Printer Friendly

Tool-wear insensitivity: why are nine out of ten machine-tool buyers ignoring this technology?

Tool-wear insensitivity

Why are nine out of ten machine-tool buyers ignoring this technology?

Tool-wear sensing has come a long way in the past few years. The eyes and ears of operators--through no fault of their own--have not kept pace with high-speed machining. Fortunately, technology has, and you can now obtain electronic watchdogs that not only bark when you break a tool, but also whine when a tool has lost its cutting edge and its time to throw that machine a fresh "bone."

Unfortunately, few in this country have the confidence yet to give tool-wear monitoring a try--hardly 5% of new-machine buyers chose this option. Yet, for hardly 5% of the total price of that machining or turning center, the rest could gain collision protection, broken-tool protection, automatic tool-replacement, and important feedback on the cutting process. Instead of waiting for the shop down the street to be the first on the block with this technology, they could be pushing their cutting tools to their full potential.

Three ways to go

There are three basic tool-wear sensing technologies--power, force, and vibration--and three corresponding levels of sophistication. All go beyond simply detecting broken tools or collisions.

The simplest approach is monitoring the power consumption of spindles and/or axis motors, comparing normal power or torque levels with preset limits for tool degradation. It uses simple current sensors and is relatively easy for the operator to understand, but yields the least sophisticated response and may fail to prevent machine damage for rapid tool failures. Because of its simplicity and track record, it is well established technology and the most widely used.

The next step up in sophistication is the mounting of force-sensing elements (usually strain gages) in the cutting-tool support structure to measure the actual machining reaction forces in spindle bearings, turret bases, or lead screws. Force sensing is gaining acceptance with the more sophisticated user and offers much more "intelligence" about cutting conditions. The sensing unit is relatively complex, expensive, and, in some cases, difficult to retrofit to a given machine tool. However, once initiated into this technology, users find they can no longer live without it.

Tops in sophistication is the acoustic-emission approach. It attempts to analyze the complex envelope of mechanical vibration emitted by the cutting tool itself, zeroing in on characteristic patterns indicating tool degradation or the "snap" of tool chipping or major breakage. It uses a relatively simple piezoelectric sensor embedded within a foot or less of the cutting tool, but the analytical systems and software quickly get very complex. Because this is still a young technology with a small database of field experience (characterizations for a multitude of cutting conditions), it places a large burden on the user to do the necessary experimenting and interpretation to establish reliable trigger responses to a variety of machining conditions.

Tool-wear sensing evolved from a simple ammeter with trip contacts that cost a few hundred dollars and required no great user insight or experience--a curiosity that set off an alarm when something was obviously amiss. Now, apparently, it's too much of a leap to ask the machine buyer to pay upwards of $10,000 for a system that may or may not be much better at reliably identifying tool-wear conditions, depending on user skills and dedication.

Acoustic emission

Let's start first at the leading edge of tool-wear sensing technology. A leading expert in acoustic emission is Dr David Dornfeld, Professor of Mechanical Engineering, Laboratory for Manufacturing Automation, University of California, Berkeley. Although this technology has been applied to machine tools for over 15 years, success has been limited. The basic sensor is a high-frequency piezoelectric transducer (usually lead zirconate titanate) that responds to frequencies in the 75 kHz to 1 MHz range-its key advantage--sensing vibrations well beyond the audio range and most machine-tool operating frequencies.

Dornfeld's group is presently working on reducing extraneous noise in the input signal, characterizing a wider variety of cutting-tool frequency signatures, and increasing the reliability of the system's decisions. They are developing a database to determine what initial setup conditions to use for a given machining range.

In single-point turning, the transducer is threaded into a drilled hole, either inside the shank of the tool-holder or in the turret, close enough to be acoustically coupled by steel-to-steel contact to the cutting tool (usually 6" to 12"). The sensor picks up structural-borne (not airborne) vibration, the high-frequency stress waves emitted by the normal cutting process or by fracture of the tool tip. A seperate instrumentation box monitors the AE energy signal, detecting changes in stress waves indicating changes in the tool geometry. Based on parameter settings, it analyzes frequency patterns and produces an output indication of tool wear and/or tool breakage.

As Dornfeld explains, "A sharp tool has a characteristic level of vibrational energy and particular mix of frequencies. As the tool wears, that energy level gradually increases as the contact areas increase, indicating tool wear. Tool fracture, on the other hand, produces a stress wave--a sudden release of energy--indicating catastrophic failure.

"An AE system can detect normal degradation, accelerating degradation (small chipping and cracking of the tool just before failure), as well as final catastrophic failure. It can also monitor chip breaking--telling whether chips are continuous or discontinuous. Of course, every cutting situation is slightly different, and it takes some specific machining experience to fine-tune the system to get reliable warnings of impending cutting hazards."

