The reliability of a series system of repairable subsystems: A bayesian approach

Consider the problem of estimating the reliability of a series system of (possibly) repairable subsystems when test data and historical information are available at the component, subsystem, and system levels. Such a problem is well suited to a Bayesian approach. Martz, Waller, and Fickas [Technometrics, 30 , 143–154 (1988)] presented a Bayesian procedure that accommodates pass/fail (binomial) data at any level. However, other types of test data are often available, including (a) lifetimes of nonrepayable components, and (b) repair histories for repairable subsystems. In this article we describe a new Bayesian procedure that accommodates pass/fail, life, and repair data at any level. We assume a Weibull model for the life data, a censored Weibull model for the pass/fail data, and a power-law process model for the repair data. Consequently, the test data at each level can be represented by a two-parameter likelihood function of a certain form, and historical information can be expressed using a conjugate family of prior distributions. We discuss computational issues, and use the procedure to analyze the reliability of a vehicle system. © 1994 John Wiley & Sons, Inc.     

Sole sourcing versus dual sourcing under stochastic demands and lead times

The use of a single vendor for each inventoried item is usually assumed in most of the inventory models. However, there are situations where the use of more than one vendor should be considered, especially when lead times are stochastic. This research presents a theoretical investigation of the effect of cost structures on the relative performance of sole-sourcing versus dual-sourcing inventory control policies. We show that except for cases where the ordering cost is high, the lead-time variability is low, or the customer service level is low, dual sourcing performs better than sole sourcing under the normally distributed demand and shifted-exponential lead times. Moreover, the computational results indicate the dual sourcing provides a better service level than sole sourcing at the optimal solutions, and that dual sourcing results in larger order quantities than sole sourcing, which suggests that attractive quantity discounts may not be in jeopardy when dual sourcing is employed. Finally, because it is generally known that multiple sourcing can enhance the competition among suppliers, material managers should consider splitting purchase orders when two equally qualified suppliers are available. © 1994 John Wiley & Sons, Inc.     

Mixture selection for maximal stochastic performance

Methodology for the optimal selection of a mixture of components for the attainment of a required performance level is developed. This accommodates both cost and stochastic variation and is applied to gasoline blending. © 1994 John Wiley & Sons, Inc.     

Properties of a three-dimensional estimator

A large-scale simulation was conducted to investigate the statistical properties and the suitability of an estimator that estimates the coordinates of a three-dimensional source, according to the angles at which several observers view the source. The estimator uses a heuristical averaging method of estimates obtained from the solution of six two-dimensional problems that utilize line-to-point transformations. The estimator was found to be approximately unbiased with low standard deviations for normal error distributions. An effective and easy-to-use forecasting formula was developed to forecast the parameters of the distributions of the estimates. © 1994 John Wiley & Sons, Inc.     

Estimating moments of the minimum order statistic from normal populations—a bayesian approach

Let Yⁱ ∽ N(θⁱ, σ), i = 1, …, p, be independently distributed, where θⁱ and σ are unknown. A Bayesian approach is used to estimate the first two moments of the minimum order statistic, W = min (Y¹, …, Y^p). In order to compute the Bayes estimates, one has to evaluate the predictive densities of the Yⁱ's conditional on past data. Although the required predictive densities are complicated in form, an efficient algorithm to calculate them has been developed and given in the article. An application of the Bayesian method in a continuous-review control model with multiple suppliers is discussed. © 1994 John Wiley & Sons, Inc.     

Employment and utilization of engineers among defense contractors

Defense contractors retain engineers for two reasons. First, engineers work on current-period production contracts. Second, engineers conduct independent research and development (IR&D) and write bids and proposals (B&P) for future contracts. The amount of IR&D/B&P activity, however, understates the amount of effort devoted to learning that reduces the cost of future production. Even some engineers who work on current-period production may engage in learning, as a by-product of their efforts. This article develops and implements a model for estimating the volume of learning activity undertaken by four large manufacturers of military aircraft. © 1994 John Wiley & Sons, Inc.     

Improvement of VIS and IR camouflage properties by impregnating cotton fabric with PVB/IF-WS2

In order to examine the possibility to improve its camouflage properties standard cotton fabric with camouflage print was impregnated with poly(vinyl butyral), PVB and fullerene-like nanoparticles of tungsten disulfide, PVB/IF-WS₂. FTIR analysis excluded any possible chemical interaction of IF-WS₂ with PVB and the fabric. The camouflage behavior of the impregnated fabric has been examined firstly in the VIS part of the spectrum. Diffuse reflection, specular gloss and color coordinates were measured for three different shades (black, brown and dark green). Thermal imaging was applied to examine the camouflage abilities of this impregnation in IR part of the spectrum. The obtained results show that PVB/IF-WS₂ impregnation system induced enhacement of the materials camouflage properties, i.e. that IF-WS₂ have a positive effect on spectrophotometric characteristics of the fabric.     

A platform for initial testing of multiple camouflage patterns

The human visual system is still an important factor in military warfare; military personnel receive training on effective search strategies, and camouflage that can effectively conceal objects and personnel is a key component of a successful integrated survivability strategy. Previous methods of camouflage assessment have, amongst others, used psychophysics to generate distinctiveness metrics. However, the population from which the human observers are drawn is often not well defined, or necessarily appropriate. In this experiment we designed a new platform for testing multiple patterns based on a camouflaged object detection task, and investigate whether trained military observers perform better than civilians. We use a two-alternative forced choice paradigm, with participants searching images of woodland for a replica military helmet displaying Olive Green, Multi Terrain Pattern, US Marine Pattern or, as a conspicuous control, UN Peacekeeper Blue. Our data show that there is no difference in detection performance between the two observer groups but that there are clear differences in the effectiveness of the different helmet colour patterns in a temperate woodland environment. We conclude that when tasks involve very short stimulus presentation times, task-specific training has little effect on the success of target detection and thus this paradigm is particularly suitable for robust estimates of camouflage efficacy.     

Real-time prediction of projectile penetration to laminates by training machine learning models withfinite element solver as the trainer

Studies on ballistic penetration to laminates is complicated, but important for design effective protection of structures. Experimental means of study is expensive and can often be dangerous. Numerical simu-lation has been an excellent supplement, but the computation is time-consuming. Main aim of this thesis was to develop and test an effective tool for real-time prediction of projectile penetrations to laminates by training a neural network and a decision tree regression model. A large number of finite element models were developed;the residual velocities of projectiles fromfinite element simulations were used as the target data and processed to produce sufficient number of training samples. Study focused on steel 4340tpolyurea laminates with various configurations. Four different 3D shapes of the projectiles were modeled and used in the training. The trained neural network and decision tree model was tested using independently generated test samples using finite element models. The predicted projectile velocity values using the trained machine learning models are then compared with thefinite element simulation to verify the effectiveness of the models. Additionally, both models were trained using a published experimental data of projectile impacts to predict residual velocity of projectiles for the unseen samples. Performance of both the models was evaluated and compared. Models trained with Finite element simulation data samples were found capable to give more accurate predication, compared to the models trained with experimental data, becausefinite element modeling can generate much larger training set, and thus finite element solvers can serve as an excellent teacher. This study also showed that neural network model performs better with small experimental dataset compared to decision tree regression model.