Search

With the increasing density of flash memory, its service life is declining. As the most important technical index of a flash device, flash retention parameters often need to consume a substantial amount of time to be tested. With a view toward reducing the test period, the most popular method currently is to speed up the retention loss of flash memory at high temperatures. However, since the retention loss of flash memory is the result of a mixture of multiple failure mechanisms, the Arrhenius model with a single apparent activation energy (Eaa) has significant limitations in calculating the high-temperature acceleration factor for the retention loss of flash memory. In this paper, based on a large number of real experiments, we have studied the retention characteristics of the flash memory in high-temperature environments and closely tracked the change of Eaa at different temperatures, retention time and program/erase (P/E) cycle count. Experimental results show that the equivalent acceleration factor (AF) grows rapidly with retention time at the beginning of flash retention and that flash with a high P/E cycle count has a faster AF growth rate.

Accelerators for machine learning (AI) inferencing applications are homogeneous designs composed of identical cores. Each core, or processing element (PE), contains multiply-and-accumulate units, control logic, and registers for storing and forwarding weights and activations. Testing homogeneous array-based AI accelerator chips by running automatic test pattern generation (ATPG) at the array level results in a high CPU time and pattern count. We propose a constant-testable (C-testable) method for test generation at the PE level such that the ATPG effort does not increase with the number of PEs. Our results show that, compared to the traditional array-level testing, the proposed method achieves up to 4.2× (3.5×), 1530× (2388×), and 170× (142×) reduction in the test pattern count, ATPG runtime, and test cycle count, respectively, for stuck-at (transition) faults in a 256×256 array, while preserving the test coverage. A reconfigurable scan architecture is introduced to enable the proposed C-testable solution for the entire accelerator array. The design-space exploration of a hierarchical test-compaction framework is presented. We also describe four debug solutions for fault localization and diagnosis.

Circular resource use in agriculture and food systems could play an important role when aiming for sufficient food output with limited environmental impact and resource depletion. Circularity, however, is not a goal in itself. With respect to nutrient use and emissions, agricultural system sustainability is currently commonly assessed by nutrient output/input ratio (O/I, nutrient use efficiency) or surplus per ha (I–O). OBJECTIVE: Our aim is to assess how these sustainability indicators are related to nutrient cycling. METHODS: Starting from basic circularity concepts, a set of equations (frame) is presented that relates nitrogen (N) and phosphorus (P) cycling to food product output, or to food use by human consumers. Circularity indicators express how many times a nutrient input cohort completes a full cycle (CyCt), or passes through the system's top trophic compartment (UseCt). Examples of such compartments are the crop (arable systems), the herd (livestock farms), and the human population (regional food systems). UseCt governs export in useful product. The frame allows to predict equilibrium O/I from system properties, and to attribute parts of O/I to direct (linear) and cycled flow. CyCtR quantifies how many cycles could be completed by nutrients in absence of product export. CyCtR allows to assess the efficacy of returning waste from exported products. Above indicators are compared against Finn cycling index and Figge circularity index, more commonly used in ecological and industrial research respectively. All indicators are calculated for systems of increasing complexity: (i) a UK wheat field, (ii) a Dutch dairy farm, and (iii) the Flanders regional food system. Their responses to changes in system properties are analysed for examples ii and iii. RESULTS AND CONCLUSIONS: Nutrient flows in UK arable field and Flanders are almost linear. In UK arable field, O/I equals 0.74 (N) and 0.66 (P), with small contributions from cycled flow (9% for N, 5% for P). In Flanders, cycled flow constitutes only 2% of total N and P flows that reach the human consumer in Flanders. The dairy farm shows largest contributions of cycled flow: 35% (N) and 60% (P) of O/I comes from cycled flow, but O/I itself is only 0.28 (N) and 0.72 (P). SIGNIFICANCE: The presented frame allows to assess the impacts of system changes on productivity, nutrient cycling, resource use and nutrient emissions. This is useful for ex-ante assessment of measures that reduce nutrient losses from the system or increase the retrieval of external waste flows.

