Daniel M. Gaines

A repository for design, process planning and assembly

Abstract This paper provides an introduction to the Design, Planning and Assembly Repository available through the National Institute of Standards and Technology (NIST). The goal of the Repository is to provide a publically accessible collection of 2D and 3D CAD and solid models from industry problems. In this way, research and development efforts can obtain and share examples, focus on benchmarks, and identify areas of research need. The Repository is available through the World Wide Web at URL http://www.parts.nist.gov/parts.

Custom-Cut: a customizable feature recognizer

Abstract The tools and processes available in a given shop greatly influence the way in which manufacturers view a part and the way in which they decompose it into machinable volumes, or features. Likewise, a feature recognizer should be able to produce a different feature decomposition when different tools are available. Additionally, it should be easy for a user to add new tool descriptions to the system in order to maintain and customize the feature recognizer. We address this challenge in two parts. First, we present an extensible representation which allows users to easily add their own custom tool descriptions to a feature recognizer’s knowledge base. Second, we present C ustom -C ut , a tool-based, resource-adaptive feature recognizer. C ustom -C ut accepts the user-defined cutting tools as input and automatically identifies the areas of the part that can be cut using the custom tools. We call C ustom -C ut resource-adaptive because the features it identifies will be different if it is given different cutting tools. The advantages of this is that it is easier for the user to maintain and customize and provides greater assurance that the features identifies are actually machinable with the given set of equipment.

MEDIATOR: A Resource Adaptive Feature Recognizer that Intertwines Feature Extraction and Manufacturing Analysis

A deterrent to practical use of many feature extraction systems is that they are difficult to maintain, either because they depend on the use of a library of feature-types which must be updated when the underlying manufacturing resources change (e.g. Cools and fixtures ), or they rely on the use of task-specific post processors, which must also be updated. For such systems to become practical, it must be easy for a user to update the system to match the current resources. This paper presents MEDIATOR (Maintainable, Extensible Design and manufacturing Integration Architecture and TranslatOR). MEDIATOR is a resource adaptive feature extraction and early process planning system for 3-axis milling, A resource adaptive system is one that changes its behavior as the manufacturing resources in a shop change, MEDIATOR allows users to select from a standard set of tools and fixtures, and automatically identifies any changes in the features that result, It attains its resource adaptive behavior by blurring the line between feature extraction and process planning; descriptions of the manufacturing resources are used to directly identify manufacturable areas of the part. A non-programmer can easily update MEDIATOR by selecting different shop resources.

Using regression trees to learn action models

Anyone who has ever driven a car on an icy road is aware of the impact the environment can have on our actions. In order to build effective plans, we must be aware of these environmental conditions and predict the effects they will have on our ability to act. We present an application of regression trees that allows a robot to learn action models through experience so that it can make similar predictions. We use this approach to allow a mobile robot to learn models to predict the effects of its navigation actions under various terrain conditions and use them in order to produce efficient plans.

MAPP: a matrix architecture for process planning

The goal of this work is to devise a structured representation of manufacturing planning that captures both the sequence of processing steps and the data functions, such as tooling, fixturing, and setups. The difficulty in accomplishing this is that although processing steps are often strongly ordered, these steps do not follow neat functional boundaries between tools, fixtures, setups, etc. The solution presented in this paper is a matrix architecture for process planning (MAPP) which captures a sequence of planning phases that cut across functional boundaries while still allowing data to be organized neatly into major functions. The matrix itself is a set of data blackboards: the rows of the matrix are the sequence of planning phases, the columns are data functions, such as tools, fixtures, and setups. This architecture combines advantages of sequential architectures and blackboard architectures, and enables more accurate modeling of the interwoven nature of tooling, fixturing and setup decisions in manufacturing planning.

A tool-centric approach to designing composable feature recognizers

Because parts are manufactured with a variety of processes, we would like feature recognizers that can support planning within these various manufacturing domains. These manufacturing domains often share common properties, such as similar tools and process capabilities. This raises the question of whether or not feature recognizers can be designed to take advantage of the similarities among these manufacturing domains. In particular, I explore the possibility of designing composable feature recognizers that can be created by using and/or adapting components from existing feature recognizers for related domains. This approach provides an alternative to other approaches including 1) developing a single, large feature recognizer for several domains and 2) developing a new feature recognizer from scratch for each domain. Composable feature recognizers will be smaller and easier to write than the feature recognizers from the first approach since they will be tailored for a smaller family of domains. However, since they will share components with related feature recognizers they will be less redundant and easier to maintain than those from the second approach. In this paper, I investigate a tool-centric approach to the design of composable feature recognizers in which knowledge and reasoning algorithms are structured around the manufacturing equipment. Equipment Module Libraries are developed consisting of manufacturing equipment with associated knowledge and reasoning algorithms. New feature recognizers are constructed by selecting and composing the relevant equipment modules.

