Deployment of a cloud pipeline for real-time visual inspection using fast streaming high-definition images

We investigate the challenges of building an end-to-end cloud pipeline for real-time intelligent visual inspection system for use in automotive manufacturing. Current methods of visual detection in automotive assembly are highly labor intensive, and thus prone to errors. An automated process is sought that can operate within the real-time constraints of the assembly line and can reduce errors. Components of the cloud pipeline include capture of a large set of high-definition images from a camera setup at the assembly location, transfer and storage of the images as needed, execution of object detection, and notification to a human operator when a fault is detected. The end-to-end execution must complete within a fixed time frame before the next car arrives in the assembly line. In this article, we report the design, development, and experimental evaluation of the tradeoffs of performance, accuracy, and scalability for a cloud system.     

Recurrent neural networks and robust time series prediction

We propose a robust learning algorithm and apply it to recurrent neural networks. This algorithm is based on filtering outliers from the data and then estimating parameters from the filtered data. The filtering removes outliers from both the target function and the inputs of the neural network. The filtering is soft in that some outliers are neither completely rejected nor accepted. To show the need for robust recurrent networks, we compare the predictive ability of least squares estimated recurrent networks on synthetic data and on the Puget Power Electric Demand time series. These investigations result in a class of recurrent neural networks, NARMA(p,q), which show advantages over feedforward neural networks for time series with a moving average component. Conventional least squares methods of fitting NARMA(p,q) neural network models are shown to suffer a lack of robustness towards outliers. This sensitivity to outliers is demonstrated on both the synthetic and real data sets. Filtering the Puget Power Electric Demand time series is shown to automatically remove the outliers due to holidays. Neural networks trained on filtered data are then shown to give better predictions than neural networks trained on unfiltered time series.     

A Dynamic Analysis of a Spatial Manipulator to Determine Payload Weight

This paper presents a methodology whereby the payload weight of a serial manipulator can be determined from a minimum set of sensor data, i.e., joint angle and joint torque measurements. The particular manipulator geometry that is analyzed is a four degree‐of‐freedom serial chain that is commonly used in excavator systems. It was quite remarkable that a relatively simple solution was obtained for the payload weight considering that there are a total of nine unknown moments and cross moments of inertia of the payload together with the unknown location of the center of mass. Example calculations are presented. © 2003 Wiley Periodicals, Inc.     

Prevention and Mitigation Strategies to Address Recent Brittle Fractures in Steel Bridges

Brittle fracture results in unplanned loss of service, very costly repairs, concern regarding the future safety of the structure, and potential loss of life. These types of failures are most critical when there is no evidence of fatigue cracking leading up to the fracture and the fracture origin is concealed from view. Hence, the failure occurs without warning and the details are, essentially, noninspectable. In these cases, it appears desirable to take a proactive approach and introduce preventative retrofits to reduce the potential for future crack development. These efforts will help ensure that the likelihood of unexpected fractures is minimized. This paper examines the behavior of two bridge structures in which brittle fractures have developed in recent times, discusses the causes of the failures, and offers suggested design strategies for prevention and retrofit mitigation techniques. In situations where considerable uncertainty exists in the prediction of accumulated damage or in the ability to reliably inspect critical details, preemptive retrofit strategies appear to be highly desirable.     

Cement versus cementless fixation in total knee arthroplasty

A prospectively studied group of 55 uncemented Press Fit Condylar total knee arthroplasties was compared retrospectively with a matched group of 51 cemented Press Fit Condylar total knee arthroplasties at a mean of 10 years after operation. For the cemented group, the pain and function scores improved from 32 and 45 preoperatively to 95 and 77, respectively. For the uncemented group the scores improved from 33 and 50 preoperatively to 93 and 60, respectively. There were 10 revisions in the uncemented group for femoral or tibial aseptic loosening or osteolysis compared with two revisions in the cemented group. Exclusive of problems related to patellar metal backing, survival to revision for aseptic failure or radiographic loosening was 72% in the uncemented group and 94% in the cemented group at 10 years. A significantly higher revision rate was found in the uncemented compared with cemented total knee arthroplasty of the Press Fit Condylar design.     

Long-term immunity to poliovirus in children immunized with live attenuated and enhanced-potency inactivated trivalent poliovirus vaccines

Eighty-six children who completed immunization with the two trivalent poliovirus vaccines, live attenuated (OPV) and enhanced potency inactivated (EIPV), in one of four schedules (OPV-OPV-OPV, EIPV-EIPV-EIPV, EIPV-OPV-OPV, and EIPV-EIPV-OPV) at 1 year of age were monitored serologically over the subsequent 4 years and challenged with OPV at 5 years of age. Each of the immunization groups exhibited an initial 10- to 100-fold decline in neutralizing antibody to poliovirus types 1, 2, and 3 during the first 2 years of follow-up; thereafter antibody titers stabilized. The EIPV-EIPV-OPV group maintained the highest antibody levels throughout the observation period, including the response to OPV challenge at 5 years of age. These data suggest that immunization with OPV, EIPV, and combinations of the two vaccines confers long-term immunity. Optimal systemic immunity was associated with two or more doses of EIPV.     

The partial testing design: a less costly way to test equivalence for sensitivity and specificity

We propose a new, less costly, design to test the equivalence of digital versus analogue mammography in terms of sensitivity and specificity. Because breast cancer is a rare event among asymptomatic women, the sample size for testing equivalence of sensitivity is larger than that for testing equivalence of specificity. Hence calculations of sample size are based on sensitivity. With the proposed design it is possible to achieve the same power as a completely paired design by increasing the number of less costly analogue mammograms and not giving the more expensive digital mammograms to some randomly selected subjects who are negative on the analogue mammogram. The key idea is that subjects who are negative on the analogue mammogram are unlikely to have cancer and hence contribute less information for estimating sensitivity than subjects who are positive on the analogue mammogram. To ascertain disease state among subjects not biopsied, we propose another analogue mammogram at a later time determined by a natural history model. The design differs from a double sampling design because it compares two imperfect tests instead of combining information from a perfect and imperfect test.     

Seroepidemiologic indicators for the evaluation of campaigns for the control of foot-and-mouth disease

A sero-epidemiological survey was conducted in two districts in Argentina between 1993 and 1995, to provide additional information on the epidemiology of foot and mouth disease (FMD) in Argentina and to assess the level of immunity in cattle populations, and the circulation of FMD virus. As part of the final stage of this survey, a comparison was made of the results obtained by the enzyme-linked immunosorbent assay (ELISA) and agar gel immunodiffusion techniques. Levels of population immunity against the four types of virus included in the vaccine increased progressively during the period of the survey until, in 1995, at the end of the vaccination period, the percentage of animals possessing adequate levels of protection was approximately 77% in yearlings, and more than 94% in cattle over one year old. During the three-year study, there was a clear tendency for viral activity to diminish, until in 1995 when between 3% and 0.6% were positive to the agar gel immunodiffusion test for the antigen associated with viral infection. By contrast, the ELISA detected antibody in about five times as many animals. The authors show how the increase in the level of population immunity was accompanied by a fall in viral activity.