journal =

    Use of Measurements for Enforcing the Necessary Conditions of Optimality in the Presence of Constraints and Uncertainty , author =. 2005 , journal =

    work page 2005

  8. [8]

    2000 , journal =

    Plantwide Control: The Search for the Self-Optimizing Control Structure , author =. 2000 , journal =

    work page 2000

  9. [9]

    2024 , month = mar, journal =

    Global Self-Optimizing Control of Batch Processes , author =. 2024 , month = mar, journal =. doi:10.1016/j.jprocont.2024.103163 , urldate =

    work page doi:10.1016/j.jprocont.2024.103163 2024

  10. [10]

    2024 , month = jun, journal =

    Dynamic Controlled Variables Based Dynamic Self-Optimizing Control , author =. 2024 , month = jun, journal =. doi:10.1016/j.jprocont.2024.103228 , urldate =

    work page doi:10.1016/j.jprocont.2024.103228 2024

  11. [11]

    Self-Optimizing Control: The Missing Link between Steady-State Optimization and Control , shorttitle =

    Skogestad, Sigurd , year =. Self-Optimizing Control: The Missing Link between Steady-State Optimization and Control , shorttitle =. Computers & Chemical Engineering , volume =. doi:10.1016/S0098-1354(00)00405-1 , urldate =

    work page doi:10.1016/s0098-1354(00)00405-1

  12. [12]

    2023 , journal =

    Lagrange-Based Global Self-Optimizing Control for Constraint Activeness Varying Processes , author =. 2023 , journal =

    work page 2023

  13. [13]

    The Existence of Perfect Controlled Variables and Numerical Design Methods , author =

    Generalized Global Self-Optimizing Control for Chemical Processes Part I. The Existence of Perfect Controlled Variables and Numerical Design Methods , author =. 2023 , month = sep, journal =. doi:10.1021/acs.iecr.3c01685 , urldate =

    work page doi:10.1021/acs.iecr.3c01685 2023

  14. [14]

    Industrial & Engineering Chemistry Research , shortjournal =

    Real-Time Optimization of Gold Cyanidation Leaching Process in a Two-Layer Control Architecture Integrating Self-Optimizing Control and Modifier Adaptation , author =. Industrial & Engineering Chemistry Research , shortjournal =. 2017 , date =. doi:10.1021/acs.iecr.6b04801 , url =

    work page doi:10.1021/acs.iecr.6b04801 2017

  15. [15]

    2022 , month = dec, journal =

    Using Stochastic Programming to Train Neural Network Approximation of Nonlinear MPC Laws , author =. 2022 , month = dec, journal =. doi:10.1016/j.automatica.2022.110665 , urldate =

    work page doi:10.1016/j.automatica.2022.110665 2022

  16. [16]

    2024 , journal =

    Performance Guaranteed MPC Policy Approximation via Cost Guided Learning , author =. 2024 , journal =. doi:10.1109/LCSYS.2024.3379115 , urldate =

    work page doi:10.1109/lcsys.2024.3379115 2024

  17. [17]

    2016 , month = nov, publisher =

    Deep Learning , author =. 2016 , month = nov, publisher =

    work page 2016

  18. [19]

    Computer vision , langid =

    Physics-Informed Deep Learning , author =. Computer vision , langid =

    work page

  19. [20]

    2023 , month = jan, journal =

    Optimal Control of PDEs Using Physics-Informed Neural Networks , author =. 2023 , month = jan, journal =. doi:10.1016/j.jcp.2022.111731 , urldate =

    work page doi:10.1016/j.jcp.2022.111731 2023

  20. [21]

    Raissi, P

    Physics-Informed Neural Networks: A Deep Learning Framework for Solving Forward and Inverse Problems Involving Nonlinear Partial Differential Equations , author =. 2019 , month = feb, journal =. doi:10.1016/J.JCP.2018.10.045 , urldate =

    work page doi:10.1016/j.jcp.2018.10.045 2019

  21. [22]

    Andersson, Joel A. E. and Gillis, Joris and Horn, Greg and Rawlings, James B. and Diehl, Moritz , year =. CasADi: A Software Framework for Nonlinear Optimization and Optimal Control , shorttitle =. Mathematical Programming Computation , volume =. doi:10.1007/s12532-018-0139-4 , urldate =

    work page doi:10.1007/s12532-018-0139-4

  22. [23]

    and Rockafellar, R

    Dontchev, Asen L. and Rockafellar, R. Tyrrell , year =. Implicit Functions and Solution Mappings: A View from Variational Analysis , shorttitle =. doi:10.1007/978-1-4939-1037-3 , urldate =

    work page doi:10.1007/978-1-4939-1037-3

  23. [24]

