Patrick M. Sammons

Iterative learning control of bead morphology in Laser Metal Deposition processes

Laser Metal Deposition (LMD) is a layer-based manufacturing process in which a laser and powdered metal are used to create a molten bead that is then traced along a path to create functional parts. The properties of the structure, including shape and material microstructure, are the result of complex interactions between the laser, the powder, the part substrate and other factors. Thus, a control algorithm is needed to accurately produce the designed part. However, feedback control of the process can create phase lag in the resulting control structure, which in turn can create dimensional instability. Additionally, the LMD process has been shown to change with part height or layer number. Taking these issues into account, a feed-forward, adaptive-type controller that changes with each fabricated layer, should be used. This paper first presents a dynamic model for the LMD process that incorporates the dependency of the process on part height. Then, an optimal Iterative Learning Process Control algorithm is presented to regulate the melt pool morphology of a deposited part using layer number as the iteration axis. A simulation study on the LMD process using the designed process controller shows that it is able to achieve good tracking performance.

Modeling and Compensation of Backlash and Harmonic Drive-Induced Errors in Robotic Manipulators

Harmonic drives, used widely in robot transmission systems, can induce significant periodic, joint-dependent position errors. Further, backlash in transmission systems, caused by wear or improper assembly, can considerably limit the overall repeatability, and therefore accuracy, of robotic manipulators. To measure the kinematic errors induced by both the harmonic drive and backlash, a laser tracker system, accurate to 10 μm at 20 m, is used to measure the end-effector position of a FANUC 200i LR Mate as its first joint is actuated randomly through ±130° (i.e., the range visible by the laser tracker). A joint-dependent model is then derived to account for the error seen in the measurements. Using a maximum likelihood estimator, the joint-dependent model coefficients and the amount of backlash are simultaneously identified. After backlash compensation is implemented, the maximum residual calculated between the nominal predicted position and the measured position of the end-effector, 0.2969 mm, is reduced by approximately 68%, to 0.0947 mm and the mean is reduced by 58% from 0.0631 to 0.0264 mm, after the error is modeled and compensation is implemented.Copyright

Control-oriented modeling of Laser Metal Deposition as a repetitive process

Laser Metal Deposition (LMD) is an additive manufacturing process whose dynamics are driven by complex heat transfer and fluid flow phenomena. The LMD process, along with every additive manufacturing process, is fundamentally a two-dimensional process possessing both temporal (or spatial) domain dynamics and propagation of information from layer to layer. However, modeling the two-dimensionality of the process for use in control has received little attention. Here, a model aimed at capturing the important nonlinear two-dimensional physical processes of the melt pool shape, while maintaining simplicity, is presented. The model is expressed in a form that lends itself to the design of repetitive process control schemes. An analytical tool is used to describe layer-to-layer stability properties of the process using the model, and insights into the fundamental stability limitations of the process are given.

Repetitive Process Control of Laser Metal Deposition

The Laser Metal Deposition (LMD) process is an additive manufacturing process in which a laser and a powdered material source are used to build functional metal parts in a layer by layer fashion. While the process is usually modeled by purely temporal dynamic models, the process is more aptly described as a repetitive process with two sets of dynamic processes: one that evolves in position within the layer and one that evolves in part layer. Therefore, to properly control the LMD process, it is advantageous to use a model of the LMD process that captures the dominant two dimensional phenomena and to address the two-dimensionality in process control. Using an identified spatial-domain Hammerstein model of the LMD process, the open loop process stability is examined. Then, a stabilizing controller is designed using error feedback in the layer domain.© 2014 ASME

Table-Based Compensation for 5-Axis Machine Tools

This paper presents a 5-axis machine tool compensation method that uses tool tip measurements recorded throughout the joint space to construct a set of compensation tables. The measurements can be taken using a laser tracker, permitting rapid measurement at most locations in the joint space. To compensate the machine tool, the measurements are used to identify a kinematic model, and then that model is used to construct an optimal set of compensation tables. The kinematic model is composed of the nominal, or ideal, kinematics with additional (unknown) six degree of freedom errors inserted between each of the joints. The error kinematics are identified using the measurement data and a maximum likelihood estimator. The identified model is then projected onto a joint-compensation space that maps to the compensation tables in the machine tool controller. Simulations of the approach are provided using measurement data from a Flow International 5-axis machine tool equipped with a Siemens 840D controller. The simulation results show a mean residual error of .076 mm, which is a 76.8% reduction from the uncalibrated machine tool.Copyright

Modeling and compensation of joint-dependent kinematic errors in robotic manipulators

Bearing systems and harmonic drives in robots introduce complex kinematic errors which result in joint kinematic errors that reduce the accuracy of their manipulators. Typical calibration methods do not consider these complex errors, thus, limiting post calibration performance. In this paper, a method of modeling and calibrating robot kinematic errors by building a joint-dependent kinematic error model is presented. Measurements are collected by a laser tracker and Active Target mounted on the end of the last robot link. A joint-dependent kinematic error model is constructed and the model parameters are identified with a mathematical algorithm based on maximum likelihood estimation. The kinematic error model is used to modify joint commands offline. Experiments are implemented on a FANUC LR Mate 200i robot. Using 250 measurements to construct the kinematic error model, the mean residual between the measured and modeled positions is reduced from 3.379 to 0.105 mm, a 96.9% reduction. Compensation is applied to an independent set of 100 measurements, and the mean residual is reduced from 3.614 to 0.131 mm, a 96.4% reduction.

A Model Predictive Repetitive Process Control Formulation for Additive Manufacturing Processes

Additive Manufacturing (AM) processes are a class of manufacturing processes in which parts are fabricated in a layer-by-layer fashion. The layer-by-layer fabrication method creates layer-to-layer dynamics. Implementing process control that neglects the layer-to-layer dynamics can lead to process instability. While repetitive process controllers which utilize only layer-to-layer feedback are a viable method, their usefulness is limited in that they are not well-suited for tracking non-periodic layer-domain references. However, since the entire reference signal is typically known a priori in AM process fabrications, a predictive control methodology can be useful for controlling fabrications in which the reference signal is non-periodic. In this paper a model predictive control formulation is extended to two-dimensions and utilized for repetitive process control Simulation results comparing open-loop and controlled fabrications for a Laser Metal Deposition process are given.Copyright