多共振分量融合CNN的行星齿轮箱故障诊断

针对行星齿轮箱中各部件所激起的振动成分混叠、早期故障特征经常被较强的各级齿轮谐波成分以及环境噪声所湮没的问题，提出一种多共振分量融合卷积神经网络（multi-resonance component fusion based convolutional neural network，简称MRCF-CNN）的行星齿轮箱故障诊断方法。首先，对振动信号进行共振稀疏分解，得到包含齿轮谐波成分的高共振分量和可能包含轴承故障冲击成分的低共振分量；其次，构建多共振分量融合卷积神经网络，将得到的高、低共振分量和原始振动信号进行自适应的特征级融合，通过有监督的方式训练模型并进行行星齿轮箱故障诊断。对行星齿轮箱实验数据的分析结果表明，该方法能够有效分类行星齿轮箱中滚动轴承和齿轮的故障，成功对行星齿轮箱故障进行诊断，同时能够进一步增强卷积神经网络对振动信号所蕴含的故障信息的辨识能力。     

Concept design of a novel reformer producing hydrogen for internal combustion engines using fuel decomposition method: Performance evaluation of coated monolith suitable for on-board applications

Hydrogen addition effectively reduces the fuel consumption of spark ignition engines. We propose a new on-board reformer that produces hydrogen at high concentrations and enables multi-mode operations. For the proposed reformer, we employ a catalytic fuel decomposition reaction via a commercial NiO–CaAl₂O₄ catalyst. We explore the physical and chemical aspects of the reforming process using a fixed bed micro-reactor operating at temperatures of 550–700 °C. During reduction, methane is decomposed to form hydrogen and carbon. Carbon formation is critical to hydrogen production, and free space for carbon growth is essential at low temperatures (≤600 °C). We define a new accumulated conversion ratio that quantitatively measures highly transient catalytic decomposition. The free space of the coated monolith clearly aided low-temperature decomposition with negligible pressure drop. The coated substrate is therefore suitable for on-board applications considering that our reformer concept also utilizes the catalytic fuel decomposition reaction.     

MgFe and Mg–Co–Fe mixed oxides derived from hydrotalcites: Highly efficient catalysts for COx free hydrogen production from NH3

Mg–Fe Layered Double Hydroxide (LDH) with M²⁺: M³⁺ 3:1 stoichiometric ratio was synthesized and employed as catalyst precursor for CO_x-free hydrogen production from ammonia. The resulting catalyst showed good catalytic activity. A series of Mg/Co–Fe layered double hydroxides were synthesized by replacing Mg²⁺ with Co²⁺ without disturbing M²⁺:M³⁺ ratio. The influence of nature and extent of Co(II) substitution on structure, morphology and surface properties were studied. A systematic study was carried out using these materials as catalyst precursors for ammonia decomposition. BET, XRD, TPR, XPS, CO₂-TPD and TEM techniques were used to characterize the synthesized catalysts. These Fe-based catalysts are highly active, highly stable and not promoting any stable surface nitridation during the ammonia decomposition reaction. Among all catalysts, the Mg₃Co₃Fe₂ catalyst showed the highest activity i.e. 100% conversion at 6,000 h⁻¹ and 60% at 50,000 h⁻¹ space velocities at 550 °C. The registered superior catalytic activity was result of the formed specific catalyst's properties like high surface area, high surface Co and Fe atomic concentration and suitable basicity. These Fe-based materials are, cost-effective, easily synthesize and highly stable, thus attractive for large-scale operation.     

Cobalt doping of α-Fe/Al2O3 catalysts for the production of hydrogen and high-quality carbon nanotubes by thermal decomposition of methane

Fe–Co/Al₂O₃ catalysts were developed and tested in the catalytic decomposition of methane (CDM) for the synthesis of multi-wall carbon nanotubes (MWCNT) and the CO₂-free hydrogen production. While Fe (54.5–66.7 mol.%) is the main active phase for the carbon formation on the catalyst, Co acts as dopant aiming to improve its overall catalytic behaviour. Catalysts with Co contents of up to 18.2 M% showed the presence of α-Fe and Fe–Co crystallites with different size and lattice parameter. Fe_1-xCo_x alloy with bcc crystal system was identified only for Co contents of 14.0% and above, and presented a lattice constant lower than α-Fe, which would modify the carbon diffusion of the metal particle during the MWCNT growth. Co inhibited the Fe₃C formation during CDM resulting in higher carbon formations and longer activity times. This phase, shown in undoped catalysts, favored the presence of bamboo-type carbon nanotubes.     

