Failure Detection in Chemical Processes Using Symbolic Machine Learning: A Case Study on Ethylene Oxidation

Julien Amblard; Niklas Groll; Matthew Tait; Mark Law; Gürkan Sin; Alessandra Russo

cs.LGMar 6, 2026

Failure Detection in Chemical Processes Using Symbolic Machine Learning: A Case Study on Ethylene Oxidation

Authors: Julien Amblard, Niklas Groll, Matthew Tait, Mark Law, Gürkan Sin, Alessandra Russo

Organizations: Department of Computing, Imperial College London, London, United Kingdom · Department of Chemical and Biochemical Engineering, Technical University of Denmark, Kongens Lyngby, Denmark · ILASP Limited, London, United Kingdom

Abstract

Over the past decade, Artificial Intelligence has significantly advanced, mostly driven by large-scale neural approaches. However, in the chemical process industry, where safety is critical, these methods are often unsuitable due to their brittleness, and lack of explainability and interpretability. Furthermore, open-source real-world datasets containing historical failures are scarce in this domain. In this paper, we investigate an approach for predicting failures in chemical processes using symbolic machine learning and conduct a feasibility study in the context of an ethylene oxidation process. Our method builds on a state-of-the-art symbolic machine learning system capable of learning predictive models in the form of probabilistic rules from context-dependent noisy examples. This system is a general-purpose symbolic learner, which makes our approach independent of any specific chemical process. To address the lack of real-world failure data, we conduct our feasibility study leveraging data generated from a chemical process simulator. Experimental results show that symbolic machine learning can outperform baseline methods such as random forest and multilayer perceptron, while preserving interpretability through the generation of compact, rule-based predictive models. Finally, we explain how such learned rule-based models could be integrated into agents to assist chemical plant operators in decision-making during potential failures.

Figures & tables

Figure 1: Simulation flowsheet of the EO process, with a legend to identify components

Panel of Table 1 — Table 1: The main process variables of the EO process we focus on in this paper — entries in the first column correspond to the steps from Figure 1 , and the units of the process variables are respectively bar, ºC and kg/s for pressure, temperature and flow rate

param	unit	default	min	max	instr
SRCR1.P	bar	2	1.6	1.98	uniform
SRCE1.P	bar	2	1.95	2.05	uniform
SRCR1.M[C2H4]	kmol	4464	4240	4360	uniform

Figure 2: A perturbation file for “low pressure at source”

	Learning parameters	Accuracy		Interpretability		Confidence
	Non-default assignments	$min(auc)$	$avg(auc)$	$avg(n_{body})$	$avg(n_{r}\|cl)$	$avg(prob)$
(1)	$exp:=trivial$	1	1	2.14	1.4	0.74
(2)	$exp:=nontrivial6$	0.82	0.92	3	1.33	0.55
(3)	$exp:=nontrivial10$	0.75	0.88	3	1.7	0.75
(4)	$exp:=nontrivial10;~n\_runs:=125$	0.7	0.85	3.11	1.8	0.73
(5)	$n\_runs:=10$	0.83	0.93	2.5	1	0.7

Table 2: A look at the impacts of adjusting the learning parameters for failure detection with task

T_{dynamic}

, using emphasized text to highlight the results that are preferred under each metric

Config.	SVM	MLP	RF	HGB	AB	Ours
(1)	0.95	0.95	0.97	0.97	0.97	1
(2)	0.85	0.83	0.87	0.89	0.87	0.92
(3)	0.84	0.77	0.85	0.84	0.85	0.88
(14)	0.87	0.86	0.97	0.97	0.97	0.94
(15)	0.85	0.83	0.87	0.89	0.87	0.92
(16)	0.83	0.82	0.89	0.89	0.87	0.9

Table 3: The AUC averaged over the selected experiments for different learning parameter configurations (as given in Table 2 ), comparing different baselines to our approach

Panel of Figure 3 — Figure 3: ROC curves generated from the default learning parameter configuration, for our approach and the baselines

