Relevant Thesis-Based Degree Programs
Graduate Student Supervision
Doctoral Student Supervision
Dissertations completed in 2010 or later are listed below. Please note that there is a 6-12 month delay to add the latest dissertations.
Sensor-based sorting of waste wood composites and the use of composite particles in cement-bonded wood composites (2024)
Sustainable wood manufacturing practices increasingly include the recycling and reuse of waste wood. To avoid cross-contamination of potentially harmful treatments found in the wood industry, sorting is necessary. Visual sorting approaches were assessed, and it was found that they cannot reliably sort waste wood without exceeding existing heavy metal limits. Consequently, near-infrared hyperspectral imaging was used to identify wood waste type in three scenarios: (1) distinguishing new wood composites, (2) differentiating composites obtained from a landfill, and (3) distinguishing waste wood composites in a simulated waste wood stream. A partial least square discriminant analysis was employed, and it was found that accuracies reduced with each scenario. Nonetheless, all models were able to accurately predict at least 80% of the samples. To assess recyclability of wood composites (plywood, ACQ treated plywood, oriented strand board, particleboard, and melamine coated particleboard) in cement-bonded wood composites, hydration tests were conducted. The exothermic curing reaction of cement is inhibited when mixed with wood particles and the temperature reduction can be used to assess compatibility of wood with cement. These hydration tests found all wood composite types compatible with cement. However, the temperature curve of melamine coated and uncoated particleboard particles was missing a second temperature peak. Preliminary boards were manufactured and found to be very brittle, so it was concluded that the absence of a second temperature peak inhibits compatibility. Hydration tests should only be used to assess initial compatibility when a second peak is present and not be used to predict overall properties. Plywood, treated plywood and oriented strand board were subsequently used to manufacture cement-bonded wood composites. Their effect on density, modulus of elasticity (MOE), modulus of rupture (MOR), thickness swelling, water absorption and thermal conductivity on cement-bonded wood composites was evaluated along with that of particle size fraction and the use of an accelerator. The MOE and MOR values were smaller than expected due to a poor interface bond between the recycled wood particles and cement. However, the thermal conductivity of the cement-bonded panels met the minimum requirements. It is recommended that future studies explore how the interface bond could be improved.
View record
Wood sawing monitoring: sensory and artificial intelligence approaches (2020)
This thesis aimed to develop expert models for intelligent monitoring of the circular sawing process. Circular sawing experiments were conducted under different cutting conditions in kiln-dried, green, and frozen wood to study cutting power and waviness. The effect of the cutting factors and wood conditions on the response variables were reported and discussed. In parallel, the process was monitored using sound, acoustic emission (AE), and vibration sensors. A new wavelet-based methodology was developed to enable sound signal monitoring in very noisy environments by identifying and conserving the sound components corresponding to the sawing process. Emphasis was then put on sensory feature selection of AE signals in time and frequency domains. Feature selection was optimised by connecting the decision-making model with a heuristic optimisation algorithm to maximise the monitoring performance. Accordingly, particle swarm optimisation was linked with a neuro-fuzzy model. To eliminate the need for sophisticated signal processing, an automatic feature selection process was studied using the vibration signals. For this purpose, segments of the signals were fed into a self-organizing map model combined with the neuro-fuzzy and multilayer perceptron neural network models. The results showed that sawing frozen wood requires more cutting power than dry and green wood. However, freezing conditions lowered the waviness indicating a reduced sawing deviation. It was shown that the proposed wavelet-based approach for sound signal monitoring could be effective in noisy applications. Optimal feature selection could increase the monitoring accuracy when using an AE sensor. Interestingly, the automatic feature selection resulted in the highest accuracy indicating that combining the self-organizing map with an intelligent decision-making model could be used in sawmilling applications using a vibration sensor. In general, the results of this study proposed expert models for online monitoring of cutting power and waviness, which is of great importance in transitioning towards smart lumber manufacturing.
View record
Master's Student Supervision
Theses completed in 2010 or later are listed below. Please note that there is a 6-12 month delay to add the latest theses.
