RESUMO
Near-infrared (NIR) spectroscopy is widely used as a nondestructive evaluation (NDE) tool for predicting wood properties. When deploying NIR models, one faces challenges in ensuring representative training data, which large datasets can mitigate but often at a significant cost. Machine learning and deep learning NIR models are at an even greater disadvantage because they typically require higher sample sizes for training. In this study, NIR spectra were collected to predict the modulus of elasticity (MOE) of southern pine lumber (training set = 573 samples, testing set = 145 samples). To account for the limited size of the training data, this study employed a generative adversarial network (GAN) to generate synthetic NIR spectra. The training dataset was fed into a GAN to generate 313, 573, and 1000 synthetic spectra. The original and enhanced datasets were used to train artificial neural networks (ANNs), convolutional neural networks (CNNs), and light gradient boosting machines (LGBMs) for MOE prediction. Overall, results showed that data augmentation using GAN improved the coefficient of determination (R2) by up to 7.02% and reduced the error of predictions by up to 4.29%. ANNs and CNNs benefited more from synthetic spectra than LGBMs, which only yielded slight improvement. All models showed optimal performance when 313 synthetic spectra were added to the original training data; further additions did not improve model performance because the quality of the datapoints generated by GAN beyond a certain threshold is poor, and one of the main reasons for this can be the size of the initial training data fed into the GAN. LGBMs showed superior performances than ANNs and CNNs on both the original and enhanced training datasets, which highlights the significance of selecting an appropriate machine learning or deep learning model for NIR spectral-data analysis. The results highlighted the positive impact of GAN on the predictive performance of models utilizing NIR spectroscopy as an NDE technique and monitoring tool for wood mechanical-property evaluation. Further studies should investigate the impact of the initial size of training data, the optimal number of generated synthetic spectra, and machine learning or deep learning models that could benefit more from data augmentation using GANs.
Assuntos
Análise de Dados , Madeira , Módulo de Elasticidade , Luz , Aprendizado de MáquinaRESUMO
Monitoring the moisture content (MC) of wood and avoiding large MC variation is a crucial task as a large moisture spread after drying significantly devalues the product, especially in species with high green MC spread. Therefore, this research aims to optimize kiln-drying and provides a predictive approach to estimate and classify target timber moisture, using a gradient-boosting machine learning model. Inputs include three wood attributes (initial moisture, initial weight, and basic density) and three drying parameters (schedule, conditioning, and post-storage). Results show that initial weight has the highest correlation with the final moisture and possesses the highest relative importance in both predictive and classifier models. This model demonstrated a drop in training accuracy after removing schedule, conditioning, and post-storage from inputs, emphasizing that the drying parameters are significant in the robustness of the model. However, the regression-based model failed to satisfactorily predict the moisture after kiln-drying. In contrast, the classifying model is capable of classifying dried wood into acceptable, over-, and under-dried groups, which could apply to timber pre- and post-sorting. Overall, the gradient-boosting model successfully classified the moisture in kiln-dried western hemlock timber.
RESUMO
The quality control of thermally modified wood and identifying heat treatment intensity using nondestructive testing methods are critical tasks. This study used near-infrared (NIR) spectroscopy and machine learning modeling to classify thermally modified wood. NIR spectra were collected from the surfaces of untreated and thermally treated (at 170 °C, 212 °C, and 230 °C) western hemlock samples. An explainable machine learning approach was practiced using a TreeNet gradient boosting machine. No dimensionality reduction was performed to better explain the feature ranking results obtained from the model and provide insight into the critical wavelengths contributing to the performance of classification models. NIR spectra in the ranges of 1100-2500 nm, 1400-2500 nm, and 1700-2500 nm were fed into the TreeNet model, which resulted in classification accuracy values (test data) of 94.35%, 89.29%, and 84.52%, respectively. Feature ranking analysis revealed that when using the range of 1100-2500 nm, the changes in wood color resulted in the highest variation in NIR reflectance amongst treatments. As a result, associated features were given higher importance by TreeNet. Limiting the wavelength range increased the significance of features related to water or wood chemistry; however, these predictive models were not as accurate as the one benefiting from the impact of wood color change on the NIR spectra. The developed framework could be applied to different applications in which NIR spectra are used for wood characterization and quality control to provide improved insights into selected NIR wavelengths when developing a machine learning model.
RESUMO
Controlling the variability in mat structure and properties in bamboo scrimber (BS) is key to producing the product for structural applications, and wide strip scrimber (WBS) is an effective approach. In this study, the effects of scrimmed bamboo bundle morphology and product density on the properties of WBS were investigated. WBS panels were manufactured and tested using wide (200 to 250 mm) bamboo strips with different fiberization intensity. Maximum strength properties (flexural, compressive, and shear strength), and lowest thickness swelling and water absorption were achieved with three or four passes due to the higher resin absorption by strips. For balanced product cost and performance, we recommend 1-2 fiberization passes and a panel density of 0.9-1.0 g/cm3. Panel mechanical properties were compared with other common bamboo composites. Bamboo scrimber products were highly variable in properties due to differing manufacturing processes, element treatments, and suboptimal mat structure. Products including laminated bamboo lumber and flattened bamboo made from nonfiberized elements show markedly different relationships between strength and elastic properties mostly due to inadequate bonding between the laminae, which causes premature bond-line failure. This study helped improve the understanding of the structure-property relationship of engineered bamboo products while providing insights into process optimization.
RESUMO
Color parameters were used in this study to develop a machine learning model for predicting the mechanical properties of artificially weathered fir, alder, oak, and poplar wood. A CIELAB color measuring system was employed to study the color changes in wood samples. The color parameters were fed into a decision tree model for predicting the MOE and MOR values of the wood samples. The results indicated a reduction in the mechanical properties of the samples, where fir and alder were the most and least degraded wood under weathering conditions, respectively. The mechanical degradation was correlated with the color change, where the most resistant wood to color change exhibited less reduction in the mechanical properties. The predictive machine learning model estimated the MOE and MOR values with a maximum R2 of 0.87 and 0.88, respectively. Thus, variations in the color parameters of wood can be considered informative features linked to the mechanical properties of small-sized and clear wood. Further research could study the effectiveness of the model when analyzing large-sized timber.