Hong-Chao Zhang

Metal additive manufacturing (AM) is a key enabling technology for fabricating high-value components, yet its qualification is hindered by complex defect mechanisms and the inability of current intelligent assistants to effectively integrate mechanistic literature with in-situ monitoring data. To address these challenges, this study proposes additive manufacturing-vision language model (AM-VLM), an agentic multimodal expert system that integrates mechanistic literature and melt pool case evidence via a physics-aware dual-index retrieval-augmented generation framework. We curate a foundation literature index and construct a multimodal melt pool case index, and retrieve evidence from both repositories in parallel. To improve evidence fidelity, we design a modality-aware evidence refinement strategy that performs cross-encoder reranking for textual candidates and applies physics-aware reranking guided by normalized melt pool morphology and intensity cues. We further implement an incrementally maintainable knowledge update workflow, supported by a network agent and manual curation, to continually incorporate new processes and operating conditions. Building on this evolving knowledge base, we develop a synergistic knowledge-data diagnosis pipeline that couples monitoring-derived evidence with mechanism-grounded reasoning to produce defect identification, root-cause attribution, and parameter-level mitigation suggestions. AM-VLM’s question-answering capability is validated on an AM Q&A benchmark, where it improves expert-rated answer accuracy from 0.368 to 0.808 compared with baselines. On literature and melt pool image retrieval tasks, the proposed cross-encoder reranking increases the nDCG@5 of retrieved literature evidence from 0.783 to 0.886, while the physics-aware reranking boosts the image defect classification F1 score from 0.824 to 0.946. Demonstrations of the synergistic knowledge-data diagnosis pipeline on directed energy deposition platforms, together with a public LPBF case study, further support that AM-VLM can deliver actionable, traceable diagnosis and process-improvement recommendations. This work establishes a comprehensive, scalable framework that unifies mechanistic reasoning with real-time monitoring for reliable decision support in metal AM.

Metal additive manufacturing (AM) is a key enabling technology for fabricating high-value components, yet its qualification is hindered by complex defect mechanisms and the inability of current intelligent assistants to effectively integrate mechanistic literature with in-situ monitoring data. To address these challenges, this study proposes additive manufacturing-vision language model (AM-VLM), an agentic multimodal expert system that integrates mechanistic literature and melt pool case evidence via a physics-aware dual-index retrieval-augmented generation. We curate a foundation literature index and construct a multimodal melt pool case index, and retrieve evidence from both repositories in parallel. To improve evidence fidelity, we design a two-stage reranking strategy that performs cross-encoder reranking for textual candidates and applies physics-aware reranking guided by normalized melt pool morphology and intensity cues. We further implement an incrementally maintainable knowledge update workflow, supported by a network agent and manual curation, to continually incorporate new processes and operating conditions. Building on this evolving knowledge base, we develop a synergistic knowledge–data diagnosis pipeline that couples monitoring-derived evidence with mechanism-grounded reasoning to produce defect identification, root-cause attribution, and parameter-level mitigation suggestions. AM-VLM’s question-answering capability is validated on an AM Q&A benchmark, where it improves expert-rated answer accuracy from 0.250 to 0.867 compared with baselines. On literature and melt pool image retrieval tasks, the proposed cross-encoder reranking increases the nDCG@5 of retrieved literature evidence from 0.783 to 0.886, while the physics-aware reranking boosts the image retrieval F1 score from 0.824 to 0.946. Demonstrations of the synergistic knowledge-data diagnosis pipeline on a directed energy deposition platform further confirm that AM-VLM delivers actionable, evidence-traceable diagnosis and process-improvement recommendations. This work establishes a comprehensive, scalable framework that unifies mechanistic reasoning with real-time monitoring for reliable decision support in metal AM.

