Accurate ensemble pruning with PL-bagging

Ensemble pruning deals with the selection of base learners prior to combination in order to improve prediction accuracy and efficiency. In the ensemble literature, it has been pointed out that in order for an ensemble classifier to achieve higher prediction accuracy, it is critical for the ensemble classifier to consist of accurate classifiers which at the same time diverse as much as possible. In this paper, a novel ensemble pruning method, called PL-bagging, is proposed. In order to attain the balance between diversity and accuracy of base learners, PL-bagging employs positive Lasso to assign weights to base learners in the combination step. Simulation studies and theoretical investigation showed that PL-bagging filters out redundant base learners while it assigns higher weights to more accurate base learners. Such improved weighting scheme of PL-bagging further results in higher classification accuracy and the improvement becomes even more significant as the ensemble size increases. The performance of PL-bagging was compared with state-of-the-art ensemble pruning methods for aggregation of bootstrapped base learners using 22 real and 4 synthetic datasets. The results indicate that PL-bagging significantly outperforms state-of-the-art ensemble pruning methods such as Boosting-based pruning and Trimmed bagging.     

Yield improvement and yield variability reduction in an EOQ model with planned shortages and random yield

This paper reports the results of an investigation of a set of continuous time, constant demand inventory models under the condition of yield uncertainty. Specifically, the impact of yield improvement programs on lot size, backorder level, and the resulting costs are examined. Models for improving yield rate and reducing yield variability are developed and examined through a series of numerical exercises. In addition, a model for the simultaneous improvement of yield rate and yield variability is presented for the case where there is a relationship between the mean and variance of the yield distribution. In all cases, investment programs improve the picture with respect to manufacturing yield for processes which are not necessarily under statistical control.     

A stochastic model for localized disturbances and its applications

A stochastic model for local disturbances, particularly for a temporal harmonic with random modulations in amplitude and/or phase, is proposed in this paper. Results for the second moment responses of a linear single-degree-of-freedom system to this type of stochastic loading demonstrate a significant change in response characteristics due to a small uncertainty. A local phenomenon may last much longer and resonance may be smeared to a broad range. Integrated with wavelet transform, the proposed approach may be used to model a random process with non-stationary frequency content. Especially, it can be effectively used for Monte Carlo simulation to generate large size of samples that have similar characteristics in time and frequency domains as a pre-selected mother sample has. The technique has a great potential for the case where uncertainty study is warranted but the available samples are limited.     

丙烯腈－苯乙烯共聚物水解磺化产物对泥浆的稳定作用

在实验室条件下对用丙烯腈－苯乙烯共聚物在硫酸存在下的水解磺化产物（ＨＳＡＳ）处理的泥浆性能进行了评定。ＨＳＡＳ处理的泥浆有较小的滤失性，较好的耐盐和抗高温性。     

Probabilistic analysis of a static frame model

This paper describes our efforts during our participation in the Sandia Validation Workshop. The focus of the paper is the calibration of material models and simulation of random fields to characterize the variations of material properties across spatial field. Both parametric and non-parametric methods were used to represent uncertainty. Part of the challenge of this problem is the small amount of data that is available for the necessary probabilistic analyses in support of calibration, validation, accreditation and prediction activities. The analysis methods and corresponding results are described.     

Cycle detection using a stack

We present an algorithm for detecting periodicity in sequences produced by repeated application of a given function. Our algorithm uses logarithmic memory with high probability, runs in linear time, and is guaranteed to stop within the second loop through the cycle. We also present a partitioning technique that offers a time/memory tradeoff. Our algorithm is especially well suited for sequences where the cycle length is typically small compared to the length of the acyclic prefix.     

Response level crossing rate of a linear system excited by a partially specified Gaussian load process

The problem of determining the maximum mean response level crossing rate of a linear system driven by a partially specified Gaussian load process has been considered. The partial specification of the load is given only in terms of its total average energy. The critical input power spectral (PSD) function, which maximizes the mean response level crossing rate, is obtained. The critical input PSD turns out to be highly narrow-banded which fails to capture the erratic nature of the excitation. Consequently, the trade-off curve between the maximum mean response level crossing rate and the maximum disorder in the input process, quantified in terms of its entropy rate, has been generated. The method of Pareto optimization is used to tackle the conflicting objectives of the simultaneous maximization of the mean response level crossing rate and the input entropy rate. The non-linear multi-objective optimization has been carried out using a recently developed multi-criteria genetic algorithm scheme. Illustrative example of determining the critical input of an axially vibrating rod, excited by a partially specified stationary Gaussian load process, has been considered.     

Study of direct and indirect naphtha recycling to a resid FCC unit for maximum propylene production

To satisfy the increasing propylene demand, direct and indirect naphtha recycling schemes around an existing resid fluid catalytic cracking (FCC) unit were investigated. To this aim, light cracked naphtha (LCN), heavy cracked naphtha (HCN) and a PolyNaphtha (PN) oligomerisation product were cracked under a wide range of operating conditions over a commercial Y zeolite based equilibrium catalyst. Experimental data were acquired in three different units: a fixed bed bench scale unit, a fixed fluidised bed unit and an adiabatic circulating fluidised bed pilot plant. It was shown that FCC naphthas require high operating severities to crack, and that even then their conversion remains relatively moderate. Hence, direct recycling to the main riser does not seem a viable pathway to increase propylene product. Feeding FCC naphthas to a second reaction zone operating at high severity allows to increase the propylene yield in a significant manner. Increasing conversion, however, not only leads to higher LPG and propylene yields, but also results in very high dry gas yields. An alternative scheme was proposed, in which the olefinic C₄ and C₅ fractions are converted into a naphtha fraction through oligomerisation in a dedicated unit before being recracked in the secondary riser. As the highly olefinic oligomerised effluent mainly consist of dimerised and trimerised butenes and pentenes, this feed is more easily cracked and high conversions can be achieved. This indirect interconversion of butenes and pentenes into propylene therefore effectively allows to convert these butenes and pentenes into propylene, resulting in a significant increase in propylene yield. Each of the three main naphtha recycle options (directly to the main riser, directly to a secondary riser or indirectly via a light olefin oligomerisation unit) have been analysed and compared to a base case. In the evaluation of each of these schemes, all heat balance effects, both on the riser and the regenerator side, have been accounted for. The proposed process scheme with an indirect recycle via an oligomerisation unit enhances the already inherent flexibility of the FCC unit. The naphtha recycle can be turned on or off, the second reaction zone can be used to crack naphtha or to crack resid feed to maximise throughput, while the effluent of the oligomerisation unit can be recycled to the FCC unit for propylene production or hydrogenated and sent to gasoline and kerosene pool.     

Comparison of finite element reliability methods

The spectral stochastic finite element method (SSFEM) aims at constructing a probabilistic representation of the response of a mechanical system, whose material properties are random fields. The response quantities, e.g. the nodal displacements, are represented by a polynomial series expansion in terms of standard normal random variables. This expansion is usually post-processed to obtain the second-order statistical moments of the response quantities. However, in the literature, the SSFEM has also been suggested as a method for reliability analysis. No careful examination of this potential has been made yet. In this paper, the SSFEM is considered in conjunction with the first-order reliability method (FORM) and with importance sampling for finite element reliability analysis. This approach is compared with the direct coupling of a FORM reliability code and a finite element code. The two procedures are applied to the reliability analysis of the settlement of a foundation lying on a randomly heterogeneous soil layer. The results are used to make a comprehensive comparison of the two methods in terms of their relative accuracies and efficiencies.