Stop on a dime?

What can you do once you've detected tool failure? "Most people are satisfied with simply doing a feed hold", he replies, admitting that it's doubtful you could stop a lathe in half a spindle revolution at high machining speeds. "In most cases, we've found we can make a decision on tool failure within 5 to 10 msec, but nothing can really stop a tool in a high-speed cut. That's why it's more promising to detect precursors to failure."

Dornfeld admits there is some danger of false signals with AE, just as there is with any tool-wear-sensing scheme, particularly in the initial learning period. "The acoustic energy from the cutting process is quite high, and we are effectively zeroing in on only a narrow slice of that high-frequency energy spectrum--only 1 kHz of bandwidth. Very little extraneous signal gets through, and that would also be damped out by structural interfaces in the machine tool. We've never found a false-triggering problem in all the tests we've done. When somebody drops a wrench on the machine next to the sensor, you do get some extraneous signals, but they are not in the typical frequency or energy pattern we're looking for. A transient signal is not focused anything like the cutting-tool signal."

His group is looking at teaming acoustic with other transducers to increase reliability over a wider range of applications. "We've made a lot of progress on applying AE to extremely small machining loads--diamond turning or finish grinding, for example--where there's really no alternative technology. But, where there's contact, large relative velocities, and chips being formed, AE works extremely well, and it also benefits from the fact that precision machines tend to have very low noise levels."

High-speed hang-ups?

Surprisingly, according to Dornfeld, high-speed machining is no more of a challenge for AE than normal machining speeds. "We're working on a transducer now that mounts on a milling-machine spindle and picks out the engagement of each milling tooth. This application is actually software dependent; it's not limited by transducer sensitivity. The question is finding the right software to extract information on fracture, progressive wear, and even the amount of material being removed."

Although energy levels increase with speed, they are looking at ways to get rid of the mean signal level (make it independent of machining velocity), and be able to look at the more subtle deviations indicating changes in tool-geometry or work-material characteristics.

Force-sensing alternative

The IntelliTool tool-monitoring package is produced in Europe by Sandvik and available here through Automation Intelligence, Orlando, FL. In it, a strain-gage load cell provides a voltage output signal that's related to the thrust force on the cutting tool. In a CNC lathe, the sensor plate is mounted between the turret and its slide (if possible) and measures cutting-force components in the machine's X-Z plane--axial and radial force combined into a single vector. A single plate sensor monitors all the tools and operations performed by that turret.

A microprocessor signal-conditioner can identify up to 998 operations, each with its own alarm limits corresponding to monitored blocks in the part program. Some interface signals from the machine control are required to maintain synchronization between sets of alarm limits and actual part-program executions. In four-axis machines, a second sensor is used under the other turret.

Explains AI product manager, Paul Magadanz, "The basic idea of bracketing the sharp-tool operating conditions with alarm limits has been enhanced for CNC applications and made more user-friendly by newer microprocessor features. You simply push a learn button on the tool monitor, and the CNC program interface drives the tool monitor through a learning process. For optimum sensitivity, you can do the cutting program on a block-by-block basis and optimize alarm limits, but this is not necessary if you're simply looking for tool breakage or crash protection."

In a single-point monitoring situation, a lower-cost, single-operation control unit can be used, without the multiple-operation capability. It also monitors tool wear, tool breakage, and missing-tool threshold conditions.

High-tech hurdles

Magadanz admits that there's some reluctance to accept this technology--for example, in low-volume FMS situations, where there's little cutting experience to benefit from. "This is not a simple clip-on device for detecting broken or dull tools. Nothing like that yet exists. The user must do a lot of trial-and-error work to get major benefits. It takes a certain amount of user sophistication, just as it takes sophistication to determine exactly what's a dull tool when you're starting up a new machining process.

"A tool monitor cannot make an arbitrary decision or absolute judgment of cutting quality. That depends on actual cutting parameters--the ability to hold a dimensional requirement, maintain a surface finish, or eliminate chatter conditions--based on practical experience developed over time. To achieve automatic tool-wear and breakage monitoring, you must correlate the level of tool dullness to a measured parameter like percentage increase in thrust force. Then you adjust alarm limits to get the tool-condition reactions you want."

Pushing the machine

Despite the present low level of tool-wear-sensing use, Magadanz notes some improvement. "We're seeing a lot more inquiries as people put their machines into more automated, less manpower-intensive cells. If you're pushing the machine, you need this new level of intelligence.