With the increasing integration level of chips, the number of transistors and functional modules inside the chip is also increasing. It is complex to design the clock tree interconnects in a chip. At the same time, under the influence of high operating frequency and advanced technology, the impact of electromigration is highlighted on the clock tree interconnects. In order to ensure the performance of chip design and improve reliability, it is particularly important to quickly and accurately analyze the impact of electromigration problems on clock tree interconnects. In this paper, we have proposed a simulation analysis method based on COMSOL for the electrothermal coupled migration of clock trees. This method takes into account the effects of environmental factors and interconnect structures on electrothermal migration, qualitatively describing their relationship and exploring the impact of several basic clock interconnect structures on electrothermal migration.

We demonstrate in-plane field-free-switching spin-orbit torque (SOT) magnetic tunnel junction (MTJ) devices that are capable of low switching current density, fast speed, high reliability, and, most importantly, manufactured uniformly by the 200-mm-wafer platform. The performance of the devices is systematically studied, including their magnetic properties, switching behaviors, endurance and data retention. The successful integration of SOT devices within the 200-mm-wafer manufacturing platform provides a feasible way to industrialize SOT MRAMs. It is expected to obtain excellent performance of the devices by further optimizing the MTJ film stacks and the corresponding fabrication processes in the future.

Almost all of the research on cracks in laser cladding is based, at present, on a horizontal substrate, which cannot be directly applied to prepare high performance coatings, especially high hardness coatings, on tilted substrate. In this work, the influence of the substrate’s tilt angle on the crack behavior of high hardness Ni60A coating is studied, based on the laser intensity distribution and energy attenuation models on the tilted substrate. Results show that the cracking rate (the crack number in the unit’s cladding length) of the coating increases with the increasing substrate tilt angle, but the tilt angle has no significant influence on the crack mechanism. The different lap direction has a certain influence on the crack, and the coating prepared by downward lap cladding has a larger cracking rate due to the greater laser energy loss. Furthermore, with the increasing substrate tilt angle, the residual stress increases due to the decreased plastic flow, and the fracture strength decreases due to the decreased dilution rate, which results in the increase in the cracking rate of the Ni60A coating. This work will broaden the application of laser cladding technology on repairing complex parts such as gear and blades.

Electric vehicles (EV) are a crucial technology to tackle environmental issues like air pollution and energy depletion. In this paper, we use a static computable equilibrium model to estimate the effects of two different policies: the subsidy and the credit regulation. The results show that both policies have negligible effects on the macro economy. The subsidy on EV creates more emissions as the substitution between conventional vehicles and EV isn’t very effective, leading to more production. The credit regulation, otherwise, suppresses the production of conventional vehicles while encouraging the production of EV, thus, emissions of CO₂, NO_x, Dust, COD are largely reduced. And the combination of the two policies has a mixed effect. However, the emissions of SO₂ and NH₃-N increase no matter which policy is implemented, suggesting that different sectors perform diversely under the stimulus. Policy-maker needs to make a trade-off between the benefits and risks of certain policies.

5G is the latest-generation network that powers the IoT in pace with energy transformation essentially demanded during the modern world, especially the wireless form. Laser charging means laser beam irradiance of the solar cell panel to generate the photovoltaic power, and its way belongs to the WPT (Wireless Power Transfer). During the process, uncertain and ungovernable factors affect the energy transfer of efficiency. We have chosen the In_0.3Ga_0.7As material as the semiconductor of the PV (photovoltaic) panel to enhance its photoelectric conversion efficiency. Nevertheless, during the PV processing, nonlinearity elements impact the prediction modelling, making it extremely difficult. This paper proposes a new forecasting model based on SVM (Support Vector Machine) which is drawn from SLT (Statistical Learning Theory). Since the forecast belongs to classification issue essentially, the main idea of SVM is extending the SLT dimensions. We used the kernel function for realization. The PV is done under the different conditions (constant temperature, gradient temperature, or various laser power). The gathered data is processed by MATLAB and the 3rd software: LIBSVM. Without the parameter optimization, the prediction accuracy of classification is 80 (32/40) against 100 (40/40). Considering that the experiment data have small quantity, further make the over 650 data size samples which are calculated by the formula to verify the model. Through the effect of a few random factors, the result still remains 80 or higher. The research work has resulted in a solution of a fast and precise prediction model in laser charging solar cell panel. It also revealed the implicit relationship between the factors of the PV.