Full-span overhead sign support structures can be found along any major highway across the US. Such structures experience different wind loading scenarios varying with time. As a result, these structures start to build up fatigue cracks within their members near the end of their fatigue life. Due to economic realities, the needed routine fatigue inspections on such highway structures can not be performed regularly. This paper is intended to present a comprehensive tool to accurately predict the remaining fatigue life of full-span overhead highway sign support structures subjected to a long and sustained wind fluctuation. Synthetic wind time histories were developed by superimposing cosine waves over a range of frequencies of 3–300 Hz and randomly generated phase angles. Kaimal spectrum was utilized to build a database of wind time histories for each daily mean wind speed along a period of 45 years in the State of Kansas. Moreover, each wind time history was modified to capture both the mean speed and high speed in each given day. After that, the wind speed vs. the number of cycle relationship, for a given time span, was extracted from the synthetic wind time history using the Rain Flow counting technique. Fatigue evaluations were conducted using axial truss member stresses extracted from a finite element solution corresponding to each wind speed in any given time range. Potential fatigue failure was assessed for each structural member after amplifying the stress range using an average dynamic amplification factor generated by integrating the frequency-response curve of harmonic excitations. These assessments evaluate the ratio of consumed fatigue cycles to ultimate fatigue cycles using Miner’s rule to estimate the fatigue life. A computationally affordable simulation package was developed to carry out the generation of wind time histories, cycle counting, structural modeling, and fatigue life calculations. This package was used to evaluate the fatigue life of a non-cantilever sign structure in Wichita, Kansas. The software predicted the end-of fatigue life of two members in this structure. Accordingly, inspections of these two members revealed the existence of unnoticed severe fatigue cracks while other members did not show any sign of distress.

This study suggests a novel investment strategy for sizing a supercapacitor in a Battery Energy Storage System (BESS) for frequency regulation. In this progress, presents hybrid operation strategy considering lifespan of the BESS. This supercapacitor-battery hybrid system can slow down the aging process of the BESS. However, the supercapacitors are relatively expensive and have a self-discharging rate of nearly 40% per day. Therefore, finding the optimal size of a supercapacitor contributes to prolonging the lifespan of the BESS at an effective cost. In the proposed operation strategy, this paper estimates the lifespan at the given supercapacitor size using an innovative variable capacity model (VCM). Because this model does not use the current and voltage characteristics to calculate the lifespan of the BESS and uses measured lifespan data sets, this method has the advantages of making quick investment decisions. It is shown through simulation that the VCM gives more realistic results than the conventional fixed capacity model estimating a battery lifespan with achievable cycle count curves. Moreover, the net benefits of the supercapacitor were measured through benefit and cost analysis and net present value analysis for the supercapacitor size in the BESS. Through this paper, frequency regulation service is considered as the application target for the proposed strategy.

The goal of this study is to assess the economics of deploying rechargeable batteries for simultaneous demand peak shaving and price arbitrage business. To this end, we present a novel modeling approach for investment and operational decision-making, which supports the selection of battery technology and size among different Li-ion batteries and optimizes the scheduling of battery charge and discharge. Battery degradation incurring due to repeated charging and discharging is accounted for by implementing the Rainflow cycle counting algorithm in an energy-economic modeling framework for operational dispatch optimization. This allows for an accurate representation of cycle battery degradation in the optimization model. Simulation results obtained for a Swiss electricity service provider show that battery deployment is currently only economically viable if the sizes of energy capacity and power units are selected judiciously. Furthermore, our case-specific findings reveal that the cycle lifetime of a battery has only a minor impact on the achievable life cycle revenues, which can be attributed to the fact that a small number of charge and discharge cycles is sufficient to exploit the most profitable revenue opportunities. Based on the conducted case study, we argue that second-use batteries could be particularly appealing for stationary energy storage.

The method is proposed to generate random processes with stable integral characteristics of cumulative distributions, which at the same time are random in nature. The algorithm consists of the two steps. The first one is generation of the random extrema sequence using the objective Markovian matrix, completed with in-service loading data. The extrema and their sequence are shown to be of importance for the metal fatigue assessment. Therefore, random testing with extrema control is more advantageous as compared to testing with the generation of spectral densities. The next step is the random process generation with its first continuous derivative. The concatenation procedure was proposed to create the continuous function reconstruction with the pointwise specification (local maxima and minima sequence). The regression analysis of two random variables was used to evaluate the frequency component. Spectral analysis of such a continuous process results in the spectral density function, which finds its application in some methods of random fatigue estimation. The major trait of the proposed method is the coincidence of the rain-flow cycle counting characteristics of the initial and resulting processes.

PurposeThe purpose of this scientific work is to simulate the fatigue damage under random loading, taking into account the mean stress effect on fatigue lifetime and using the Rainflow counting technique to assess the fatigue damage by the Ansys software. The used material is aluminum alloy 6082-T6. A comparison with literature results has confirmed this investigation in this paper.Design/methodology/approachThe study of fatigue under random loading is based on the same concepts as constant loading with the addition of damage summation. The proportion of damage caused by a stress cycle depends not only on the alternating stress but also on the mean stress.FindingsAnalysis of the fatigue damage shows that the number of relative damage due to each cycle.Originality/valueThis paper aims to simulate the fatigue damage under random loading for aluminum alloys.

