Date of Award

2026

Keywords

artificial intelligence, BOD, chemometric modelling, essential oil, essential oil yield, Indian sandalwood, industrial wastewater treatment, instant controlled pressure drop (DIC/CID), machine learning, near-infrared spectroscopy, portable spectroscopy, predictive modelling & simulations, process intensification, sandalwood processing, Santalum album, soft sensor, steam vacuum decompression, wood classification

Document Type

Thesis

Publisher

Edith Cowan University

Degree Name

Doctor of Philosophy

School

School of Engineering

First Supervisor

Muhammad Rizwan Azhar ORCID iD 0000-0002-5938-282X

Second Supervisor

Hussein A. Mohammed ORCID iD 0000-0002-8730-3674

Abstract

Indian sandalwood (Santalum album L.) is a high-value aromatic tree native to South Asia and now cultivated commercially in tropical regions, including northern Australia. Its principal marketable product is a heartwood essential oil rich in the sesquiterpene alcohols alpha- and beta-santalol, which underpin its use in perfumery, cosmetics, pharmaceuticals, aromatherapy and specialty products. The sandalwood industry is constrained by a convergence of process, analytical and environmental limitations: long extraction cycles, diffusion-limited mass transfer through oil-bearing heartwood, destructive laboratory-based assays, delayed quality decisions and slow wastewater feedback. This thesis addresses these limitations as a single integrated process systems problem rather than as separate extraction, spectroscopy and wastewater studies. It develops a digital-physical framework for designing and optimising Instantaneous Controlled Pressure Drop (CID) technology for Santalum album L. essential oil extraction, while embedding near-infrared spectroscopy (NIRS), machine learning and water resource recovery facility (WRRF) soft sensing into the same industrial decision architecture.

The research first establishes the scientific and industrial basis for intensifying sandalwood oil extraction. Conventional hydrodistillation (HD) and steam distillation (SD) are shown to be inherently constrained by slow heat and mass transfer through a lignocellulosic matrix in which valuable sesquiterpene alcohols are structurally trapped in heartwood tissues. CID is investigated as a thermo-hydro-mechanical alternative in which short saturated-steam exposure is followed by rapid decompression to deep vacuum. This pressure discontinuity induces autovaporisation, internal tensile stress, pore opening and oil-cell rupture, thereby shifting extraction from diffusion-limited distillation to permeability-controlled release.

At pilot scale, the thesis maps CID operation across cycle number, treatment pressure, condensation strategy, wood particle size, pressure-drop intensity and derived severity descriptors. The experimental results identify a bounded operating window, rather than a maximum-severity strategy, as the basis of high-quality extraction. Within this window, CID achieved substantially higher oil recovery than 24 h HD and SD benchmarks in less than one hour, while maintaining an acceptable compositional profile. Multi-scale optical and scanning electron microscopy confirmed the mechanism by revealing fibre delamination, open lumens, exposed resin canals, and ruptured oil-bearing structures after cyclic decompression. Reactor-level analysis further demonstrated that condensation rate, vacuum stability, expansion volume and volume compensation are scale-up variables that directly control extraction severity and repeatability.

The second part of the thesis addresses the information latency that prevents intensified extraction from becoming an industrially controllable process. A unified NIRS-artificial intelligence framework was developed across solid, powdered, liquid and multiphase sandalwood matrices. Benchtop and portable spectrometers were used to predict essential oil yield, alpha- and beta-santalol content, heartwood sapwood ratio, oil moisture, ethanol extract yield and CID-derived emulsion yield. The models showed strong predictive performance and external-validation capability, demonstrating that NIRS can encode both chemical and structural variability across heterogeneous sandalwood materials. These models provide the feedstock, quality, and endpoint intelligence needed to control the experimentally identified CID operating window.

The third part of the thesis extends the same digital logic to environmental process control. Because CID, HD and SD all generate organic-rich condensate, separator drainage, emulsion water and wash-down streams, the effluent from extraction becomes part of the process design problem. A machine-learning soft sensor was developed to rapidly predict the final effluent biological oxygen demand (BOD) at an industrial WRRF treating essential oil wastewater. By linking process and laboratory variables to BOD outcomes, the soft sensor reduces reliance on five-day laboratory testing and enables earlier operational response to organic-load shocks, pH excursions, hydraulic changes and treatment instability.

The final integration is delivered through a techno-economic and sustainability assessment comparing CID, HD and SD on the functional unit of 1 kg of market-ready sandalwood oil. The assessment links extraction physics and monitoring outputs to mass, energy, greenhouse gas, water, wastewater, capital and operating cost inventories. The analysis shows that CID's economic and environmental value is structural: higher recovery, shorter residence time, reduced thermal duty and lower wastewater burden are achieved simultaneously when the process remains inside the validated severity window. Sensitivity and uncertainty analyses identify yield stability, installed capital, operating days, condensation reliability and WRRF burden as the most important scale-up validation variables.

Collectively, the outcomes of this thesis are directly aligned with industrial benefit across the full sandalwood processing value chain. In extraction, the pilot-scale CID framework provides a practical pathway to reduce processing time, improve oil recovery and define controllable operating windows for industrial deployment. In quality assurance, the NIRS-AI models enable rapid, non-destructive prediction of feedstock properties, oil quality and process endpoints, reducing dependence on delayed laboratory assays and supporting faster production decisions. In environmental management, the WRRF soft-sensor framework provides early prediction of BOD and supports timely operational response to wastewater variability generated during essential oil processing. Finally, the integrated techno-economic and sustainability assessment translates these technical advances into scale-up-relevant evidence on cost, energy use, greenhouse-gas emissions, water demand and wastewater burden. Although the operating values are sandalwood-specific, the methodological architecture - mechanism-based process intensification, rapid spectroscopy, soft sensing, risk assessment and TEA/LCA - is transferable to other essential-oil, natural-product and bioresource-processing systems in which product quality, utilities and wastewater load are tightly coupled.

Access Note

Access to this thesis is embargoed until 9th October 2027

Download

Available for download on Saturday, October 09, 2027

Share

 
COinS
 

Link to publisher version (DOI)

10.25958/1swv-ff60