重新找回人工智能的可解释性

针对深度神经网络AI研究的可解释性瓶颈，指出刚性逻辑（数理形式逻辑）和二值神经元等价，二值神经网络可转换成逻辑表达式，有强可解释性。深度神经网络一味增加中间层数来拟合大数据，没有适时通过抽象把最小粒度的数据（原子）变成粒度较大的知识（分子），再把较小粒度的知识变成较大粒度的知识，把原有的强可解释性淹没在中间层次的汪洋大海中。要支持多粒度的知识处理，需把刚性逻辑扩张为柔性命题逻辑（命题级数理辩证逻辑），把二值神经元扩张为柔性神经元，才能保持强可解释性。本文详细介绍了从刚性逻辑到柔性逻辑的扩张过程和成果，最后介绍了它们在AI研究中的应用，这是重新找回AI研究强可解释性的最佳途径。     

Microstructure evolution during thermomechanical processing in 3Mn-0.1C medium-Mn steel

ABSTRACT

Medium-Mn steels are energetically investigated as a candidate of the third generation advanced high strength steels (AHSSs). However, their phase transformation and microstructaure evolution during various heat treatments and thermomechanical processing are still unclear. The present study first confirmed the kinetics of static phase transformation behaviour in a 3Mn-0.1C medium-Mn steel. Hot compression tests were also carried out to investigate the influence of high-temperature deformation of austenite on subsequent microstructure evolution. It was found that static ferrite transformation was quite slow in this steel, but ferrite transformation was greatly accelerated by the hot deformation in austenite and ferrite two-phase regions. Characteristic dual-phase microstructures composed of martensite and fine-grained ferrite were obtained, which exhibited superior mechanical properties.

This paper is part of a Thematic Issue on Medium Manganese Steels.     

Mechanical and thermal properties and microstructural evolution of Ta-doped Nb4AlC3

(Nb_1-xTa_x)₄AlC₃ (x = 0–0.5) ceramics were prepared by the hot press sintering method. The XRD results show that the second phase (Nb_1-xTa_x)C is formed when the Ta content increases to 25 mol%. The SEM micrographs show that (Nb_1-xTa_x)C has a core/rim structure, whose formation mechanism was also investigated. Substituting some Ta for Nb can significantly improve the mechanical properties of Nb₄AlC₃. (Nb_0.75Ta_0.25)₄AlC₃ exhibits an excellent fracture toughness of 8.3 ± 0.3 MPa m^1/2 at room temperature (RT). The highest Young's modulus (349 ± 16 GPa) and Vickers hardness (4.5 ± 0.3 GPa) at RT are exhibited by the (Nb_0.5Ta_0.5)₄AlC₃ sample, which correlate to increases of 18% and 80%, respectively, compared with those of Nb₄AlC₃. The flexural strengths of (Nb_0.5Ta_0.5)₄AlC₃ are 439 ± 18 MPa at RT and 344 ± 22 MPa at 1100 °C, which correlate to increases of 27% and 45%, respectively, compared with those of Nb₄AlC₃. The solid solution of Ta and the formation of (Nb_1-xTa_x)C are beneficial to the strengthening of Nb₄AlC₃. The coefficient of thermal expansion (CTE) increases slightly from 7.08 × 10⁻⁶ K⁻¹ for Nb₄AlC₃ to 7.24 × 10⁻⁶ K⁻¹ for (Nb_0.75Ta_0.25)₄AlC₃ at 25–1400 °C. The thermal conductivity of (Nb_0.75Ta_0.25)₄AlC₃ (28.4–29.8 W/m·K) is higher than that of Nb₄AlC₃ (18.1–21.2 W/m·K) over the whole test range (25–1000 °C). Owing to their excellent mechanical and thermal properties, Ta-doped Nb₄AlC₃ ceramics have good potential as structural materials.     