Coordinator: a robust setup planner that detects fixture-feature interactions

This paper presents Coordinator, a robust, resource adaptive process planner for CNC machining centers that uses a blend of geometric and manufacturing information to detect a broad variety of manufacturing interactions and create setups plans. This system is designed to be used by expert users to reduce the time and effort required to produce good plans. One of Coordinators contributions is the ability to detect fixture-feature interactions, interferences between manufacturing operations in which one operation destroys the clamping surfaces required by another. These are an important class of interactions, which have a major (if not the major) impact on ordering setup plans. However, they are rarely addressed in the literature, mainly because traditional process planning organizations do not compute the information required to detect them at the right time. Fixturing information, which is needed for identifying feature-fixture interactions, is usually computed after all plan sequencing has been done. Coordinator is part of a large system, MAPP, which reorganized the tasks in process planning by computing the required abstract fixturing information prior to sequencing so that Coordinator can detect a wide variety of interaction types, including feature-fixture interactions. Coordinator represents an important advance in making practical, robust, and flexible manufacturing planners.

Simultaneously searching for planning goals, operators, and effectors

In this work, we describe an approach to automated manufacturing feature extraction and operations planning that combines the two processes to benefit both. When viewed in planning terms, if feature extraction is the process of identifying planning goals, and operations planning is the the process of selecting and sequencing planning operators, then what we are doing can be viewed as combining the normally separate processes of finding goals and developing instantiated operators to satisfy them. Thus, we call this approach Ebgoc (Effector-Based Goal and Operator Construction), and we have implemented it in a computer program called Mediator. Effectors are the physical equipment, such as tools and machines, that are used to transform the initial materials into the desired end product (goal). We say that this approach is effector-based (rather than feature-based) because we do not recognize a fixed set of features, but instead we geometrically derive the set of shapes that can be machined with the currently available set of effectors. Thus, each shop can customize Mediator so that it identifies machinable volumes appropriate for the resource in that specific shop. This approach gives Mediator several important properties. It can 1) effectively handle feature interactions (e.g., volumetric intersections), 2) be customized to produce features appropriate to specific shop resources, 3) identify areas that the current resources cannot machine, and 4) handle nonstandard, user-defined tooling.

Learning action models to support efficient navigation planning for unmanned ground vehicles

An effective navigation planner must have knowledge not only of the effects its actions will have, but also the effect that the environment will have on its actions (e.g. the UGV may travel more slowly over rough terrain). This is needed because the shortest path to the goal is not always the most efficient when you consider the rate of travel over the terrain. To address this issue, we have developed an approach called ERA which uses regression tree induction to learn action models that predict the effect terrain conditions will have on a UGVs navigation actions. The action models support a high level mission planner that finds efficient navigation plans consisting of way-points through which the UGV should travel. We will present results from our experiments in a simulated environment and on an RWI ATRV-Jr robot. The studies evaluate the performance of ERA in different mission scenarios with different amounts of sensor and actuator noise. Advantages of our approach include the ability to automatically learn action models, generate efficient high level navigation plans taking into account terrain conditions and transfer learned knowledge to other missions.

A model for building planners that design

Computer-aided process planning (CAPP) tools have the potential for assisting designers and manufacturing engineers in producing high-quality, low-cost products. AI planning research offers promising technologies for the development of such tools. However, current planning models limit the role that the planner can play in problem solving. Designers and manufacturers do more than simply generate process plans, they re-evaluate design decisions and design new tools to improve the manufacturability of a product. To assist in this effort, we need planners that understand the relationships between design and planning, and can suggest modifications to the part or to the tools used to create parts. To this end, we present a model for integrating planning and design. This model allows a planner to suggest changes to a problem description if it cannot solve the problem or to suggest new tools that will enable it to solve the problem in its original form. Thus, the model allows planners to take a more active role in the design process. We illustrate the use of this model by describing a system we have implemented which generates re-design suggestions to improve the manufacturability of machined parts.