Software Drives Abrasive Waterjet Machining
Proper selection and application of software to waterjet operations can enhance workpiece accuracy
By John H. Olsen
By Carl C. Olsen
Today, software plays a vital and growing role in improving machine-tool performance. Nowhere is this change more apparent than with abrasive waterjet machining. Software controls the jet to make precise parts in a fraction of the time achievable by any manual method, including manual programming. Software has made this once-exotic machining technique available to any machine shop with water and power.
Waterjet machining has some very unique characteristics. Imagine trying to precisely profile mill a 1" (25.4-mm) thick steel plate with a 0.032" (0.8-mm) diam end mill. That is a loose analog to the jet-machining situation. The first thought that comes to mind is to move very slowly to keep the chip load light and allow the tool to remain straight. That is what was done to make precise parts in the early days of waterjet machining.
When used as a cutting tool, an abrasive-laden jet of water introduces several kinds of errors when it moves quickly:
- Lag: As the nozzle moves, the bottom of the stream lags behind the top. The amount of the lag varies, depending on how quickly the nozzle moves. Lag is important on curves and corners, but not on straight lines.
- Striations: As the nozzle moves faster, the bottom portions of the jet begin to oscillate from side to side, producing marks much like flame-cutting marks. These affect surface finish.
- Taper: If the nozzle moves quickly, the kerf is widest at the top. As the speed decreases, taper reduces and eventually reverses, with the kerf becoming widest at the bottom.
- Width vs. speed: As the jet slows down, the kerf becomes wider.
- Kickback: The jet shape depends on acceleration as well as speed. A sudden acceleration at an inside corner causes the jet to kick back and gouge deeply into the bottom surface of the workpiece at the corner.
Software can predict and compensate for the complex behavior of the jet, to enable users to achieve precision while producing parts even more quickly. Because the jet behaves in a mathematically predictable way, intricate calculations can predict how it will behave.
Prediction starts with building a mathematical model of the many effects that influence the abrasive waterjet. The model is then tested in the real world, and adjusted to improve its predictions. With time and experience, the predictions become more accurate, and the model can be expanded to take more factors into consideration, further improving its accuracy.
The limiting factor in precision and productivity of a waterjet machine isn't the machine itself, but the machine's control software and the predictive model. Nozzle speed along the path, the shape of the path, water pressure, nozzle size, and other factors determine the shape of the cutting tool. Software allows compensation for this shape, which results in production of more precise parts.
Today, the human part programmer need not consider all these factors by hand to choose an optimum speed to move at every point along the path. He simply defines the part geometry and the quality of part that he requires. Software then generates the toolpath including the proper speeds at every point along the path required to achieve the specified part quality in the minimum time.
Software has been employed on waterjets from the very beginning of the development of this type of equipment with the part geometry created using a computer-driven drawing tool. Any modern CAD or vector-based drawing program can be used to define part geometry. Other geometry input means can also be used.
The next step is to convert the drawing data to a toolpath. Originally, the toolpath contained geometry and speed data, usually in the form of G code. The programmer manually coded the toolpath, made a stab at choosing proper speeds, and then adjusted the speeds either with the speed control override during cutting or by an iterative process involving trial parts and recoding. Today, there are two approaches to improve this situation.
In the first approach, G code is produced by CAD/CAM software and the speeds are estimated by using lookup tables or other means. An attempt is made to control some of the jet errors by proper speed selection. The G code is then run on a machine with a more-or-less conventional controller. This is an improvement, but it's not state of the art.
In the second approach, a CAD/CAM program generates a data file similar to G code that contains no speed data. In the place of speed, there are data describing the quality of result desired along each portion of the path. This file is then passed to the machine along with specification of the material to be cut and the material thickness. The machine has a special-purpose controller that chooses the speed and acceleration at each point along the path, using its built-in predictive model to deliver a part of the specified quality in the minimum-possible time.
The advanced controller can carry out all the functions that any ordinary controller performs. Controllers have always managed speed, acceleration, and jerk to remain within the machine's hardware capabilities. These functions are not unique. Our focus here is on the special features required for precision jet machining, which are described here and discussed in three patents on our software and controller.
First, the path geometry is preprocessed to optimize piercing and cornering. The predictive model is applied to predict the optimal pierce length for the shortest pierce time, and all lead-in lengths are adjusted to this length if space is available. If there is not enough room to grow the pierce to the optimum length, the controller will use as much room as is available, and adjust the feed as necessary to compensate and maintain the optimum pierce for that condition.
Next, the path approaching external corners is extended past the corner, so that the exit point where the jet is lagging will also pass the corner. After passing the corner, the nozzle returns to begin cutting the next leg of the path, taking into account the behaviors of the jet in terms of kickback as the jet resumes cutting the part. This approach is much faster than slowing the jet until it becomes straight and lag-free.
Next, the path is interpolated to the resolution of the motor commands. Speeds are chosen at every point to be no larger than those that will achieve the desired surface finish. If minimum taper is requested for a particular segment of the path, speeds are set for that requirement rather than surface finish. Speeds are then adjusted downward according to acceleration constraints to avoid kickback errors in corners. In fact, there are four different accelerations chosen by the predictive model that are used to manage the jet in corners, depending upon whether the corner is internal or external, and whether the jet is decelerating or accelerating. In addition, the sharpness of the corner (or radius of the arc) is considered for further refinement. At the end of these speed adjustments, a two-axis machine is ready to run, and the machine can begin cutting.
In 1997 Axel Henning of the Fraunhofer Institute showed how to automate the removal of taper with a five-axis machine. A five-axis machine that automatically removes taper has further computation to do. At this point, speeds are known at every point on the path, so the predictive model is used to predict the taper in the kerf. Further commands are then generated to tilt the head so that the part edge is vertical, and all taper is placed in the scrap.
Let's recap: The operator entered geometry data and quality information to specify the 2-D part he wanted. The control accepted those data, and then moved in five axes adjusting the speeds to produce the part quality requested in the minimum possible time. For parts of equal quality, production time can be a factor of two or more faster than for a similar part programmed by hand.
Software that adjusts speeds for quality is not an accessory for modern abrasive waterjets, but an integral part of the machine. Updating and improving the software can significantly reduce the time required to create parts.
This article was first published in the March 2007 edition of Manufacturing Engineering magazine.