The article presents the results of the research of dependability of mobility assets (MA) of weapon and military equipment (WME) according to operating hours and energy consumption which allow to control their operational reliability. The subject of the study is the process of investigation of operational reliability of mobility assets of of weapon and military equipment. The purpose of the article is theoretical determination of the resource provision by the criterion of a single permissible failure on the basis of the probabilistic model of fatigue resistance of structural elements and data obtained under real operating conditions. The research established a correlation between operating hours and energy consumption of mobility assets of weapon and military equipment and justified the feasibility of planning maintenance and repair according to mileage. The given studies are based on probability-statistical methods of determination of car reliability. The results of the carried out researches showed that the mathematical model of resource provision by the criterion of the only one permissible failure can be obtained on the basis of the probabilistic model of fatigue resistance of structural elements. A possibility to calculate the parameters of the failure flow on the basis of the data obtained under real operation conditions is shown. Thus, the failure flow parameter is calculated with a guaranteed probability of more than 0.90. The provided method allows, provided that the loading cycle count values are known, to determine the reliability of the individual unit (system) and the mobility means as a whole. Determination of the number of load cycles of of MA of WME during use as intended is the subject of further research. The normative maintenance period shall be established on the basis of statistical data obtained under specific operating conditions, on the basis of the labour-intensivity of the one maintenance or one fault elimination. The judgement has been confirmed that as the operating time increases, at the beginning the specific value of the labour-intensity of the current repair increases and then gradually decreases.

Designing resource-efficient deep neural networks (DNNs) is a challenging task due to the enormous diversity of applications as well as their time-consuming design, training, optimization, and evaluation cycles, especially the resource-constrained embedded systems. To address these challenges, we propose a novel DNN design framework called accuracy-and-performance-aware neural architecture search (APNAS), which can generate DNNs efficiently, as it does not require hardware devices or simulators while searching for optimized DNN model configurations that offer both inference accuracy and high execution performance. In addition, to accelerate the process of DNN generation, APNAS is built on a weight sharing and reinforcement learning-based exploration methodology, which is composed of a recurrent neural network controller as its core to generate sample DNN configurations. The reward in reinforcement learning is formulated as a configurable function to consider the sample DNNs' accuracy and cycle count required to run on a target hardware architecture. To further expedite the DNN generation process, we devise analytical models for cycle count estimation instead of running millions of DNN configurations on real hardware. We demonstrate that these analytical models are highly accurate and provide cycle count estimates identical to those of a cycle-accurate hardware simulator. Experiments that involve quantitatively varying hardware constraints demonstrate that APNAS requires only 0.55 graphics processing unit (GPU) days on a single Nvidia GTX 1080Ti GPU to generate DNNs that offer an average of 53% fewer cycles with negligible accuracy degradation (on average 3%) for image classification compared to state-of-the-art techniques.

The results of fatigue tests of samples made of S355JR steel under random tension-compression with nonzero mean stress are presented. The procedure of experimental research is described. The obtained experimental results are presented with the use of Wohler fatigue graphs. The algorithm for the determination of the fatigue life uses, among other things, a rainflow cycle counting procedure, as well as the Palmgren–Miner linear damage accumulation hypothesis and six selected models that take into account the effect of mean stress on the tested durability. We also display the plots for comparison of the experimental and computed lives. It is indicated, which of the considered models describes the impact of the mean stress on the fatigue life of the tested material within the scatter error 3.

This paper proposes a cycle life model for lithium-ion batteries. The main objective of this work is to facilitate the electrical simulation of lithium-ion battery aging (due to cycling), and its impact on battery capacity and internal resistance. Most of the reported cycle life models are either: a) physics based, with parameters difficult to retrieve or b) semi-empirical, where the parameter identification process requires large amount of experimental data, huge manpower and test duration lasting from months to years. Moreover, these models tend to be valid only for the underlying battery and not for other battery types. This makes the simulation of lithium-ion batteries cycling effects, difficult to achieve, expensive and time consuming. The model proposed in this paper is based on simple physical equations from fatigue theory and equivalent cycle counting. The parameter identification process is straightforward and requires only few data from battery datasheets and limited (or short duration) cycling experiments. The proposed model is generic and able to represent the impact of common cycle life factors such as: depth-of-discharge (DoD), temperature and C-rate. The model is validated using two lithium-ion battery types (LFP-LiFePO4 and NMC-LiNiMnCoO2) and simulation results are close to reality with an error within ±1.5% compared to experimental results.