Panel of Figure 4 — Figure 4: Rules generated by DisPLAS for $T_{static}$ ( $\texttt{low2}<\texttt{low1}<\texttt{normal}<\texttt{high1}<\texttt{high2}$ ) when including all nontrivial experiments and process variable measurements over 125 runs of the simulation

Panel of Figure 5 — Figure 5: Rules generated by DisPLAS for $T_{dynamic}$ , modifying only the learning parameter $t_{short\_term}$ — the predicate unchanged indicates process variables with equal initial and short-term values, while remaining predicates in rule bodies refer to measurements taken at $t_{short\_term}$ and (absolute) changes in measurement since $t=0$ (increase: $\nearrow$ , decrease: $\searrow$ )

Fault Detection

Fault detection identifies abnormal behavior in machines, sensors, infrastructure, software, and autonomous systems before it causes harm or disruption.

See all 102 papers

Explainability for Fault Detection System in Chemical Processes
Feb 18, 2026Georgios Gravanis, Dimitrios Kyriakou, Spyros Voutetakis +2Industrial Fault Diagnosis Fault Diagnosis
Department of Information and Electronic Engineering, International Hellenic University, Greece · Chemical Process and Energy Resources Institute, Centre for Research and Technology Hellas, Greece · Department of Industrial Engineering and Management, International Hellenic University, Greece
Atoms to Processes: The Role of Artificial Intelligence and Machine Learning in Chemical Engineering
Oct 1, 2026Michael Baldea, Linda J. Broadbelt, Marianthi G. Ierapetritou +17Scientific ML Physics-Informed ML
McKetta Department of Chemical Engineering and Oden Institute for Chemical Engineering and Sciences, The University of Texas at Austin, Austin, TX 78731, USA · Department of Chemical and Biological Engineering, Northwestern University, Evanston, IL 60208, USA · Department of Chemical and Biomolecular Engineering, University of Delaware, Newark, DE 19716, USA +13
Compliance-Aware Predictive Process Monitoring: A Neuro-Symbolic Approach
Mar 27, 2026Fabrizio De Santis, Gyunam Park, Wil M. P. van der Aalst +1Logic Tensor Networks Predictive Process Monitoring
University of Parma, Parma, Italy · Eindhoven University of Technology, Eindhoven, Netherlands · RWTH Aachen University, Aachen, Germany

Feb 18, 2026J/K move · Enter open · S save

Explainability for Fault Detection System in Chemical Processes

Georgios Gravanis, Dimitrios Kyriakou, Spyros Voutetakis, Simira Papadopoulou+1

Department of Information and Electronic Engineering, International Hellenic University, Greece · Chemical Process and Energy Resources Institute, Centre for Research and Technology Hellas, Greece · Department of Industrial Engineering and Management, International Hellenic University, Greece

Industrial Fault Diagnosis Fault Diagnosis Gradient-Based Attribution SHAP Feature Attribution

Open paper AskPDF arXiv

In this work, we apply and compare two state-of-the-art eXplainability Artificial Intelligence (XAI) methods, the Integrated Gradients (IG) and the SHapley Additive exPlanations (SHAP), that explain the fault diagnosis decisions of a highly accurate Long Short-Time Memory (LSTM) classifier. The classifier is trained to detect faults in a benchmark non-linear chemical process, the Tennessee Eastman Process (TEP). It is highlighted how XAI methods can help identify the subsystem of the process where the fault occurred. Using our knowledge of the process, we note that in most cases the same features are indicated as the most important for the decision, while insome cases the SHAP method seems to be more informative and closer to the root cause of the fault. Finally, since the used XAI methods are model-agnostic, the proposed approach is not limited to the specific process and can also be used in similar problems.

Abstract

Fault Detection

Explore similar work

Explainability for Fault Detection System in Chemical Processes

Atoms to Processes: The Role of Artificial Intelligence and Machine Learning in Chemical Engineering

Compliance-Aware Predictive Process Monitoring: A Neuro-Symbolic Approach