Search

Time-series 1-D signals are ubiquitous in industrial applications for monitoring and control. However, it is lacking of efficient tools to deal with simultaneously multiple time-series 1-D signals. To this end, in this article, a novel theory of image formation is proposed that converts multiple 1-D signals to 2-D images and takes advantages of convolutional neural network for feature extraction and classification of a sequence of images. A case study is carried out for the classification of working conditions in photovoltaic power systems. In total, 23 1-D signals are mapped to a sequence of 2-D images to derive six different models through image formation-based deep learning. They are tested through the outdoor experiments under time varying working conditions. We discover that physical implication in 2-D images affects significantly the classification performance such that 2-D images with the clustered currents or voltages tend to create better results while randomly arranged image patterns are prone to generate worse results. Excellent performance with an accuracy 96.09% is guaranteed when physical advantages are incorporated in the proposed tools. Driven by the deep learning approaches, the proposed tools are promising for complicated industrial applications.

The cutting-edge monitoring technique of photovoltaic power systems tends to employ cloud servers for big data analysis in the wake of smart sensors and internet of things (IoT). Various mismatch phenomena of photovoltaic arrays are generated due to the external and internal interactions, which would be identified by an appropriate monitoring approach. However, high dimensional sequential data provided by multiple sensors challenges the existing monitoring technologies. This paper proposes a dimension reduction technology mapping multiple sequence signals to a sequence of images which are processed further by a convolutional neural network (CNN), resulting in a novel condition monitoring system for photovoltaic array systems. Firstly, multiple sources of 1-dimensional time-series data are rearranged to construct 2-dimensional time-series images. Then, the CNN algorithm automatically extracts the underlying graphical features from data of 2-dimensional images for condition monitoring. Experiments were carried out upon self-made solar power stations to verify the effectiveness of the proposed method for real-world solar power systems. It shows that the data driven approach could identify effectively key operation conditions from the historical data with a negligible loss of features at the presence of mismatched phenomena.

Unprecedented mismatch occur frequently in the photovoltaic (PV) array systems, challenging the classical monitoring systems. However, the study of artificial intelligence methods lacks some well-designed experiments for a systematic verification because mismatch are influenced by many aspects such as the PV array, the DC–DC converter, the DC–AC inverter, the loads as well as the environmental variables. The objective has two folds. One is to design some identical apparatuses to emulate the ‘same’ real-world solar power station that is operated under multiple mismatch for a controllable, repeatable and comparable experimental study. Another is to propose a novel condition monitoring strategy based on the backward propagation neural network and a decision-making formula. Some controllable indoor experiments and uncontrollable outdoor experiments are carried out to verify the ideas. Such work has never been studied before. The experimental results show that the derived monitoring systems identify and classify accurately different mismatch in both indoor tests and outdoor experiments. Moreover, real-time experimental results infer that the data-driven approach with the self-learning capabilities is adaptive to the environmental changes. Therefore, the supervised learning-based condition monitoring strategy is promising in the solar power industry in terms of operation management and performance enhancement.

Photovoltaic array systems operated in harsh and versatile environments generate complex nonlinear characteristics due to various mismatched phenomena, which undermines the system performance unprecedentedly and challenges the existing monitoring technologies. To this end, a class of modern data acquisition and smart condition monitoring systems are introduced with the advanced technologies such as sensor networks, smart combiner boxes and smart inverters. Furthermore, a novel condition monitoring approach for identifying and classifying mismatched photovoltaic arrays is proposed that consists of pre-filter, Elman neural network and decision-making rules to deal with various mismatches such as partial shading due to tree leaves or dirt dropping, and open circuit faults. Novel micro solar power stations are built and used to prove the concept via a comparable experimental study that has never done before. Experimental results show that applying pre-filters on time series input attributes improves the performance up to 3.70% and modifying the decision making formulae improves further the performance up to 1.78%. Eventually, the entire network could identify all of fault conditions if the parameters are chosen properly. Therefore, the proposed method is promising in solar power systems and the constructed apparatuses are useful in the study of modern photovoltaic power systems.