基于改进ReaxFF 力场的全氟己酮热分解机理研究

采用基于密度泛函的紧束缚模型（DFTB）和改进的反作用力场（ReaxFF）对全氟己酮的热分解机理进行反应分子动力学仿真研究，分析了热分解的初始反应路径、微观反应过程以及体系内气体的生成机理，探讨了温度对全氟己酮热分解的影响，建立了以仿真数据为基础的一级动力学方程并对热分解过程进行一级动力学评估。结果表明，全氟己酮热分解的起始反应为：全氟己酮→全氟异戊烷+一氧化碳，全氟己酮热分解的主要产物为一氧化碳、四氟甲烷、六氟丙烯、全氟异丁烯、十氟丁烷、全氟异戊烷及大量活性基团。经一级动力学评估得到指前因子和表观活化能分别为5.021 7×1014 s-1和1 771.96 kJ/mol。     

基于VMD瞬时能量与GA- RBF的滚动轴承故障诊断

针对滚动轴承早期故障的有效识别,提出了一种基于VMD瞬时能量与GA优化的RBF神经网络的滚动轴承故障诊断方法,可以有效对滚动故障做出诊断。首先,VMD将滚动轴承振动信号进行分解成合适数目的本证模态函数;其次,计算本证模态函数分量的瞬时能量并组成特征向量;最后,将特征向量输入到GA优化的RBF神经网络实现轴承故障识别。通过滚动轴承故障诊断实验对该方法进行验证。结果表明,该方法识别滚动轴承故障的准确率为96.43%,较默认参数的RBF神经网络和EEMD瞬时能量与GA-RBF神经网络有明显的提高,证明了所提方法的可行性。     

回转窑人工智能专家系统的应用实践

根据水泥熟料煅烧过程操作控制原理研发的"回转窑人工智能专家系统"采用了人工智能模拟技术、实时数据清洗技术、控制数据关联匹配技术。该系统由分解炉出口温度自动控制系统、篦速自动控制系统和氨水用量自动控制系统三个模块组成。该系统投运后,分解炉出口温度控制有效率由之前人工控制的平均60%~70%提高至平均80%~95%;篦下压力和二次风温稳定性提高;氨水消耗量下降显著;窑系统投料量增加10 t/h左右,熟料标准煤耗降低2.85 kg/t,熟料强度均有提高。     

Production of hydrogen by catalytic methane decomposition using biochar and activated char produced from biosolids pyrolysis

Catalytic methane decomposition (CMD) was studied by employing biochar and activated char of biosolids’ origin under different reaction temperatures and methane concentrations. Higher reaction temperatures and lower inlet methane concentrations were found to be favourable for achieving higher methane conversion. A maximum initial methane conversion of 71.0 ± 2.5 and 65.2 ± 2.3% was observed for activated char and biochar, respectively at 900 °C and for 10% CH₄ in N₂ within the first 0.5 h of experiment. Active sites from oxygen containing carboxylic acid functional groups and smaller pore volume and pore diameter were attributed to assist in higher initial methane conversion for biochar and activated char respectively. However, rapid blockages of active sites and surfaces of biochar and activated char due to carbon formation have caused a rapid decline in methane conversion values in the first 0.5 h. Later on, crystalline nature of the newly formed carbon deposits due to their higher catalytic activity have stabilised methane conversion values for an extended experimental period of 6 h for both biochar and activated char. The final conversion values at the end of 6 h experiment with biochar and activated char at 900 °C and for 10% CH₄ in N₂, were found to be 40 ± 1.9 and 35 ± 1.6% respectively. Analysing carbon deposits in detail revealed that carbon nanofiber type structures were observed at 700 °C while nanospheres of carbon were found at 900 °C.     

Methane thermocatalytic decomposition to COx-free hydrogen and carbon nanomaterials over Ni–Mn–Ru/Al2O3 catalysts

Thermocatalytic decomposition of methane is proposed to be an economical and green method to produce CO_x-free hydrogen and carbon nanomaterials. In this work, the catalytic performance of Ni–Mn–Ru/Al₂O₃ catalyst under different reaction parameters (such as, pre-reduction temperature, reaction temperature, space velocity, etc.) were investigated to obtain optimum reaction conditions. The catalysts were characterized by N₂ adsorption/desorption, X-ray diffraction, inductively coupled plasma optical emission spectrometer and hydrogen temperature programmed reduction. For the 60 wt% Ni-5 wt% Mn-10 wt% Ru/Al₂O₃ catalyst using Ru(NO)(NO₃)_x(OH)_y(x + y = 3) as Ru precursor, the methane conversion rate obtained is high as 93.76% under optimum reaction conditions (reduction at 700 °C for 1 h, reaction at 750 °C, GSHV = 36,000 mL/g_cat h). Carbon nanomaterials formed during the process of methane thermocatalytic decomposition were characterized by scanning electron microscopy, thermal gravimetric analyzer and Raman spectroscopy. Carbon nanofibers were formed over all the Ni–Mn–Ru/Al₂O₃ catalysts.     

综合改进奇异谱分解和奇异值分解的齿轮故障特征提取方法

针对齿轮故障特征微弱，在强背景噪声下难以有效提取的问题，提出了一种改进奇异谱分解（ISSD）结合奇异值分解（SVD）的齿轮故障特征提取方法。针对奇异谱分解（SSD）算法中模态参数需凭经验选取的缺陷，基于散布熵优化算法对SSD算法进行了改进，在得到既定的一组奇异谱分量的基础上，根据峭度值最大准则筛选出了最佳奇异谱分量并进行了SVD处理，采用奇异值能量标准谱自适应地确定了信号重构阶数以还原信号和提高降噪效果。最后对信号进行包络解调以提取齿轮故障特征，将所提方法运用到仿真信号和齿轮实测信号中，并同传统包络谱、SSD包络谱以及经验模态分解结合SVD(EMD-SVD)方法进行了对比分析，结果表明，所提方法的降噪和特征提取效果更佳，能够更加有效地实现齿轮故障的判别。