Given the recent global pandemic and domestic labor shortage, there is a pressing demand for digital means that enable construction site managers to obtain information more efficiently in support of their daily tasks. Site-moving employees experience difficulty with conventional software applications. These applications rely on forms and necessitate multiple finger actions, like keystrokes and mouse clicks, making them inconvenient and reducing the desire to utilize them. By providing an intuitive method for user input, conversational AI, also known as a chatbot, can significantly improve the usability and ease of use of any system. This study presents a prototype for an AI-based chatbot, powered by a demonstrated Natural Language Understanding (NLU) model, facilitating site managers' daily inquiries into building component dimensions. The chatbot's answering component utilizes Building Information Modeling (BIM) methodologies. The preliminary assessment of the chatbot's performance indicates its capability to accurately predict intents and entities within queries submitted by site managers, achieving satisfactory levels of accuracy for both intent prediction and answer generation. Site managers can now leverage alternative approaches for obtaining the information they need, as indicated by these results.
Industry 4.0's influence extends to the radical transformation of physical and digital systems, significantly improving the digitalization of maintenance plans for physical assets in an optimal manner. Road network conditions and the prompt implementation of maintenance schedules are fundamental to the success of predictive maintenance (PdM) in road infrastructure. Our PdM strategy utilizes pre-trained deep learning models to efficiently and accurately classify and recognize diverse road crack types. This work investigates the application of deep learning neural networks for the purpose of classifying roads based on the measure of deterioration. Training the network involves teaching it to discern various types of road damage, such as cracks, corrugations, upheavals, potholes, and others. Due to the quantity and severity of the damage sustained, we can quantify the rate of degradation and implement a PdM framework that allows us to identify the intensity of damage occurrences, enabling us to prioritize maintenance strategies. Through the use of our deep learning-based road predictive maintenance framework, stakeholders and inspection authorities can make decisions on maintenance for different types of damage. The effectiveness of our approach was validated by strong results in precision, recall, F1-score, intersection-over-union, structural similarity index, and mean average precision, showcasing the significant performance gains of our proposed framework.
This paper outlines a CNN-based method for detecting algorithm faults within scan-matching to enable accurate simultaneous localization and mapping (SLAM) in dynamic environments. Environmental readings from a LiDAR sensor change in response to the presence of dynamic objects. As a result, the attempt to match laser scans based on scan matching techniques is anticipated to encounter problems. Therefore, a more powerful scan-matching algorithm is crucial for 2D SLAM, surpassing the limitations of existing scan-matching techniques. Laser scan data from a 2D LiDAR, originating from an environment of unknown characteristics, is processed initially. This is subsequently subjected to ICP (Iterative Closest Point) scan matching. Finally, the matched scans are transformed into visual images, which feed a CNN for training the system to detect faults within the scan matching procedure. The trained model, in its final analysis, detects the faults contained within the new provided scan data. The training and evaluation are carried out in various dynamic environments, designed to replicate real-world situations. Across a range of experimental environments, the proposed method's experimental validation demonstrated a high degree of accuracy in detecting scan matching faults.
This study introduces a multi-ring disk resonator, characterized by elliptic spokes, for the purpose of counteracting the aniso-elasticity of (100) single-crystal silicon. To control the structural coupling connecting each ring segment, one can swap out the straight beam spokes with elliptic spokes. An optimized design of the elliptic spokes allows for the degeneration of two n = 2 wineglass modes. For the design parameter of an aspect ratio of 25/27 for the elliptic spokes, a mode-matched resonator could be produced. this website Evidence for the proposed principle was provided by both numerical simulations and physical experiments. surgeon-performed ultrasound Demonstrating an experimentally validated frequency mismatch of just 1330 900 ppm, the current study notably outperforms the 30000 ppm maximum achievable by conventional disk resonators.