When a high-speed droplet impacts on mesh screens, part of the droplet penetrates the screen through its pores and generates smaller secondary drops, which spray downstream in a conical distribution. This instantaneous phase fragmentation phenomenon has been widely utilized in liquid spray applications and multiple-phase liquid separation. During droplet deformation, the intense liquid–gas fragmentation can lead to high nonequilibrium effect, which makes it hard to simulate by traditional fluid computational method. In this study, for the first time, we provided a numerical method to simulate the entire process of penetration dynamic behaviors. This 3D droplet-impact model based on MDPD (many-body dissipative particle dynamics) method exhibits high stability. A special solid–liquid boundary condition was proposed and successfully reduced the massive computational resources wasted on the solid mesh surface. To verify our model, the impacting of a droplet on a flat surface and on a mesh screen were simulated, respectively. The result showed a good match with our previous drop impact study and our experiment of the whole process about a droplet fragmented into hundreds of small drops. We further studied the mass transfer ratio (the ratio of penetrated drops to the initial droplet) and the ejection angle (the angle of the spray cone). The mass transfer ratio and ejection angle can be approximated as a function of Weber number, solid fraction and mesh number by summarizing the regular drop-penetrated behaviors over initial speed and mesh number.

AlCoCrFeNiTi high-entropy alloy coatings (HEACs) were prepared by mechanical alloying (MA) and laser cladding (LC) process on H13 hot-working die steel substrate. Phase evolution, microstructure, and mechanical properties of the alloyed powder and HEACs were investigated in detail. The final milling AlCoCrFeNiTi coating powders exhibited simple body centered cubic (BCC) phase and mean granular size of less than 4 &mu, m. With the increase of heat input of the laser, partial BCC phase transformed into minor face centered cubic (FCC) phase during LC. AlCoCrFeNiTi HEACs showed excellent metallurgical bonding with the substrate, and few defects. Moreover, the microhardness of AlCoCrFeNiTi HEACs reached 1069 HV due to the existence of the hard oxidation and the second phase grains, which are about five times that of the substrate. The laser surface cladding HEACs exhibited deteriorated tensile property compared with that of the substrate and the fracture generally occurred in the region of HEACs. The fracture mechanism of AlCoCrFeNiTi HEACs was dominated by the comprehensive influence of brittle fracture and ductile fracture.

Background PS916, chitosan derivative with shown activities in atherosclerotic and fatty liver, is being investigated as an anti-atherosclerotic agent in clinical trials in China. Methods Fluorescein-labeled PS916 (PS916-FTC) was prepared by the reaction with fluorescein isothiocyanate. The pharmacokinetics and bio-disposition of PS916-FTC were studied in rats after oral or intravenous administration. Results Analysis of the plasma, urine, fecal and tissue samples collected at intervals up to 72 h revealed that PS916-FTC exhibited moderate volume of distribution (Vss, 0.650~0.748 L/kg), and low clearance (60.9~107 mL/h/kg) after intravenous administration. The pharmacokinetics of PS916-FTC was characterized by low bioavailability (8.40%) after oral administration. The average accumulation ratio for PS916-FTC exposure after steady-state administration was 1.04. A two-compartmental pharmacokinetics model was employed. The urinary route was the major pathway (54.4%), and the fecal route was a minor pathway (6.29%) for PS916-FTC elimination after intravenous administration; the fecal route was the major pathway (79.0%) for PS916-FTC elimination after oral administration. Conclusion PS916-FTC was widely distributed to most tissues in rats; relatively high levels of PS916-FTC in kidney and liver were observed after intravenous or oral administration. These findings supported the understanding of pharmacokinetics and bio-disposition of PS916 in rats and provide relevant information for future design of clinical studies. Highlights 1) Fluorescein-labeled PS916 was successfully synthesized. 2) A rapid and sensitive analytical method of PS916-FTC was validated. 3) The pharmacokinetic of PS916-FTC in rats was investigated. 4) The bio-distribution of PS916-FTC in rats was investigated.