A method is introduced for efficient reliability-based design of laser peening (LP) surface treatment to extend fatigue life of metal components. The method includes nonparametric probability density estimation, surrogate modeling using a new finite element (FE or FEA) approach, and reliability analysis with correlated random variables (RVs). Efficient LP simulation is achieved via a new technique termed single explicit analysis using time-dependent damping (SEATD), which reduces simulation times by a factor of 6. The example study of a three-point bend coupon reveals that fatigue life reliability significantly affects optimal LP design, as 52 laser spots are needed for 99% reliability versus 44 spots for 95%.

Since pioneering work in 1972, the proportional hazard model (PHM) has been widely studied for survival analysis in the area of medicine. Recently, applying the PHM in the area of reliability engineering attracts significant research attentions. In this paper, a two-zone PHM is investigated to predict equipment remaining useful life (RUL) based on the practice that the equipment lifecycle could be divided into two zones: a stable zone and a degradation zone. Results from the numerical experiment illustrate that RUL prediction by applying the proposed two-zone PHM is more accurate and reliable than prediction using the traditional PHM for the entire lifecycle. In practice, this improvement is crucial for real-time maintenance decision making to prevent equipment from catastrophic failures.

Machine failures in manufacturing systems interrupt production operations and cause production loss. The conventional methods for failure prevention are to perform preventive maintenance before failure occurs. In these methods, a fixed maintenance threshold (FMT) is obtained using the lifetime distribution of each machine. This threshold can then be used to trigger maintenance work-orders. A problem with the conventional technique is that it does not consider the updated state of the system, which continues to change before and after maintenance. Therefore, unnecessarily high costs can be incurred due to unexpected equipment failure (lack of maintenance) or excessive maintenance. In this paper, a reliability-based dynamic maintenance threshold (DMT) is calculated based on the updated equipment status. The benefits of the DMT are demonstrated in a numerical case study on a drilling process. The results illustrate that the maintenance policy using the DMT can reduce unscheduled downtime, increase equipment availability, and utilize the equipment remaining useful life more effectively than a conventional FMT-based maintenance policy.

The blank holder force (BHF), which regulates the amount of metal drawn into the die cavity, has been used effectively to draw defect-free sheet parts. It has been shown that a properly designed variable BHF profile can provide greater drawability as compared to constant BHF. Extensive research has been carried out in the past to determine this optimal profile. However, these deterministic efforts neglect the sensitivity of this optimal BHF design to unintended process variations inherent in sheet forming. This paper investigates the reliability of optimal variable BHF in the presence of process uncertainties by minimizing the magnitude of wrinkling and fracture defects under probabilistic constraints. To demonstrate the efficacy of this approach, it analyzes the conical cup drawing of an aluminum killed deep-drawing quality steel under process uncertainties including sheet thickness, punch speed, interface friction at the punch surface, and interface friction at the die surface. The results show that probabilistic design improves the yield (probability of good parts) to 99.98% as compared to the 48.04% yield realized by the traditional deterministic design. In addition, Monte Carlo simulations verify that the probabilistic design handles the uncertainties better due to its inherent robustness. This is primarily due to the design moving away from the process constraints and at the same time balancing the probability of failure between the two failure modes.

This paper analyzes the impact of machining parameter selection on machining system productivity. The method involves prediction of machine tool reliability from mechanical failure based on process choices and production rate. The machine reliability is then input to a machining system production configuration to estimate productivity. Through a simple turning model example, it is demonstrated that an optimal production rate for maximum productivity exists. The best system configuration for maximum productivity is shown to be affected by the level of machine reliability and the choice of process parameters.

Signals from several sensors were employed for real-time laser weld quality monitoring. Sheet-metal butt-joint laser welds of three quality classes (full penetration, partial penetration, gapped) were produced in experimental trials. Optical, air-born acoustic and plasma charge signals acquired during welding were subsequently Fourier-transformed and the spectra were analyzed individually to determine relationships to laser weld quality. The frequency bands most highly correlated to weld quality were identified by stepwise linear discriminant analysis (LDA) of the spectra. Testing of the quality discriminators formulated by LDA of the spectral data showed that the acoustic signal was most reliably correlated with weld quality. Fusing the data from all three sensors prior to LDA analysis produced a discriminator that had about the same reliability as one based on acoustic data alone.

In this paper, a new algorithm, based on a nonlinear optimization method (NOM), has been developed. The accuracy as well as the reliability/robustness of the new algorithm have been verified by applying it to more than 200 CMM measured data sets on differently manufactured parts. The results have been compared with that of Least Squares Method (LSM) and Convex Hull (CVH) method applied to the same data sets. A data filter is proposed to be enclosed in the new algorithm to detect and delete outliers in the data sets. [S1087-1357(00)01102-3]

In a tandem cold-rolling mill, a strip is successively reduced in gauge at each stand as it passes through the mill. The optimal scheduling of a tandem mill is an important but difficult task. In this work, a scheduling problem is considered as an optimization problem which minimizes the total specific power and satisfies certain constraints. Since the material properties and friction coefficient are not known precisely, they are treated as fuzzy numbers. The fuzzy set theory is applied to find out an optimum drafting pattern. The methodology is illustrated by means of a few examples. It is observed that the schedule in which the maximum reduction is achieved in the first pass results in minimum specific power; however, its reliability is poor. Optimization using fuzzy set theory provides a solution which meets the twin requirements of high reliability and minimum power. [S1087-1357(00)00902-3]

