Journal Design Engineering Masthead
African Civil Engineering Journal | 12 May 2025

Randomised Field Trial of Process-Control System Methodologies for Yield Optimisation in South African Industrial Plants

P, i, e, t, e, r, v, a, n, d, e, r, M, e, r, w, e, ,, T, h, a, n, d, i, w, e, N, k, o, s, i, ,, K, a, g, i, s, o, N, a, i, d, o, o
process-control systemsyield optimisationrandomised field trialindustrial plants
Multi-site randomised field trial across 27 industrial plants in South Africa
Model predictive control yielded 7.3% mean increase versus control group
Statistical process control showed 3.1% improvement, expert system 1.8%
Linear mixed-effects model with robust standard errors clustered at plant level

Abstract

{ "background": "Industrial process-control systems are critical for operational efficiency, yet there is a paucity of rigorous field evidence from the region comparing the efficacy of different methodological approaches for yield optimisation.", "purpose and objectives": "This study aimed to empirically evaluate and compare the yield improvement performance of three distinct process-control system methodologies—model predictive control, statistical process control, and a rule-based expert system—within operational industrial plants.", "methodology": "A multi-site randomised field trial was conducted across 27 plants in the chemical and manufacturing sectors. Plants were randomly assigned to one of the three intervention methodologies or a control group. Yield was measured as mass output per unit of raw material input over a standardised period. The primary analysis used a linear mixed-effects model: $Y{ij} = \\mu + \\alphai + \\beta X{ij} + \\epsilon{ij}$, where $Y_{ij}$ is the yield for plant $j$ in sector $i$, with robust standard errors clustered at the plant level.", "findings": "The model predictive control system generated a statistically significant mean yield increase of 7.3% (95% CI: 5.1% to 9.5%) compared to the control group, outperforming both the statistical process control (3.1% increase) and rule-based expert system (1.8% increase) methodologies.", "conclusion": "The trial demonstrates that advanced model-based control methodologies can deliver substantial yield gains in real-world industrial settings, whereas simpler systems offer more modest improvements.", "recommendations": "Plant managers should prioritise investment in model predictive control systems where feasible, given its superior performance. Further research should investigate the cost-benefit analysis of implementation across different industrial scales.", "key words": "process control, randomised controlled trial, yield optimisation, industrial engineering, manufacturing", "contribution statement": "This paper provides the first comparative, experimental evidence from a randomised field trial on the effectiveness of different process-control methodologies for yield improvement in an