Intelligent monitoring of cutting power in circular sawing (2025)
This thesis investigated cutting power in circular sawing with two primary objectives: analyzing the effects of five main factors—wood species (Douglas-fir and Western maple), wood condition, rake angle, feed speed, and tool wear level—on cutting power and developing predictive models using sensor data. Exploratory data analysis (EDA) and regression modeling identified tool wear level and rake angle as the most impactful factors on cutting power, with significant interaction effects observed between feed speed and tool wear, feed speed and wood condition, tool wear and rake angle, and tool wear and wood condition. These findings provide valuable insights for optimizing the sawing process.To improve prediction accuracy, sensor-based models were developed using accelerometer, microphone, and acoustic emission (AE) sensors. Sensor data was preprocessed with discrete wavelet denoising to remove noise and enhance signal clarity. Subsequently, multiple time-domain and frequency-domain features were extracted for each sensor, with the most relevant selected using LASSO. Four sensor-based models—multiple linear regression (MLR), random forest (RF), fully connected neural network (FCNN), and 1D convolutional neural network (CNN)—were developed for predicting cutting power. Among the models, RF achieved the best performance (R²test = 0.910), surpassing both deep learning and linear models. Based on the feature importance analysis, microphone data emerged as the most predictive of cutting power among all three sensors, emphasizing its potential for real-time monitoring. In general, this study highlighted the value of sensor data integration to enhance cutting process efficiency and established a foundation for future research in real-time monitoring and predictive maintenance.
View record
Investigation of the effect of tool wear and wood species on surface quality and performance of finger-joint (2025)
Finger-jointed (FJ) lamellas are widely used in mass timber where bonding performance is critical for structural reliability. However, tool wear during FJ can affect wood surface characteristics, potentially affecting adhesive behavior and joint strength. In this study, Coastal Douglas-fir (DF), Hem-Fir (HF), and Western Redcedar (WRC) were machined using freshly sharpened and worn cutting tools and bonded with a polyurethane (PUR) adhesive to evaluate changes in surface topography, surface energy, adhesive wetting, ultimate tensile strength (UTS), and initial friction properties.The findings indicate that worn tool increased surface roughness (SA, SP, SV) for all wood species, which improved initial friction strength (IF-UTS) and coefficient of friction (IF-CoF), enhancing pre-cured joint integrity and stability. For DF, tool wear also increased UTS due to improved mechanical interlocking and adhesive penetration. In contrast, HF and WRC showed decreased UTS under worn tool conditions, despite increased roughness, IF-UTS, and IF-CoF, primarily due to reduced adhesive wettability and penetration. Glueline delamination remained low and was not significantly affected by wood species or tool wear, demonstrating the effectiveness of the PUR adhesive in accommodating surface variability.Overall, all FJ lamellas exceeded industry standards for UTS and failure modes, confirming reliable bonding performance. This study highlights the importance of balancing tool wear and wood species selection in FJ manufacturing. Moderate tool wear may be beneficial for DF, while maintaining sharp tools appears critical for HF and WRC. The findings provide valuable insights for optimizing machining strategies and improving the quality and consistency of structural FJ products in the wood industry.
View record
Cutting power and dust emission in sawing kiln-dried, green and frozen douglas-fir and western maple wood (2024)
Sawing is a critical machining process in the wood manufacturing industry, transforming logs into usable lumber for various applications, including construction and furniture making. Despite significant research, optimizing the sawing process to balance productivity, cost efficiency, and environmental sustainability remains challenging. This thesis investigates the effects of five key cutting parameters (wood species, tool wear, wood condition, rake angle, and feed speed) on cutting power and dust emission during circular sawing of Douglas-fir and Western maple, with the overall aim to identify optimal parameter combinations to enhance sawmill operations. It was found that all cutting parameters significantly impact cutting power and dust emission. Notably, sawing Western maple and dry wood required higher cutting power and produced more dust compared to sawing Douglas-fir and green or frozen wood. Increasing feed speed resulted in higher cutting power but reduced dust emission, while a worn tool and a smaller rake angle increased both cutting power and dust emission.The results indicate that maximizing feed speed can improve productivity and reduce dust, but at the cost of higher cutting power. By offering insights into the complex interactions between cutting parameters, this research contributes to the development of more efficient, cost-effective, and environmentally sustainable sawing practices in the wood manufacturing industry.
View record
Achieving energy efficiency and quality control during industrial veneer drying with data-driven approaches (2022)
Veneer drying plays an essential role in manufacturing veneer-based composites and consumes a significant amount of energy from burning fossil fuels, wood residues and electricity. The soaring energy price, the growing carbon tax rate, and the substantial social-environmental concerns regarding fossil fuel use have urged veneer manufacturers to adapt and become more efficient in energy consumption. Therefore, the ultimate objective of this research was to provide solutions for the industrial veneer drying process to reduce energy use without encumbering product quality. Unlike the physics-based methods commonly seen in the literature, this research embraced a data-driven approach for its promising potential in modelling a dynamic, complicated, and interrelated process such as veneer drying. In partnership with a veneer manufacturer in British Columbia, industrial drying data and climatic information corresponding to a several-month period were extracted, processed, compiled, and cleaned in two formats for analyzing unit energy consumption and drying quality turnout. Both parametric and non-parametric algorithms were deployed to predict the unit gas and electricity usage of each dryer. Based on cross-validation evaluations, the random forest model with all explanatory variables slightly outperformed two linear models regarding almost all accuracy metrics, but linear models had the advantage of providing an easy-to-interpret solution. In parallel, logistic regression classifiers and a random forest classifier were developed to estimate the quality level of individual veneer sheets, but the classification ability of all models was limited. Nonetheless, the results suggested that a more refined sorting of the initial moisture content of veneers could reduce unit energy consumption and lessen the number of re-dries and over-dries, thus improving quality turnout. Besides, the relations between unit energy consumption and Zone 1a temperature or drying speed were not consistent among different types of veneer products or between dryers, which posed challenges in modifying the veneer drying schedule for energy savings. Finally, models predicting energy consumption and veneer quality were combined in a preliminary assessment to optimize the drying schedule for reducing energy use.
