Leandro G. Barajas

Stencil Printing Process Modeling and Control Using Statistical Neural Networks

This paper presents a neural network model for the stencil printing process (SPP) in surface-mount technology (SMT) manufacturing of printed circuit boards (PCBs). A practical model description that decomposes the overall steady-state process in independently modeled subspaces is provided. The neural network model can be updated in real-time procuring a method to control the process by dynamically searching the optimal set point of the control variables. The optimization is performed by minimizing the weighted mean squared error with respect to the desired solder brick height or volume; furthermore, in the case when multiple solutions exist, the set point that yields the lowest variance is used. The process simulator is mainly suitable for offline testing and debugging of more complex closed-loop control algorithms for the SPP optimization providing a common and realistic framework for algorithm performance evaluation. An important consideration in this paper is based on the fact that the estimation of the sampled moments of the probability distributions is made using a statistically significant number of data samples from each board, for each component type, for each printing direction, and for each pad orientation.

Local recurrence based performance prediction and prognostics in the nonlinear and nonstationary systems

This paper presents a local recurrence modeling approach for state and performance predictions in complex nonlinear and nonstationary systems. Nonstationarity is treated as the switching force between different stationary systems, which is shown as a series of finite time detours of system dynamics from the vicinity of a nonlinear attractor. Recurrence patterns are used to partition the system trajectory into multiple near-stationary segments. Consequently, piecewise eigen analysis of ensembles in each near-stationary segment can capture both nonlinear stochastic dynamics and nonstationarity. The experimental studies using simulated and real-world datasets demonstrate significant prediction performance improvements in comparison with other alternative methods.

3D visual perception system for bin picking in automotive sub-assembly automation

Industries are being swept up by the tide of market innovations originated from pervasive application of Robotics and Automation (RA). Given the critical role of RA in industry, it has become relevant for RA to have human-like decision making capabilities. Such enablers intrinsically require the use of flexible and robust 3D perception and control systems. In the process of automating complex automotive sub-assemblies, 3D vision-based recognition as well as grasping of complex objects is required not only for detection and categorization but also for pose estimation and robotic pick-and-place operations. In this paper, we propose a novel 3D visual perception system for sub-assembly automation based on a structured light 3D vision system. We use a novel geometric surface primitive patch segmentation approach based on Hough transforms to obtain accurate surface normal estimations from 3D point clouds for the identification of patch primitives. The most relevant primitives for our application include planar, cylindrical, conic and spherical surface patches. We extract primitive surface patches from automotive CAD models in DXF format. The models are then decomposed to simple entities such as planar polygons, vertexes and lines. Our resulting models based on the 3D point clouds are composed only by simple planes and cylinders. Our system takes advantage of the available CAD data for both object recognition and for pose estimation. Our experimental results demonstrate that we can achieve, in only a few seconds, a highly accurate pose and object class estimation.

Estimating object grasp sliding via pressure array sensing

Advances in design and fabrication technologies are enabling the production and commercialization of sensor-rich robotic hands with skin-like sensor arrays. Robotic skin is poised to become a crucial interface between the robot embodied intelligence and the external world. The need to fuse and make sense out of data extracted from skin-like sensors is readily apparent. This paper presents a real-time sensor fusion algorithm that can be used to accurately estimate object position, translation and rotation during grasping. When an object being grasped moves across the sensor array, it creates a sliding sensation; the spatial-temporal sensations are estimated by computing localized slid vectors using an optical flow approach. These results were benchmarked against an L∞ Norm approach using a nominal known object trajectory generated by sliding and rotating an object over the sensor array using a second, high accuracy, industrial robot. Rotation and slid estimation can later be used to improve grasping quality and dexterity.

Real-Time Diagnostics, Prognostics and Health Management for Large-Scale Manufacturing Maintenance Systems

Traditional technologies emphasize either experience or model-based approaches to the Diagnostics, Prognostics & Health Management (DPHM) problem. However, most of these methodologies often apply only to the narrow type of machines that they were developed for, and only support strategic level assessments as opposed to real-time tactical decisions. By enabling widespread integration of diagnostics and prognostics into our manufacturing business processes, we have reduced spacio-temporal uncertainties associated with future states and system performance and therefore enabled more informed and effective decisions on manufacturing activities. For large-scale systems, the usual approach is to aggregate multidimensional data into a single-dimensional stream. These methods are generally adequate to extract key performance indicators. However, they only point to observable effects of a failure and not to their root causes. An integrated framework for DPHM requires the availability of bidirectional cause-effect relationships that enable system-wide health management rather than just predicting what its future state would be. This paper summarizes best practices, benchmarks, and lessons learned from the design, development, deployment, and execution of DPHM systems into real-life applications in the automotive industry.Copyright