A non-contact sensing system has been developed to measure the temperature response of a rapidly rotating roll under spray cooling encountered in metal rolling processes. A thin-film sensor with a fast response time, on the order of microseconds, has been designed and fabricated to measure the roll surface temperatures. To eliminate contact noise, an electro-optical data acquisition system has been developed to provide fast acquiring rate without physical contact to the rotating roll. The system has been implemented on a roll simulator to demonstrate sensor reliability and system feasibility. Also, using the measured temperatures as boundary conditions, a numerical calculation has been performed to provide the temperature distributions of the roll and the corresponding heat transfer coefficients of the spray cooling. Results indicate that the cyclic peak value of the heat transfer coefficient varies from 75 to 100 kW/m 2 -K directly under the impingement of the spray jet. These values are consistent with the existing literature.

This paper introduces a new technique for monitoring multiple tooth rotating tools. The technique is based on laser diffraction. A narrow slit is constructed between the cutting edge of a cutter and a sharp straight blade placed parallel to the cutting edge as a reference. A laser beam is projected onto the slit and consequently a diffraction pattern consisting of a number of fringes is created behind the slit. By comparing the fringe spacing variation, cutting edge wear can be obtained. This technique has high resolution and a wide range of measurement. In addition, the measurement can be correlated to the commonly used flank wear measure (VB). The technique can be used for rotating tools such as milling and drilling. Experimental results in end-milling processes show that the technique is repeatable and accurate. With its fast speed, high resolution and reliability, it is expected that the technique has great potential for industrial applications.

This paper establishes the concept of manufacturing risk assessment based on design features. It also presents a methodology for assessing risks associated with producing feature-rich products with short time-to-market windows. Product design features are classified as inherent and value-added. Inherent features are functionalities of a product that provide a core benefit to the user. Without inherent features, a product cannot exist in the marketplace. Value-added design features augment the customer-perceived value of a product. A product with value-added features provides the manufacturer with a competitive and lucrative edge in the market place. To quantify risks associated with making a product, manufacturing is divided into three subprocesses: reliability, inherent feature augmentation, and value-added tolerance capabilities. Reliability estimates the availability to manufacture the selected product consistently; feature augmentation quantifies risks associated with incorporating inherent design features; and value-added tolerance capabilities measures the consistency in manufacturing value-added design. An example of a leading-edge technology product with a short life cycle is provided. Because of a short time-to-market window coupled with constraints in environmental variables, the bootstrap method is used to estimate key statistical parameters. Once manufacturing risks have been identified and evaluated, the next logical step for future work will be to couple these estimates with uncertainties involved in financing and marketing the product in question.

The cost of feeding parts to a robot for either machine loading or for assembly has been recognized as excessively expensive. The lack of a cost-effective generic part-presenter has prevented the flexibility of the overall manufacturing automation to be fully exploited. For this reason, the design concept of an integrated microprocessor-controlled vision system based on retroreflective vision sensing has been developed at the Material Handling Research Center in Atlanta. Where the orientation of a part can be characterized by its silhouette, outline, or structured “engineering landmarks,” the concept of using retroreflective materials in vision-guided part-presentation system has been proven to have significant potential for improving reliability, reducing computation, and lowering implementation cost. This paper presents optic design concepts for collocated illumination for cost effective part-presentation. The design concepts are illustrated with practical applications and experimental results. The uniformity of the image irradiance across the sensor surface, in general, depends on several factors such as the intensity distribution of the light source, the reflectance characteristics of the object and background, and the geometrical relationship between the imaging sensor, the illumination source, and the target. The trust of this paper focuses on the analytical model which characterizes the influences of these factors for optic design of retro-reflection. This engineering basis is not only necessary for developing practical yet potentially cost-effective optic design concepts, but also provides a useful tool for performance prediction to drive the design for vision-guided part-presentation.

Artificial neural networks are used for on-line classification and measurement of drill wear. The input vector of the neural network is obtained by processing the thrust and torque signals. Outputs are the wear states and flank wear measurements. The learning process can be performed by back propagation along with adaptive activation-function slope. The results of neural networks with and without adaptive activation-function slope, as well as various neural network architectures are compared. On-line classification of drill wear using neural networks has 100 percent reliability. The average flank wear estimation error using neural networks can be as low as 7.73 percent.

The application of microprocessor-based diagnostic strategies to industrial processes provides improved system reliability while reducing maintenance costs. The prompt detection and classification of system anomalies enables reduced troubleshooting by technicians and minimizes the misclassification of system degradations. In this paper, an experimental-based multiple hypothesis failure isolation scheme for small-scale thermofluid systems is presented. The proposed classification strategy combines dynamic modeling, control theory, multivariate statistics, and pattern recognition to develop a methodology that may be tailored for a variety of applications. The primary goal of this study is the reduction of modeling activities associated with the set of hypothesized failure modes. The overall performance of the isolation scheme is investigated for a series of experimentally simulated heat pump failures.