View record
An approach to assess the effect of improved lodgepole pine and white spruce on an Alberta sawmill and the impact on net present value (2022)
Optimization of log processing is important in maximizing wood resource utilization andfinancial benefits of sawmills, while genomic-assisted tree breeding (GATB) can accelerate theselection of trees in the tree improvement process. The main research objective was to assess thebenefits for sawmills to process improved planting stock generated through GATB as comparedto a traditional phenotype-based tree breeding (TB) method. In this thesis, the process from treeseedling to lumber manufacturing was simulated. The project was conducted in two parts. First, Iassessed the impact of the selection approach, growth parameters, including age of measurement(AM), rotation age (RA), and site index (SI) on four sawmill key performance indicators (KPI),for lodgepole pine (LP) and white spruce (WS). An individual tree growth and yield model (MGM18) was used to grow trees to harvest, while SAWSIM® was used to assign an end-productto each tree. In the second part, I analyzed the net present value (NPV) using the same parameters for various discount rates, log prices, lumber values and seedling costs. The results indicated that using data from older AM resulted in larger tree height and DBH and was due to the age-age correlation effects which were more important for LP. In practice, these results show that an underestimation is to be expected in all four KPIs when using a young AM (year 20-23). Furthermore, environmental factors such as SI had a significant impact on tree growth and lumbervolume, while RA had a positive effect on the final products value. The analysis from the lumbervalue data revealed that the value from GATB trees should be higher than that of TB trees. Thiswas reinforced by the NPV analysis. The results also revealed that the optimal economic RA (lower than 60 years) would be earlier than the age at which the maximum mean annual increment is reached. Results showed a positive NPV could be obtained when the lumber value is over 25%larger than in the base case scenario and the log price premium is not higher than 50%. Theseresults were consistent among the seven selection scenarios explored.
View record
Using x-ray scanned reconstructed logs to predict knot characteristics and log value (2021)
Amabilis fir and western hemlock are an important softwood lumber resource in British Columbia (BC). For several reasons, the available wood volume in BC has consistently decreased since 1990. This decrease suggests that sawmilling processes should become more efficient in utilizing wood. In lumber manufacturing, considering knot characteristics and distribution within lumber pieces in the optimisation of the primary breakdown patterns would significantly impact the product quality or grade. The goal of the project was therefore to investigate the impact of growth history on knot characteristics and how they, in turn, influence the manufacturing of lumber in Coastal and Interior sawmills and build predictive models. Computed tomography (CT) scanning can non-destructively detect knots in wood and is gaining acceptance in the wood industry. Seventy-two amabilis fir and western hemlock trees from three plots located on Vancouver Island, BC were scanned, and images were processed to extract knot characteristics and distribution to reconstruct three-dimensional (3D) log models. The effects of three diameter at breast height (DBH) classes (30, 40 and 50 cm) and three sites on knot characteristics, including knot volume, number of knots, average knot area on CT image sections and knot/tree volume ratio, were investigated. As expected, the knot characteristics of both species increased with the DBH. The difference of knot distribution between amabilis fir and western hemlock suggests that the latter is more sensitive to growth conditions of temperature, precipitation and sunlight. The 3D log models were then processed in Optitek to simulate the sawmill production and assess the impact of the DBH classes and sites (including knots) on the lumber and value recovery in Coastal and Interior sawmills under normal, optimistic, and pessimistic economic cycles. The sawmilling simulations revealed that the Coastal mill produced a lower lumber volume but a higher value due to the type of products manufactured and the primary breakdown patterns being used. The sawmilling simulation results were compared and used in predicting the value of standing amabilis fir and western hemlock trees. Models were developed based on the knot characteristic and tree features to predict the value of a standing tree.
View record
If this is your researcher profile you can log in to the Faculty & Staff portal to update your details and provide recruitment preferences.
Membership Status
Program Affiliations
Academic Unit(s)