Palmer amaranth and waterhemp are two invasive pigweed species, which have become most troublesome to crops, especially corn and soybean. Among these two weed species, Palmer amaranth is more harmful to crops as it can grow faster, spread rapidly, and reduce crop yields significantly when compared to waterhemp. Distinguishing Palmer amaranth from waterhemp is important for effective weed management and an increase in crop production. However, differentiating these two weeds in the early stage is considerably difficult owing to their similar morphological characteristics. In the current study, three artificial intelligence approaches, namely machine learning (ML), deep learning (DL), and object detection (OD) were employed to automate the identification of greenhouse-grown Palmer amaranth and waterhemp within two weeks after emergence, from their RGB images. Aspect ratio, roundness, and circularity were measured and supplied as the input for the ML classification models. Among the four ML models employed, the random forest model achieved the top classification accuracy of 70% with only 312 training instances. In the case of deep learning, the proposed convolutional neural network model trained on a single-object RGB image of Palmer amaranth and waterhemp achieved a classification accuracy of 93%, outperforming the top ML model. The image dataset used for the DL model increased from the original size of 2,000 to 16,000 by various augmentation techniques. Finally, a transfer-learning-based object detection model for localized identification of the weeds was designed. The OD model was developed by fine-tuning the head of YOLOv5 trained on the COCO dataset with 3,200 single-object images (images with single foliage of either Palmer amaranth or waterhemp). The OD model developed in this study achieved an accuracy of 83.5% and it can identify the weed foliages irrespective of their size and proximity to each other.
Weeds pose a major challenge in achieving high yield production in corn. The use of herbicides although effective can be expensive and their excessive use poses ecological concerns and herbicide resistance. Precise identification of weeds using Machine Learning (ML) models significantly reduces the use of herbicides. In this study, we provide a brief overview of the important ML methods used for identifying weeds in corn i.e., classification and object detection. The various metrics that are used for the evaluation of the performance of ML methods are also discussed. In the end, we identify some important research gaps which warrant future investigation. Most ML methods for the identification of weeds use digital images as input data, however, in some cases, hyperspectral data were used. Most of the current studies employ support vector machines and neural networks for the identification of weeds. Classification accuracy and F1 score are the two most frequently used accuracy metrics to evaluate the performance of ML models used. Future research on the identification of weeds may focus on improving the data volume using data augmentation, transfer learning to benefit from existing models, and interpretability of neural networks to avoid overfitting and make models more transparent.
Assessing the quality of soybean seeds for tofu production traditionally requires the actual creation of tofu, a process that demands considerable time and effort. This study addresses this issue by employing machine learning to predict tofu quality from Hyperspectral Imaging (HSI) images of soybean seeds. Two hundred varieties of soybean seeds scanned with HSI have been classified into four categories based on their qualities of tofu products. Upon comparison, XGBoost was employed to pinpoint ten critical HSI wavelengths that show a potential correlation with the protein, carbohydrate, and oil contents in the soybean seeds. Subsequently, a Convolutional Neural Network model was formulated to forecast tofu quality, basing its predictions on HSI data of soybean varieties. Remarkably, the model successfully segregated the soybeans into four unique classes, demonstrating a predictive accuracy that varied between 96% and 99%. This research amalgamates cutting-edge technologies to revolutionize the conventional assessment of soybean seeds.
A new failure criterion has been developed to improve modeling of orthotropic structural composites subjected to quasi-static and impact loadings. Rather than using an analytical expression that traditionally has been employed to predict failure, a point cloud failure surface is constructed in the stress/strain space using a combination of virtual and laboratory testing. These discrete points are obtained by building a micromechanical model that is subjected to uniaxial and multiaxial states of stress until the first failure of a finite element in the model is detected. The post-peak response behavior is then activated till the element meets the erosion criterion and is deleted from the model. One of the challenges in using the generated point cloud data is the ability to correctly predict when failure onset in a finite element takes place without being too conservative. Three predictive methods are compared in this paper – Approximate Nearest Neighbor (ANN), Simplified Approximate Nearest Neighbor (SANN), and Neural Network (NN). Point cloud data from a unidirectional composite, the T800-F3900, commonly used for aerospace applications, is used for comparative evaluation of these methods. The performances are first evaluated using a standalone program not connected with FE analysis. Finally, two of these methods (SANN and NN) are implemented in a commercial finite element program, LS-DYNA, and their performances are evaluated by simulating a laboratory impact test. Results indicate that the SANN and NN implementations are robust, efficient, and accurate.
In this study, we investigated the composition and mechanical properties of metallurgical phases present in the ASTM A36 steels subjected to postfire temperatures using nanoindentation testing in conjunction with the K++ clustering method. The specimens are exposed to target temperatures from 500°C to 1,000°C, with increments of 100°C. We extracted two nanomechanical properties, namely, hardness and Young's modulus, from the nanoindentation tests and used them as descriptive features for the clustering analysis. Results obtained from this analysis show that average volume fractions of ferrite and pearlite were 84% and 16%, respectively. The results also revealed that the mean hardness values were in the range of 2.46 to 3.01 GPa for ferrite and 3.11 to 4.27 GPa for pearlite for the different temperature exposures. The Young's moduli of ferrite ranged from 171.7 to 203.3 GPa, whereas the pearlite phase ranged from 181.1 to 206.8 GPa for the different temperature exposures. The obtained results also indicated the existence of a quadratic relation between the pearlite's mean nanoindentation hardness and the yield and tensile strengths of different postfire ASTM A36 steels.
In this study, we propose a complex-step convolutional autoencoder to identify the regions that are important in a metal microstructure for compact representation and secure sharing. Firstly, the architecture of a convolutional autoencoder is designed for the compact representation of microstructural images. The designed autoencoder achieved a high image compression ratio of 32 without loss of important information. Secondly, an in-home developed model agnostic sensitivity analysis using complex step derivative approximation is implemented on convolutional autoencoders to identify regions of the microstructure that are important for reconstruction. Finally, saliency maps that highlight the importance of pixels for reconstruction are generated for three grades of dual-phase structural steels. The saliency maps indicated secondary phase regions and grain boundaries are important for microstructure image reconstruction. The proposed approach produces more tenable saliency explanations compared to guided backpropagation and layer wise relevance propagation methods. The decoder part of the convolutional autoencoder can be used as a key that could be used to reconstruct the actual microstructure from encoded image information contributing to secure and efficient sharing of microstructure data. The proposed framework is generic and can be extended to identify important microstructural regions for other metals, composites, biomaterials, and material systems.
'%3e%3ccircle r='20' fill='%23008'/%3e%3ccircle r='17.5' fill='%23fff'/%3e%3ccircle r='3.5' fill='%23008'/%3e%3cg id='d'%3e%3cg id='c'%3e%3cg id='b'%3e%3cg id='a' fill='%23008'%3e%3ccircle r='.875' transform='rotate(7.5 -8.75 133.5)'/%3e%3cpath d='M0 17.5.6 7 0 2l-.6 5L0 17.5z'/%3e%3c/g%3e%3cuse xlink:href='%23a' transform='rotate(15)'/%3e%3c/g%3e%3cuse xlink:href='%23b' transform='rotate(30)'/%3e%3c/g%3e%3cuse xlink:href='%23c' transform='rotate(60)'/%3e%3c/g%3e%3cuse xlink:href='%23d' transform='rotate(120)'/%3e%3cuse xlink:href='%23d' transform='rotate(-120)'/%3e%3c/g%3e%3c/svg%3e)