Zhejiang Research Institute of Chemical Industry

Federated learning effectively addresses the issues of data privacy and communication overhead in traditional deep learning through distributed local training. However, its open architecture is seriously threatened by backdoor attacks, where malicious clients can implant triggers to control the global model. To address these issues, this paper proposes a novel three-stage defense mechanism based on local clients. First, through text readability analysis, each client’s local data is independently evaluated to construct a global scoring distribution model, and a dynamic threshold is used to precisely locate and remove suspicious samples with low readability. Second, frequency analysis and perturbation are performed on the remaining data to identify and disrupt triggers based on specific words while preserving the basic semantics of the text. Third, n-gram distribution analysis is employed to detect and remove samples containing abnormally high-frequency word sequences, which may correspond to complex backdoor attack patterns. Experimental results show that this method can effectively defend against various backdoor attacks with minimal impact on model accuracy, providing a new solution for the security of federated learning.

Excessive Desulfovibrio (Des) forms biofilm to enable dominant occupation of the intestinal niche, representing a common pathogenic driver of multiple inflammatory bowel disease (IBD) types. Colonization resistance constitutes the primary barrier to antagonistic probiotic efficacy, and this is driven by the pathogen-favorable microenvironment established by Des. Here, probiotic Lactiplantibacillus (Lap) is modified by calcium ions (Ca2+) and calcium-regulative polyphenol (kaempferol-3-O-rutinoside, KAE) via coordinate interaction to achieve intestinal niche reconfiguration. Targeting the tripartite mechanisms of Des-mediated colonization resistance, the Ca2+/KAE@Lap platform optimized niche competition through Ca2+-bridged interfacial binding with directional bactericidal activity, and this enables bacterial replacement at occupied sites. Ca2+/KAE@Lap reestablishes calcium homeostasis disrupted by Des via synergistic Ca2+/KAE regulation, dually restoring epithelial energy metabolism and mucus layer reconstitution, counteracting Des-induced colonized sites contraction and regenerated site impairment. This drives phenotypic shift in biofilm composition from Des-dominated to Lap-enriched consortia, which is concomitant with the redirection of intestinal colonization resistance from a pathogen-permissive to a probiotic-favored state. This calcium-based biofilm transformation strategy overcomes the transient colonization limitation inherent in conventional probiotic therapies by effectively disrupting colonization resistance in IBD treatment.

To overcome the problem of sticky oxide films hindering the complete energy release of boron powder during ignition, graphene fluoride (FG) was used. FG, selected for its high fluorine content, low hydrogen content, low surface free energy, and high thermal conductivity, reacts with boron to produce high combustion heat and large amounts of gaseous products, effectively removing the viscous oxide film and preventing boron agglomeration. FG was synthesized via shear emulsification cutting, and FG/B composites were fabricated via solvent evaporation. Thermal analysis and combustion heat testing revealed that the prepared FG flake had an average thickness of 1.56 nm. For 20% FG addition, the FG/B composite exhibited a 1.86-fold increase in combustion heat. Moreover, 20% FG addition effectively improved the combustion heat release of boron powder, making it comparable with the combustion heat release of high-energy metal powder/boron powder composites.

The prediction of key quality variables plays an important role in industrial status identification and monitoring. Due to process disturbance and hard device limitation, data collection in modern industries often exhibits high noise and irregular data sampling. To solve the above problems, this article proposes a stacked supervised and reconstructed input denoising autoencoder integrated with internal attention long short-term memory (SSRDAE-IALSTM) network for soft sensing modeling. First, a stacked supervised and reconstructed input denoising autoencoder (SSRDAE) is designed. Compared with the original DAE, each supervised and reconstructed input DAE (SRDAE) can simultaneously reconstruct the process data and quality data at the output layer, aiming to reduce information loss and extract quality-related features. Second, the denoised features are fed into the interval attention LSTM (IALSTM) to adjust the influence of different historical samples on the current sample in irregular sampling data to capture long-term temporal features. Finally, performance validations are carried out on an industrial debutanizer column and a penicillin fermentation process. The experimental results show that the proposed model can enhance the learning ability of process features and obtain better prediction performance than other comparison methods.

ABSTRACTTrifluoromethyloxadiazole compounds (TFMO) are a class of anti‐rust disease targeting histone deacetylases inhibitors (HDACs). In this paper, a series of 5‐(trifluoromethyl)‐1,2,4‐oxadiazole substituted benzamide derivatives containing 5‐(trifluoromethyl)‐1,2,4‐oxadiazole moiety were designed and synthesized. Anti‐rust bioassay results showed some of them possessed excellent activities against Puccinia sorghi at 200 mg/L. Furthermore, the structure–activity relationship was established using in silico methods, and the most active compound binding mode was carried out based on the crystal structure of human HDAC1, human HDAC4, and human HDAC6. This work provided an excellent fungicide against rust for further optimization.

Abstract The efficiency of water electrolysis for hydrogen production is significantly hindered by the slow kinetics of the oxygen evolution reaction (OER), posing challenges for its industrial application. Herein, a Pt/Ni(OH) 2 self‐supported electrode with a Pt loading as low as 0.036 mg·cm −2 is synthesized via a one‐step sodium dodecyl sulfate assisted anodization coupled with electrodeposition method. The electrode exhibits superior performances for both the hydrogen evolution reaction (HER) and the electrooxidation of hexamethylenediamine (HMDAOR) compared to commercially available 20% Pt/C in both acidic and alkaline systems. Notably, it can be operated continuously over 600 h at an industrial grade current densities of 1400 and 1000 mA·cm −2 in acidic and alkaline electrolytes, respectively, demonstrating its outstanding long‐term stability. It can achieve a current density of 1000 mA·cm −2 for hydrogen production at a cell voltage of 1.561 V in HMDA‐assisted water electrolysis, demonstrating its practical application in water electrolysis. The enhanced performance can be attributed to the Ni‐induced modulation of the Pt electronic structure, strong Pt‐support interaction enhanced the intrinsic activity of Pt, and synergistic effects between Pt and Ni. These provide valuable insights to design advanced catalysts for water electrolysis, and to broad their application in hydrogen production and wastewater treatment areas.