Computer vision (CV) applications are gaining significant traction within intelligent transportation systems (ITS) as technology continues its development. These applications are crafted to boost the intelligence and safety of transportation systems, along with their efficiency. Improvements in computer vision significantly contribute to solutions in the realms of traffic flow monitoring and regulation, accident discovery and mitigation, adaptable road usage cost systems, and road surface condition monitoring, and many more associated sectors, by employing a higher degree of efficiency. Analyzing CV applications in the literature and ITS methodologies of machine learning and deep learning, the applicability of computer vision in ITS contexts is evaluated. This survey details the advantages and drawbacks of these technologies and emphasizes emerging research areas crucial for increasing the effectiveness, safety, and efficiency of ITS. This review, which gathers research from various sources, intends to display how computer vision (CV) can contribute to smarter transportation systems. A holistic survey of computer vision applications in the field of intelligent transportation systems (ITS) is presented.
Robotic perception algorithms have greatly benefited from the significant progress in deep learning (DL) technologies observed over the past ten years. In fact, a substantial percentage of the autonomy infrastructure in both commercial and research platforms is reliant on deep learning for environmental perception, specifically with regard to data gathered from vision sensors. Deep learning perception algorithms, which include detection and segmentation networks, were assessed for their suitability to process image-equivalent outputs from advanced lidar devices. Unlike processing volumetric point clouds, this work, as far as we are aware, is the initial endeavor concentrating on low-resolution, 360-degree images acquired by lidar sensors. These images encode depth, reflectivity, or near-infrared light within their respective pixels. fine-needle aspiration biopsy General-purpose deep learning models, following appropriate preprocessing, were shown to be capable of processing these images, making them suitable for use in environmental contexts where vision sensors inherently have limitations. Utilizing both qualitative and quantitative methods, we scrutinized the performance of various neural network architectures. Visual camera-based deep learning models showcase considerable advantages over point cloud-based perception, largely attributed to their much wider proliferation and mature state of development.
For the deposition of thin composite films of poly(vinyl alcohol-graft-methyl acrylate) (PVA-g-PMA) and silver nanoparticles (AgNPs), the blending approach (ex-situ) was chosen. Employing ammonium cerium(IV) nitrate as the initiator, a copolymer aqueous dispersion was synthesized through the redox polymerization of methyl acrylate (MA) onto poly(vinyl alcohol) (PVA). Employing a green synthesis approach, lavender water extracts, derived from essential oil industry by-products, were used to create AgNPs, which were then combined with the polymer. Dynamic light scattering (DLS) and transmission electron microscopy (TEM) measurements were made to determine nanoparticle size and assess their stability over 30 days in suspension. Employing the spin-coating technique, thin films of PVA-g-PMA copolymer were fabricated on silicon substrates, incorporating silver nanoparticles in concentrations ranging from 0.0008% to 0.0260%, subsequently enabling optical property characterization. Film refractive index, extinction coefficient, and thickness were established via UV-VIS-NIR spectroscopy coupled with non-linear curve fitting techniques; concurrently, room-temperature photoluminescence measurements facilitated the study of film emission. Experiments on the film's thickness response to nanoparticle weight concentration revealed a consistent linear trend. The thickness grew from 31 nanometers to 75 nanometers as the nanoparticle weight percentage climbed from 0.3% to 2.3%. Controlled atmosphere tests of the sensing properties toward acetone vapors involved measuring reflectance spectra on a single film spot, both before and during analyte exposure, and the swelling degree was determined and compared to the corresponding undoped films. In films, the concentration of 12 wt% AgNPs proves to be the optimal level for improving the sensing response towards acetone. The films' attributes were investigated, and the consequences of AgNPs were highlighted and expounded.
Advanced scientific and industrial equipment mandates magnetic field sensors possessing high sensitivity, small dimensions, and the ability to function efficiently across a large range of temperatures and magnetic field intensities. Nevertheless, commercial sensors are scarce for gauging high magnetic fields, spanning from 1 Tesla to megagauss. For this reason, the dedicated pursuit of advanced materials and the strategic engineering of nanostructures exhibiting exceptional properties or emerging phenomena is vital for high-magnetic-field sensing applications. The central theme of this review revolves around the investigation of thin films, nanostructures, and two-dimensional (2D) materials, which show non-saturating magnetoresistance across a broad range of magnetic fields. Findings from the review indicated that modifying the nanostructure and chemical makeup of thin, polycrystalline ferromagnetic oxide films (manganites) can produce a noteworthy colossal magnetoresistance, reaching a level of up to megagauss.