CNC Controller | Control Process

CNC's fast moves

Noaker, Paula M

Rarely is the control the limiting factor in a high-speed machining application. More common productivity roadblocks are machine tool dynamics, long NC programs, and the wrong speeds and feeds. o optimize metal removal rates, you can often use the CNC to navigate around the machining system's limitations.

The CNC functionality you require may not cost as much as you think. One reason, says Steve Manolis at control manufacturer Heidenhain (Schaumburg, IL), is that advances in functionality often offer lean-frog levels of productivity that far outpace the cost of a new control.

Additional savings come from using aids, such as distributed numerical control to feed programs to controls and real-time process monitoring technology. These help controls react fast to changes in the manufacturing process to reduce scrap and improve part quality. There also may be an advantage to rethinking the way you machine parts, for example, by using fewer, more versatile tools to hog out complex contours in aluminum blocks.

TIME TRIALS

Heidenhain's latest high-speed control is the TNC 426. Besides digital servo control, it has a block processing time of 4 msec and look-ahead capability of 126 NC blocks for feed adjustment. This is a substantial gain in functionality over the company's TNC 407, which offers 30-block look ahead and 25-msec block processing time. The specifications are impressive, but they aren't worth much if the machine tool isn't dynamically capable of moving at high speeds and feeds--or you don't know how to use them to improve processing efficiency.

For example, most control builders recommend examining several characteristics when specifying a control for high-speed work. Block processing speed is a good place to begin, but it shouldn't be the only variable used to compare controls. (Manolis defines a basic block in an NC program as the bytes of information required to control all the functions for one machining position or operation. A single program will contain a variety of block sizes.)

According to Bill Griffith at GE Fanuc Automation North America (Charlottesville, VA), control builders may provide two different block processing times, with the first measured from when the control accesses a part program from its resident memory and the second measured from when the CNC accesses a part program from an external device. "The first value is easier to determine than the second," says Griffith. "You must also understand how block processing time will vary based on how many simultaneous axes will be programmed in a block and other programmed conditions, such as cutter radius compensation."

Griffith suggests that end users also ask control vendors about other specifications that can bottleneck processing, such as the following:

* Servo sample time. The sample time of the servo velocity feedback and the position feedback must be faster or equal to the system's maximum block processing time.

* Maximum position feedback pulse rate. The maximum feed rate at which the axis can be programmed depends on both feedback resolution and the maximum position feedback pulse rate.

* Interpolation rate. This is the cyclical rate at which the main CNC processor passes commands to axis processors. If block processing time is 2 msec and the interpolation rate is 10 msec, there is a bottleneck.

Digital servo control also promotes high-speed control. "Some controls still provide analog output to the motors and servodrives," says Manolis. "In analog systems, there are two loops between the control and machine interface. The controller monitors the position loop. The velocity loop is monitored by tach feedback between the servoamplifier and the servomotor. The result can be slow response times in telling the servomotor to speed up or slow down."

With digital control, NC monitors both the position and the velocity loop. Manolis reports response is faster and resolution finer, especially in contouring. Heidenhain's TNC 426, for example, can maintain resolution of 0.1 micron, 10X better than analog-controlled NC.

When specifying a control for a new machine tool or for a retrofit, you should understand the two types of acceleration and deceleration provided by controls. According to Griffith, high-speed and high-precision machining applications require a combination of the two.

"To get the best path accuracy with Fanuc's Series 15 M," says Griffith, "the control must calculate acceleration and deceleration before doing any interpolation. Preprocessing this data is necessary to look ahead to detect a part's comers and curvatures.

"Acceleration and deceleration after interpolation are important to eliminate shock on the machine. Think of turning a corner. At the transition point, you need acceleration and deceleration to smooth the comer. Older machine tools require more of this acceleration and deceleration because they really aren't designed to accelerate at the rates required for high-speed machining."

Parallel processing also is important. On Fanuc's Series 15 MB, most control functions, such as graphics, communications, individual axis control, and PLC functions, have their own processor. These work in parallel with the 32-bit main processor and communicate over a common 32-bit bus. The reason is simple. Too many serial functions on a processor would slow it down.

An option on the 15 MB is to have a 64-bit RISC processor working in parallel with the 32-bit processor. The 64-bit processor is a requirement when you need block processing rates of 2 msec or faster and more precision than a 32-bit processor can provide. "The 64-bit processor won't resolve problems with machine tool and servo systems, though," says Griffith. "A conventional servo, for example, still requires a fast enough interpolation time and update rate."

EXPLORING MACHINE DYNAMICS

High-speed processors, combined with features such as real-time dynamic compensation of the machine tool, allow end users to rethink how they machine parts. LeBlond Makino promotes use of high-speed, high power machining centers, fast controls and data communications, as well as Geometric Intelligence software, which can predict and compensate for machine dynamics on the fly. This permits high feed rates, as well as high speeds through corners and geometry changes--for example, cutting a 1/2" (12.7 mm) boss at 630 ipm (1600 cm/min) with roundness to within 0.00058" (0.0147 mm).

By zeroing out the effects of axis reversals, geometric control capability on LeBlond production machines allows end users to interpolate holes with a small-diameter end mill. They can form holes to different diameters and depths, even chamfered and counterbored, in one operation--LeBlond calls it the Tornado Process.

User-friendly macros provided by the company allow end users to generate entire interpolation routines for holemaking, as well as threading (with a special tool), from a few input parameters.

According to LeBlond engineers, interpolation can achieve higher removal rates and faster cycle times than conventional drilling. Other benefits include the following:

* Coolant flooding and chip removal improve because the tool doesn't fill the hole.