    Chemical Engineering Research and Design , year =

    Alabi, SI and Morris, AJ and Martin, EB , title =. Chemical Engineering Research and Design , year =

    work page

  24. [25]

    , title =

    Alstad, V. , title =

    work page

  25. [26]

    2007 , month = jan, journal =

    Null Space Method for Selecting Optimal Measurement Combinations as Controlled Variables , author =. 2007 , month = jan, journal =. doi:10.1021/ie060285+ , urldate =

    work page doi:10.1021/ie060285 2007

  26. [27]

    Alstad and S

    V. Alstad and S. Skogestad and E. S. Hori , title =. J. Process Control , year =

    work page

  27. [28]

    , title =

    Andersson, J. , title =. 2013 , owner =

    work page 2013

  28. [29]

    Araujo, A. C. B. D. and Hori, E. S. and Skogestad, S. , title =. Ind. Eng. Chem. Res. , year =

    work page

  29. [30]

    2003 , author =

    Real-time optimization by extremum-seeking control , publisher =. 2003 , author =

    work page 2003

  30. [31]

    AIChE journal , year =

    Baldea, Michael and Araujo, Antonio and Skogestad, Sigurd and Daoutidis, Prodromos , title =. AIChE journal , year =

    work page

  31. [32]

    Industrial & engineering chemistry research , year =

    Basak, Kaushik and Abhilash, KS and Ganguly, Saibal and Saraf, DN , title =. Industrial & engineering chemistry research , year =

    work page

  32. [33]

    2013 , author =

    Nonlinear programming: theory and algorithms , publisher =. 2013 , author =

    work page 2013

  33. [34]

    Biegler, L. T. , title =. Chemical Engineering and Processing: Process Intensification , year =

    work page

  34. [35]

    2009 , month = mar, journal =

    Large-Scale Nonlinear Programming Using IPOPT: An Integrating Framework for Enterprise-Wide Dynamic Optimization , author =. 2009 , month = mar, journal =. doi:10.1016/J.COMPCHEMENG.2008.08.006 , urldate =

    work page doi:10.1016/j.compchemeng.2008.08.006 2009

  35. [36]

    1995 , author =

    Neural networks for pattern recognition , publisher =. 1995 , author =

    work page 1995

  36. [37]

    and Tsitsiklis, J

    Blondel, V. and Tsitsiklis, J. N. , title =. SIAM Journal on Control and Optimization , year =

    work page

  37. [38]

    IEEE Control Systems Magazine , year =

    Bonvin, Dominique and Srinivasan, Bala and Hunkeler, David , title =. IEEE Control Systems Magazine , year =

    work page

  38. [39]

    2005 , author =

    Iterative algorithms for multilayer optimizing control , publisher =. 2005 , author =

    work page 2005

  39. [40]

    1975 , author =

    Applied optimal control: optimization, estimation and control , publisher =. 1975 , author =

    work page 1975

  40. [41]

    , title =

    Cao, Y. , title =. Int. J. Autom. Comput. , year =

    work page

  41. [42]

    Cao and V

    Y. Cao and V. Kariwala , title =. 2009 , note =

    work page 2009

  42. [43]

    Cao and V

    Y. Cao and V. Kariwala , title =. Comput. Chem. Engng. , year =

    work page

  43. [44]

    and Rossiter, D

    Cao, Y. and Rossiter, D. and Owens, D. H. , title =. Proc. IFAC symposium on. 1998 , pages =

    work page 1998

  44. [45]

    Cao and D

    Y. Cao and D. Rossiter and D. H. Owens , title =. Transaction of IChemE, Part A , year =

    work page

  45. [46]

    and Saha, P

    Cao, Y. and Saha, P. , title =. Chem. Engg. Sci. , year =

    work page

  46. [47]

    Journal of Process Control , year =

    Cao, Zhixing and Lu, Jingyi and Zhang, Ridong and Gao, Furong , title =. Journal of Process Control , year =

    work page

  47. [48]

    and Marchetti, A

    Chachuat, B. and Marchetti, A. and Bonvin, D. , title =. J. Proc. Control , year =

    work page

  48. [49]

    and Srinivasan, B

    Chachuat, B. and Srinivasan, B. and Bonvin, D. , title =. Comput. Chem. Eng. , year =

    work page

  49. [50]

    Industrial & engineering chemistry research , year =

    Chen, Chun Y and Joseph, Babu , title =. Industrial & engineering chemistry research , year =

    work page

  50. [51]

    Chen and U

    Q. Chen and U. Kruger and M. Meronk and A. Y. T. Leung , title =. Control Engineering Practice , year =

    work page

  51. [52]