DeviceNet network health monitoring using physical layer parameters

Since the 1980s, the manufacturing environment has seen the introduction of numerous sensor/actuator bus and network protocols for automation systems, which led to increased manufacturing productivity, improved inter- changeability of devices from different vendors, facilitated flexibility and reconfigurability for various applications and improved reliability, while reducing installation and maintenance costs. However, such heightened manufacturing integration facilitated by industrial networks also leads to dramatic consequences of improper or degraded network operation. This paper presents a novel Network Health Management system that provides diagnostic and prognostic information for DeviceNet at the device and network level. It extracts features from analog waveform communication signals, as well as logic and timing features from digital packet logs. These features are used to evaluate the network system performance degradation by applying multidimensional clustering techniques. In addition, this work proposes a hybrid prognostics structure using combined physical and logic layer features to provide fault location information that cannot easily be realized with analog or digital data independently. Furthermore, an intermittent connection diagnostic algorithm which analyzes patterns of interrupted and error packets on the network was developed. This tool can be used as a packet source identification method which uses joint analog features and digital information inferred from analog waveforms. A test-bed was constructed and the experiments of network impedance mismatch, cable degradation, and intermittent connections were conducted in laboratory environment. Experimental results show that the proposed system can detect degradations of the network and identify the location of the intermittent connection successfully. Field tests performed in an industrial environment were also conducted and their results are discussed.

Dexterous robotic hand grasping method for automotive parts

Kitting processes are fundamental enablers in flexible manufacturing environments derived from minomi principles. A typical kitting application will sort loose and unpacked parts without dunnage into a tray or “kit” and then place them near the point of assembly for easy reach by assembly workers. In order to prepare, sort and sequence the kits, it is necessary to have adaptable robots and automation able to pick-and-place a variety of parts at line production rate. This requires the assembling any of hundreds of kit types, each one having about 10 different parts within 60 seconds. These requirements are a fundamental challenge for kitting automation given that it requires several parts of different shapes, which are presented in either random or semi-structured fashion, to be grasped and placed into the kit. Highly flexible manufacturing (HFM) requires grasping previously unknown objects for which a computer 3D model may not be available. A methodology that integrates vision and flexible robotic grasping is proposed to address HFM herein. The proposed set of hand grasping shapes presented here is based on the capabilities and mechanical constraints of the robotic hand. Pre-grasp shapes for a Barrett Hand are studied and defined using finger spread and flexion. In addition, a simple and efficient vision algorithm is used to servo the robot and to select the pre-grasp shape in the pick-and-place task of 9 different vehicle door parts. Finally, experimental results evaluate the ability of the robotic hand to grasp both pliable and rigid parts.

Continuous flow modelling of multistage assembly line system dynamics

The wide spread availability of real-time plant floor systems (PFS) information has made the modern automotive assembly line a data-rich environment. Information from these data sources offers an unprecedented opportunity to model and simulate the performance of an assembly line as a dynamic system, as opposed to a conventional static manner. The dynamic models in turn can enable fast and accurate prediction of aggregate performance in multistage assembly line operations. This paper presents a data-driven continuous fluid flow approach, founded on nonlinear system dynamics (SD) principles, to model assembly line dynamics. The movement of entities is treated as a fluid flow, buffer stocks are water tanks, the conveyor belt is water pipe and manufacturing stations are the valves which control the rates of flow. A set of ordinary differential equations (ODE) is derived to model the system of buffer stocks and production flows between the interacting machines. The proposed continuous flow models are implemented in Matlabs Simulink® environment with real-world data from a production line segment of 18 machines. The results show that the instantaneous (i.e. approximately 1 week of actual operation time) throughput rate values from the continuous flow model were within 5% of the historical data averages while the results from an equivalent discrete event simulation (DES) model were inferior. In addition, the steady-state results of a 20-h simulation run (i.e. approximately 1 year of actual operation time) match well between the DES Model and the continuous flow model. This investigation is strongly indicative of the potential use of continuous flow models to capture the aggregated assembly line dynamics and yield deeper insights into the interrelations between the different parts of a complex manufacturing system.

Dynamics and performance modeling of multi-stage manufacturing systems using nonlinear stochastic differential equations

Modern manufacturing enterprises have invested in a variety of sensors and IT infrastructure to increase plant floor information visibility. This offers an unprecedented opportunity to track performances of manufacturing systems from a dynamic, as opposed to static, sense. Conventional static models are inadequate to model manufacturing system performance variations in real-time from these large non-stationary data sources. This paper addresses a physics-based approach to model the performance outputs (e.g., throughputs, uptimes, and yield rates) from a multi-stage manufacturing system. Unlike previous methods, degradation and repair dynamics that influence downtime distributions in such manufacturing systems are explicitly considered. Sigmoid function theory is used to remove discontinuities in the models. The resulting model is validated using real-world datasets acquired from the General Motorpsilas assembly lines, and it is found to capture dynamics of downtime better than traditional exponential distribution based simulation models.

Process control in a high-noise environment using a limited number of measurements

In this paper, we develop a hybrid control algorithm that produces control values for processes where only a limited number of function evaluations are available for the control law generation. This situation arises, for example, in stencil printing processes in printed circuit board manufacturing, where the cost associated with multiple function evaluations is prohibitive: The proposed control algorithm is given by a modified version of a constrained conjugated-gradient method, transitioned into a windowed-smoothed block-form of the least-squares affine estimator.