Abstract：The airborne pod system can reduce the difficulty of power line inspection and enhance the degree of automation for inspection tasks.At the same time it also has a higher requirement of the structure performance of the airborne pod system.Therefore the finite element statics analysis and modal analysis for the structure of the airborne pod system was carried by using ANSYS WORKBENCH software.From the analysis result, the maximum stress of the airborne pod system structure is 16.561MPa，the maximum deformation is 0.14705 mm and the first three order natural frequencies are 19.258Hz, 21.187 Hz and 63.44 Hz.And the weak links relatively of the structure were found out.Analysis results show that the structure of the airborne pod system has a good performance and can meet the requirement of airborne conditions.
Abstract Needle‐punched carbon/carbon composites (NP C/Cs) are advanced materials widely used in aerospace applications. The needle‐punching technique improves the structural integrality of carbon‐fiber plies; however, it also introduces many defects affecting the mechanical behavior of NP C/Cs. To investigate the behavior of needle‐punched carbon/carbon composites under bending loads, a circular‐arc beam element with a needle‐hole defect is developed to model the punched carbon fibers and an extended spring element is suggested to model the matrix. This model allows quick predictions of the bending modulus, strength and progressive damage of NP C/Cs. It is shown that efficient numerical predictions agree well with the experimental results, thus validating the proposed method.
Large language models (LLMs) have shown remarkable abilities in diverse natural language processing (NLP) tasks. The LLMs generally undergo supervised fine-tuning (SFT) followed by preference alignment to be usable in downstream applications. However, this sequential training pipeline leads to alignment tax that degrades the LLM performance. This paper introduces PAFT, a new PArallel training paradigm for effective LLM Fine-Tuning, which independently performs SFT and preference alignment (e.g., DPO and ORPO, etc.) with the same pre-trained model on respective datasets. The model produced by SFT and the model from preference alignment are then merged into a final model by parameter fusing for use in downstream applications. This work reveals important findings that preference alignment like DPO naturally results in a sparse model while SFT leads to a natural dense model which needs to be sparsified for effective model merging. This paper introduces an effective interference resolution which reduces the redundancy by sparsifying the delta parameters. The LLM resulted from the new training paradigm achieved Rank #1 on the HuggingFace Open LLM Leaderboard. Comprehensive evaluation shows the effectiveness of the parallel training paradigm.
Different Al2O3 carriers were synthesized by co-precipitation and sol-gel method. From them, 4%NiO-20%MoO3/Al2O3 catalysts were prepared by incipient wetness impregnation. The catalysts were characterized by X-ray diffraction analysis (XRD), N2 adsorption-desorption, NH3-temperature programmed desorption (TPD) and H2-temperature programmed reduction (TPR) and subsequently used for selective hydrogenation of naphthalene to high-value tetralin. The results showed that Ni-Mo/so-ge Al2O3 (900) exhibited better catalytic performance than Ni-Mo/commercial Al2O3, achieving 99.56% naphthalene conversion and 99.43% tetralin selectivity.
The research on the fatigue performance of reclaimed asphalt mixture and the establishment of its fatigue life prediction model is the key to its large-scale engineering application. The multi-factor comprehensive four-point bending fatigue tests of the fresh and recycled asphalt mixtures under the strain control of cyclic loading were conducted to investigate influencing factors on their fatigue performance and establish a new fatigue prediction model for RAP. The hot recycled asphalt mixture samples of AC-16C were prepared by a uniform test design method, with the RAP content of 0%, 20%, 30%, and 40%. Through the fatigue test data of 20 groups of fresh asphalt mixture with 0% RAP and 40 groups of recycled asphalt mixtures with the content of 20%, 30%, and 40% RAP, the simplified applicability of the JTG fatigue model and its parameters modification were studied, thereby a comprehensive fatigue prediction model of recycled asphalt mixture is established on consideration of strain level, asphalt content, void ratio, and RAP mixing amount. This fatigue prediction model of recycled asphalt mixture is proposed by modifying the parameters of the fatigue formula of the new asphalt mixture in the specification and comparing the fatigue test data of the recycled asphalt mixture. The research results show that: (1) The simplified JTG fatigue prediction model can effectively predict the fatigue life of fresh asphalt mixture with the content of 0% RAP, and the model parameter a is a constant of the order of 10 16 ; (2) The correction coefficients α and β of the initial bending stiffness modulus S 0 and the Voids Filled with Asphalt ( VFA ) in the improved JTG model of the hot recycled asphalt mixture considering the RAP content are 0.006 and 0.0136, respectively; (3) The model can accurately predict the fatigue life of recycled asphalt mixtures with the range of RAP content of 0-40%, with an average deviation of only 0.106, and has higher prediction accuracy for fatigue life measured close to or higher than 10 6 times.
Name disambiguation plays an important role in text processing. Names in natural language text are usually ambiguous, as many people share a same name, so do organizations and entities in other domains. The goal of name disambiguation is to identify the different referents of a same name in corpus. Technology of name disambiguation can be applied in web searching, information extraction, translation system, questioning-answer system, and so on. The key part to solving name ambiguity problem is to model context of names and feature of candidates, and then measure similarity. We proposed a representation, Bag of Entity with Property, BoEwP, to model a document, the context of names. Then, we presented an online disambiguation method for person or organization names based on entity and property co-occurrence. We performed experiments on Chinese, Spanish and English corpus. Our approach got nice precision and recall measurements, and required low system resources.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)