Motion control of underactuated linkage robot based on gymnastic skill

In this paper, a reproduction of a swing‐up and a giant swing motion of underactuated robots based on technique of the horizontal bar gymnast is discussed with focusing on an equivalent center of mass (ECM) of underactuated robots and the gymnast. At first, the behaviors of the ECM of the gymnast (ECMG) are analyzed by using a motion capturing technique and an efficient motion of the ECMG for the swing‐up and for the giant swing motion are identified from analysis results, respectively. Next, a partial linearization method, which can realize that the ECM of the Acrobot (ECMA) replicates this efficient motion, is designed and reproduces the underactuated robots the swing‐up and giant swing motion like the gymnast. Finally, an effectiveness of the proposed controller is shown by numerical simulations.     

Prediction of cooling efficiency of forced-air precooling systems based on optimized differential evolution and improved BP neural network

This paper proposes an approach to predict the efficiency of forced-air cooling of fresh apples that combines the optimized differential evolution (DE) algorithm and the back-propagation (BP) neural network algorithm. First, to balance population diversity and fast convergence, the individual mutation operation of the basic DE algorithm was optimized by dividing the entire population into two equal parts according to the fitness value of individuals, and DE-best-1 and DE-current-to-rand-1 are used as individual mutation operations for the superior- and inferior-part individuals, respectively. Moreover, the selection operation of basic DE was also changed by using a crowding scheme, which helps maintain population diversity and discover more regions containing the global optima. Second, an optimized DE-BP neural network model was established by using the optimized DE to determine the initial weights and thresholds of the BP neural network to avoid being trapped in local minima, following which the effect of input parameters on the network output was subjected to a comprehensive sensitivity analysis based on the trained neural network. The results show that the optimized DE-BP model accurately predicts the efficiency with which apples are cooled. Furthermore, the airflow velocity and total opening area have a significant negative correlation with the average apple temperature and a positive correlation with the cooling rate of the apples. Finally, the most important factor influencing the cooling efficiency of the pre-cooling system is the total opening area of the ventilated packaging.     

Protection through property: from private to river-held rights

ABSTRACT

This article explores how private owners can protect bodies of water through private property rights. It compares the use of conventional property rights in the Netherlands and New Zealand with a novel approach whereby a New Zealand river owns itself.     

Tunable Fe/N co-doped 3D porous graphene with high density Fe-Nx sites as the efficient bifunctional oxygen electrocatalyst for Zn-air batteries

Nitrogen-doped transition metal materials display promising potential as bifunctional electrocatalysts for oxygen reduction reaction (ORR) and oxygen evolution reaction (OER). Herein, Fe/N co-doped three-dimensional (3D) porous graphene (FeN-3D-PG) is prepared via a template method using sodium alginate as the carbon source and low polymerization degree melamine resin as the nitrogen source. The low polymerization degree melamine resin can form complexes with Fe³⁺ in the aqueous solution and further forms high density Fe-N_x active sites during pyrolysis. Meanwhile, the formed 3D porous structure efficiently promotes the uniform distribution of Fe-N_x active sites. The FeN-3D-PG catalyst exhibits pH-independent ORR activity. For OER, the catalyst possesses a low over potential (370 mV at 10 mA cm⁻²) in alkaline electrolyte. The Zn-air batteries (ZABs) using FeN-3D-PG as cathode exhibits a power density up to 212 mW cm⁻², a high specific capacity of 651 mAh g⁻¹, and the charge-discharge stability of 80 h. This work provides new sight to transition metal materials based ZABs with excellent performance.     

A dinuclear iron complex as a precatalyst for water oxidation under alkaline conditions

Water oxidation is a key reaction for water splitting. The decomposition of Fe-based-molecular structures toward Fe-based oxides is a promising method for oxygen-evolution reaction (OER) through water oxidation. The decomposition of Fe-based-molecular structures method results in a slow decomposition of precatalysts and forms Fe oxide-based catalysts. In this study, the Fe species formed through the decomposition of a dinuclear Fe(III) complex under OER is investigated by X-ray photoelectron spectroscopy, scanning electron microscopy, energy dispersive spectrometry, X-ray diffraction, and the electrochemical method. In addition, using Ni(OH)₂, a new approach is reported for detecting trace Fe species on the electrode surface. The resulting Fe oxide-based catalyst shows a catalytic current with an onset of 621 mV overpotential and the Tafel slope of 113.7 mV/decade at pH 11 in a buffer phosphate.