* Edge contact with the metal is as little as 30deg per rev for greater heat dissipation, allowing high spindle rpm without burning up the tool.

* The interrupted cut breaks chips to prevent bird nests, particularly with aluminum and stainless steel.

Interpolating the end mill also addresses the problem that speed at the center of a drill is always 0 sfm. By generating cutting action at every point on the tool, helical interpolation removes much of the cutting resistance. You may then be able to create holes in difficult work materials that conventional drills have trouble penetrating.

FEEDING THE BEAST

DNC communications technology provides a key link in high-speed machining applications, such as moldmaking. First, mold machining programs are often too long for the memory of the CNC, requiring an outside file server and communications processor. Second, communication speed must keep pace with the machine control. If data flow to the machine is too slow, the control will suffer from data starvation, which will show up as dwell marks on the finished core or cavity surface.

"DNC systems can use a PC as a communications processor for drip-feeding programs to the CNC," explains Carl Billhardt, president, FMS Technology Services (Westerville, OH). "This often costs less than adding CNC memory, while removing any file length limitations."

Billhardt reports the communication rate is critical with 3-D sculptured surface machining. For example, assume the data required to define a single-axis move is 9 characters, and 27 for a three-axis move, a DNC system with 19,200-baud communication can only support a maximum block cycle time equal to 1920 characters/sec divided by 27 character/block or 71 blocks/sec. Complex NC programs for moldmaking can require instantaneous communication rates beyond 100,000 baud.

"The program size determines whether programs can be downloaded in batch form directly into the memory of the CNC or whether the program must be drip-fed," says Billhardt. "The DNC solution for drip-feeding long files depends on the control. Some controls require a special interface card consisting of a large, volatile buffer memory and control logic.

"When executing short moves, the control may want to process data faster than the communications: system can send it. Executing long moves allows the system to supply data faster than required. As long as the average rate of data execution is less than the communications rate, the CNC will never starve for data. The size of the buffer determine the period over which the difference between communication input and execution output can be averaged. The larger the buffer, the less likely the control will be starved."

Billhardt says some controls permit operation in a direct tape mode with input taken from an EIA 232 port rather than the tape reader. Starvation is more likely because of the very limited buffer between input and the point of execution. For this reason, he recommends avoiding direct tape mode processing without an extended buffer when programs have a lot of short increments that must be processed at high feed rates.

When evaluating the DNC capabilities of CNCs, Billhardt recommends asking the following questions:

* What is the error detection and recovery mechanism? With simple software interfaces, Billhardt says that communications stop when a parity error is detected. High-level DNC interfaces can request that the sender retransmit the bad packet. If the same packet has data transfer errors after several tries, the system will signal an alarm.

* What mechanism handles data flow control?

*What must the CNC operator do to initiate a file transfer?

* Can the operator upload and download files between CNC and host system and drip-feed long files?

* When drip-feeding, what other control features are available? Block search, restart, or program execution?

"Evaluate these factors as a system," says Billhardt. "One control may permit drip-feeding of long programs with the standard hardware but only with DC1/DC3 flow control and no error detection and correction. A second control may require a special hardware option for drip-feeding in addition to providing a high-level software interface with error detection and correction."

FINETUNING FEEDBACK

Ideally, says Professor Tlusty, head of the University of Florida's Machine Tool Research Center (Gainesville), the NC programmer should know about the limitations of each high-speed machine and its tooling when writing programs. Often this isn't the case, yet these data are just as important as recommended cutting speeds and feeds in maximizing metal removal rates in high-speed machining.

With spindles providing 20,000 or 40,000 rpm, for example, chatter is a major detriment. "The best combination of depth of cut' and speed occurs when machining is chatter-free," says Scott Smith at the Machine Tool Research Center. Manufacturing Laboratories Inc. (Gainesville, FL), which is conducting high-speed, high-power spindle research with Manufacturing Laboratories Inc. (Gainesville, FL). "Mapping this zone of stability isn't easy without real-time monitoring and control of the manufacturing process."

Manufacturing Laboratories makes a chatter-recognition and control (CRAC) system that Smith helped develop as a graduate student at the University of Florida. Using the CRAC system, engineers start cutting at the maximum spindle speed and a stable depth of cut, then increase depth of cut incrementally until chatter occurs. The CRAC system then commands a spindle-speed change to another stable cut, and automatically increases axial depth of cut until the cut is again unstable. The process repeats until the system can no longer find a stable cut. The metal removal rate that results for the last stable cut is often higher than that for initial cutting conditions.

A test of a milling cutter machining aluminum illustrates CRAC results. The two-fluted, 19.05-mm-diam tool extends 57.0 mm from its toolholder. For the test conditions, maximum spindle speed of 36,000 rpm and chatter-free slotting DOC of 1 mm produced a metal removal rate of 274 cm sup 3 /min with a 0.2 mm/tooth chip load. Working with the CNC, CRAC regulated cutting to 26,582 rpm and 4.45 mm DOC. This boosted the metal removal rate 128% to 900 cm sup 3/ min at the same chip load.

Tlusty reports chatter-recognition and control also can produce results in high-speed, high-power machining of cast iron. In both work materials, the search for optimum cutting conditions can be done on the work material before making the required cuts to produce a part shape. Since cutting conditions are then optimized before part machining begins, less adverse forces act on fixturing and tooling.

WANT MORE INFORMATION?

SME offers video-based training courses called "Machining Center CNC Programming" and "Custom Macro Programming." Call SME Customer Service at 1-800-733-4SME 8 am to 6 pm Eastern time Monday through Friday.