An understanding of the mechanics of paper separation is very important to the design of reliable photocopiers, and in this paper a model is developed to simulate friction feed paper separation. The importance of such system parameters as roller acceleration and compressive load is demonstrated. The variability of friction coefficients is shown to have a significant effect on feeding reliability. Experimental data are shown to be in good agreement with predicted results. Recommendations are made for the design of reliable friction feed paper separators.

The dynamic characteristics of rotating machinery are almost always influenced by the impedances of the machine bearings (including the bearing supports) and it is particularly important to know of their likely values at the design stage when predicting machine critical speeds, unbalance response, and stability. In many instances the vital data may be obtained from theory which must be confirmed by experimental measurement beforehand, or alternatively the data may be obtained directly from experimental measurement when the appropriate bearing has already been manufactured. The experimental techniques available for use in the measurement of bearing impedances are varied, both in their experimental procedure and in the theoretical treatment of the data that is gathered. The aim of this paper is to provide a concise account of these techniques, and to discuss the potential advantages and disadvantages of each approach. In this document, the procedures described are classified in terms of the type of experimental measuring equipment required. These vary from the simplest of test rigs, which provide only a static load in one direction and where impedances can be evaluated by hand calculation, to sophisticated apparatus capable of providing static and dynamic loads in several directions and which require computer software to extract the bearing impedances from the measured data. The type of experimental procedure selected by the rotor dynamicist will depend upon the experimental facilities available, on the time available to carry out the testing, and on the required reliability of results. Each of these considerations is discussed for each procedure described. It is concluded that measurements made using multifrequency test signals provide more reliable data, although all measurement methods yield coefficient values which agree with theoretical predictions to within about 20 percent in general, and all methods have a significant scatter of results associated with their use.

Several models for metal cutting temperatures which could be applied in simulation programs have been reported in the literature. Since the temperature predicted by the models are difficult to measure, however, there is not sufficient experimental data to determine which available model is most accurate and whether further theoretical refinement is needed. In this paper calculations from four steady-state cutting temperature models are compared with simultaneous infrared and tool-chip thermocouple temperature measurements from end turning tests on 1018 steel, 2024 aluminum, free machining brass, and gray cast iron tubes. Deformation zone temperatures calculated using the models are compared to source temperatures determined from infrared measurements using a new inverse method. Calculated tool-chip contact temperatures are compared to rake face temperatures measured by the widely used tool-work thermocouple method. The data indicates most models, though quantitatively accurate, overestimate cutting temperatures. Models based on Jaeger’s friction slider solution which include workpiece thermal property variations, however, generally give results accurate to within the reliability of experimentai methods for the materials tested. Loewen and Shaw’s model, recently generalized to three-dimensional cutting by Venuvinod and Lau, seems most accurate over a broad range of workpiece and cutting conditions. No model accurately predicts tool-chip temperatures for cast iron or 2024 aluminum, indicating that further theoretical refinement for discontinuous chip formation is needed.

A new method for the diagnosis of the tapping process using the information measure and multiple probability voting scheme is proposed. Considering the features of short cutting-duration and the large uncertainties existing in a tapping process, a set of indices based on the time-domain statistical analysis has been formed. These indices have then been evaluated and ranked using an algorithm that calculates the information gain of each index about the tapping process. The final classification decision can be made by a voting scheme based on the conditional probability functions for multiple indices. Furthermore, the information gains estimated in the index evolution process can be used as a weighting function during voting to improve efficiency and reliability. From a tapping test, which includes five different tapping conditions, a success rate of 95 percent has been achieved.

Transfer lines have long been used for machining a single product at high production rates. This study deals with the transfer line, in particular, estimating the production time per part and the effect of using probability models in describing tool and machine lives. The production time is estimated using three major causes of line stoppage. An aspiration level criterion is used to establish a scheduled tool replacement interval. The aspiration decision parameter is system reliability. It is shown how a scheduled tool replacement interval could be established to obtain this reliability level. A methodology for selecting the replacement interval for a group of tools is developed and the impact of changing this interval on tool changing costs and tool failure costs is discussed. In this study, only a single premature tool failure is assumed to occur; however, the results are compared with multiple tool failures using the renewal function. It is shown that this single premature failure analysis is equivalent to the renewal function approach for machining systems having a reliability greater than 50 percent and tool failure variabilities having a coefficient of variation less than 1.0. Using the model developed in this paper, an example problem is presented. For this problem, electromechanical equipment failure and tool wear failures are modeled using the exponential and the Weibull probability distributions, respectively. Scheduled tool replacement intervals are developed for a range of target reliabilities. The part cycle time and operating cost for the transfer line based on these reliabilities are computed. These results are compared to the minimum cost reliability level. The minimum cost operating condition is dependent on the ratio of the scheduled replacement costs over the failure replacement costs. For the example presented, the reliability level at minimum cost is much lower than the desired target value of .9. This result illustrates the importance of conducting a careful analysis and using the results to help guide operating practices toward more efficient tool change practices and maintenance policies.