    Chen and J

    T. Chen and J. Morris and E. Martin , title =. Journal of the Royal Statistical Society C (Applied Statistics) , year =

    work page

  52. [53]

    and Sun, Y

    Chen, T. and Sun, Y. , title =. Control Engineering Practice , year =

    work page

  53. [54]

    , title =

    Chen, X.-W. , title =. Pattern Recognition Letters , year =

    work page

  54. [55]

    Cheng and Chiu, Min-Sen , title =

    C. Cheng and Chiu, Min-Sen , title =. Chem. Eng. Sci. , year =

    work page

  55. [56]

    , publisher =

    Fault Detection and Diagnosis in Industrial Systems. , publisher =. 2001 , author =

    work page 2001

  56. [57]

    S. W. Choi and E. B. Martin and A. J. Morris , title =. Industrial and Engineering Chemistry Research , year =

    work page

  57. [58]

    S. W. Choi and J. H. Park and I.-B. Lee , title =. Computers and Chemical Engineering , year =

    work page

  58. [59]

    2009 , author =

    Introduction to algorithms , publisher =. 2009 , author =

    work page 2009

  59. [60]

    A directional modifier-adaptation algorithm for real-time optimization , journal =

    Costello, Sean and Fran. A directional modifier-adaptation algorithm for real-time optimization , journal =. 2016 , volume =

    work page 2016

  60. [61]

    Mathematics of control, signals and systems , year =

    Cybenko, George , title =. Mathematics of control, signals and systems , year =

    work page

  61. [62]

    and Narasimhan, S

    Dahl-Olsen, H. and Narasimhan, S. and Skogestad, S. , title =. Proc. of American Control Conference , year =

    work page

  62. [63]

    Deshpande, Saurabh Anand , title =

    work page

  63. [64]

    SIAM Journal on Control and Optimization , year =

    Deshpande, Saurabh A and Bonvin, Dominique and Chachuat, Benoit , title =. SIAM Journal on Control and Optimization , year =

    work page

  64. [65]

    IFAC Proceedings Volumes , year =

    Deshpande, Saurabh and Srinivasan, Bala and Bonvin, Dominique , title =. IFAC Proceedings Volumes , year =

    work page

  65. [66]

    AIChE Journal , year =

    Diwekar, Urmila M and Kalagnanam, Jayant R , title =. AIChE Journal , year =

    work page

  66. [67]

    Downs, J. J. and Skogestad, S. , title =. Ann. Rev. Control , year =

    work page

  67. [68]

    J. J. Downs and E. F. Vogel , title =. Comput. Chem. Eng. , year =

    work page

  68. [69]

    2001 , author =

    Optimization of Chemical Processes , publisher =. 2001 , author =

    work page 2001

  69. [70]

    Engell, Sebastian , title =. J. Process Control , year =

    work page

  70. [71]

    Proceedings of the 19th IFAC World Congress , year =

    Faulwasser, Timm and Bonvin, Dominique , title =. Proceedings of the 19th IFAC World Congress , year =

    work page

  71. [72]

    1983 , publisher =

    Fiacco, Anthony V , title =. 1983 , publisher =

    work page 1983

  72. [73]

    2013 , publisher =

    Fletcher, Roger , title =. 2013 , publisher =

    work page 2013

  73. [74]

    1995 , author =

    Nonlinear and Mixed-Integer Optimization , publisher =. 1995 , author =

    work page 1995

  74. [75]

    Francois and D

    G. Francois and D. Bonvin , title =. Ind. Eng. Chem. Res. , year =

    work page

  75. [76]

    Francois and B

    G. Francois and B. Srinivasan and D. Bonvin , title =. Ind. Eng. Chem. Res. , year =

    work page

  76. [77]

    Francois and B

    G. Francois and B. Srinivasan and D. Bonvin , title =. J. Eur. Syst. Autom. , year =. doi:10.3166/JESA.46.291-305 , keywords =

    work page doi:10.3166/jesa.46.291-305

  77. [78]

    and Srinivasan, B

    Francois, G. and Srinivasan, B. and Bonvin, D. , title =. J. Process Control , year =

    work page

  78. [79]

    Use of convex model approximations for real-time optimization via modifier adaptation , journal =

    Fran. Use of convex model approximations for real-time optimization via modifier adaptation , journal =. 2013 , volume =

    work page 2013

  79. [80]

    Comparison of six implicit real-time optimization schemes , journal =

    Fran. Comparison of six implicit real-time optimization schemes , journal =. 2012 , volume =

    work page 2012

  80. [81]

    and Kano, M

    Fujiwara, K. and Kano, M. and Hasebe, S. and Takinami, A. , title =. AIChE J. , year =

    work page

Showing first 80 references.