Feature attraction

In the face of increasingly complex multi-axis machine tools, CAM developers will continue to automate more work that can be handled faster and more accurately by computers. Martin Oakham reports

The increasE in CAM programming tools incorporating automatic feature recognition may well signal a shift in the route to increased productivity.

There is only so much that can be achieved using increased spindle speeds/feeds, pallet shuttles and 5-axis machining cells as manual programming ties up resources and holds up workflow.

Automatic feature recognition — the pinnacle of 25 years’ research — has been arrived at through steady developments in ‘knowledge-based machining’ technology.

Feature recognition really started when traditional ‘process-based’ CAM systems began to reduce the number of individual manual-based operations needed to generate a toolpath by grouping operations, such as defining boundaries in standard processes. These processes could be remembered by the system and repeated in future operations, making it easier to program similar toolpaths next time round.

The vast majority of CAM developers exploring the possibilities of automatic CAM have based their developments on a system first proposed by Lyc Kyprianou at Cambridge University in 1980. Kyprianou devised a system to algorithmically extract higher level entities such as manufacturing features from lower level elements, such as surfaces and edges, of a CAD model. When he first proposed the method it aimed to encode parts for group technology (GT).

The purpose of GT is to systematically classify objects based on their manufacturing method. Kyprianou’s work involved classifying faces into primary and secondary groups and then identifying features according to patterns of these primary or secondary faces. A primary face comprises multiple boundaries (hole-loops) or mixed concave and convex boundaries. A concave boundary is a set of concave edges, where the solid angle over the edge is more than 180º. Secondary faces are all other faces.

However, the drawback of these earlier recognition systems was that the programmer still needed to verify that the correct operations were used and that the correct tools, stepovers, feeds/speeds were selected because the process did not automatically adapt to every component or material used.

Today, there are numerous techniques used in feature recognition, but three algorithmic approaches are emerging as dominant. These are volumetric decomposition techniques — where the amount of volume that each area is occupying is determined —graph-based algorithms and hint-based geometric reasoning. In the graph-based feature recognition, a graph showing the topology of the part [connection of faces] is created. This graph is then analysed to extract subsets of nodes and arcs that match with any predefined template.

Hint-based systems search for a ‘minimal indispensable portion of a feature’s boundary’ rather than complete feature patterns. For example, the presence of two opposing planar faces is a hint for potential existence of a slot feature. It is likely that future automatic CAM system will use a hybrid of many algorithms rather than focusing on one or two.

When knowledge based machining is combined with feature-based machining, CNC programming becomes close to automatic. If machinable features can be identified automatically, they can be linked to corresponding machining routines stored in a database. This linkage can be automated as well. The next step is toolpath generation, which is already highly automated in most CAM packages.

The majority of marketed ‘Automated CAM systems’ such as FeatureCAM, ProCAM II from TekSoft, CAMWorks and Vero Software use a set of interrelated ‘machineable’ features to describe the model to be machined. These machineable features, however, do more than just describe the shape — they are made up of a number of associative operations that inform the CNC machine tool of the preferred method of cutting the shape.

These include the preferred machining strategies for roughing and finishing, whether to use conventional or climb cutting strategies, the step down and over depths and preferred drill sequences. The associative operations are in the form of machining rules and user preferences applied via the expert systems making up the knowledge base.

Therefore, knowledge-based machining streamlines the manufacturing process by building the machining intelligence into the CAM cycles and toolpath strategies. So, the CAM system evaluates the part geometry and part material, selects the most appropriate tools and operations, recommends machining strategies, calculates feeds and speeds, then automatically generates the NC code.

When machineable features are associative with the solid model of the design, updating changes is only a matter of re-importing the CAD model. The CAM features will be regenerated and the toolpaths updated. Any areas not recognised automatically are dealt with in a more conventional manner. In other words, they are established as machinable features through an interactive feature recognition function, where the user is prompted to respond to a series of queries in a dialogue box until the system has enough information to define the area as a feature.

Vero Software’s VISI 16 incorporates new ‘compass’ technology, allowing the user to define a number of rules that will uniquely link specific geometry with a sequence of predefined operations/parameters. This will enable a company to implement adaptable toolpath strategies and proven cutting parameters to any geometric model.

For example, the program will recognise 3D features, determine which toolpath strategy is best suited according to the rules defined, then rough and finish them. Operations can be added and saved within the CAM template for use on subsequent tasks, ultimately teaching the software company specific manufacturing requirements.

Another important development is Vero’s adaptive roughing strategy. This comes into play when an area that needs to be roughed does not fit efficiently into a standard, such as Z-level type roughing cycle. The toolpath deviates to pick the area out using a trochoidal-type motion, and then returns to the previous roughing strategy used.

The use of undercuts is also permitted so lollipop-type tooling can be used. All toolpaths generated can have 15-30º of lead or lag programmed on them to increase cutting efficiency. Tilting vectors can also be set up for 5-axis toolpaths along with the option to tilt through an arbitrary point, which is particularly useful for machining deep pockets with short tooling, and tilting from a point — useful for machining islands.

In the face of ever-complex multi-axis machine tools, CAM developers will continue to automate more and more of manufacturing’s work that can be handled faster and more accurately by computers. It is possible that this process will continue until the CNC programming process becomes completely automatic based on the design and the ‘intelligence’ captured within it. Whether this allows machinists to retain all the control they want will be an issue of debate, but it is likely that this will be lost in the name of progress.