"There are three main incentives: minimize consumption of expensive tools, minimize the amount of machine downtime for replenishing tools, and boost machining rates by pushing cutting tools more aggressively. Then, it becomes more important to detect sudden tool failure and shut down the process before it becomes a very expensive situation to correct."

So, these monitoring systems provide a damage-limiting function, not an insurance policy against all possible disasters. "In one case," Magadanz recalls," a user wanted to run indexable carbide drills on an undersized lathe. Previously, they consistently knocked the turret out of alignment when they broke a drill point. In adding our tool-monitoring system, they were not necessarily concerned with saving the drill holder, just avoiding having that machine down for a day and a half to re-align the turret.

"In another case, gundrilling of a $30,000 titanium turbine ring, we were able to offer the ability to detect a chipped cutting edge prior to catastrophic failure of a gundrill and save the workpiece."

Price phobias

Typical package price for AI's force-sensing system is about $8000, evenly divided between sensor, control unit, and OEM installation. In a drilling application, the sensor would become a bearing housing--a cylindrical sleeve that houses the spindle bearings. In some retrofit situations, it may be a problem finding space for the spindle sensor. For a drilling station, it may be necessary to use a plate sensor between the station slide and the spindle to be monitored.

Yet, $8000 seems reasonable, considering the potential benefits. Why aren't more people choosing it? User phobia that they don't have the sophistication to benefit from this option? "We're seeing less and less of that these days," he replies. "Operators of multispindle automatics, for example, still have a fairly high level of expertise. These companies are fairly receptive to the tool-sensing idea, that their operator can key in alarm limits and experiment to get the right parameters to avoid tool-breakage problems.

"None of these systems work unless you make them work. You can set up a tool monitor, neglect it for a couple of months, and come back and find that the operators have dialed up alarm limits that don't represent reality, effectively disabling the system."

AE insight

When Magadanz looks at his acoustic competition, he sees two key points. "The advantage of the AE approach over force sensing is that their sensors are more easily retrofitted and lower cost. The disadvantage is that, for each individual operation, you need a very sophisticated learning system to identify at which frequencies these energy fluxuations occur. Making that usable on the shop floor level is the major hurdle.

"Our strategy is to provide a system that requires some tuning initially, but is better understood at the shop-floor level. As flanks wear, the land gets larger, thrust force increases, indicating tool wear, and allowable values can be established for a given tool. The user can readily understand the raising and lowering of alarm limits. But when you get into spectral analysis of an acoustic-emission signal, there are many more parameters involved than the shop person is trained to evaluate."

But he concedes that creating a usable, broad-based shop-floor AE package is something Sandvik is looking into. What other tool-wear technologies are out there? "Not vision or any other tool-wear control strategies," he feels. "Our developmental efforts are directed toward easier retrofits and lower-cost sensors to do essentially what our sensors can do today."

Smaller sensors? "Yes, and perhaps self contained, with contactless data transmission--something that can be built into a machining center without slip-ring worries or other contact devices. There's a tremendous market for something you could bolt on to a standard tapping or drilling spindle on an existing transfer line or build into a machining center."

A combination approach

Montronix Inc, Raleigh, NC, was recently spun off as an independent tool-monitoring company by Kennametal, who retains an ownership interest. Unlike Sandvik or Dornfeld's group, concentrating on a single technology, Montronix offers both: an acoustic emission system for tool breakage, the ATM line; and a cutting-force system, the TS line. Although they don't recommend the ATM acoustic product for lathes, says John Powell, Montronix president, "that system works exceptionally well in machining centers for drills, taps, and reamers, particularly for smaller sizes. You get very little acoustic noise from the cutting process itself and significant AE when the tool breaks."

Doesn't the ATM provide some tool-degradation information also? "Yes, you get some, and we can provide an alarm output to pick that up, but we do not consider it an exceptionally reliable wear sensor. We have twelve-spindle drilling applications for an automotive transmission ring where we detect the breakage of any one of those drills with a single system. In addition to protecting a transfer-line station, it also saves cycle time, so the benefits are significant. We have done quite a bit of work with acoustic, and do not see it as being reliable in a global sense."

Thus, the ATM, he feels, will remain a special instrument for tool breakage until such time as AE gets sophisticated enough to provide reliable tool-wear sensing. "I don't think you'll see a general-purpose product based on acoustics for a turning machine, for example."

In the meantime, they use force sensing to cover turning situations. "That has been very successful for us in lathe operations. Installed prices for force sensing range from a basic system for collision-type protection for $10,000, to a full-blown system for wear, breakage, and collision for up to $40,000."

Powell agrees that only about 5% of machines sold today have tool-monitoring capability, but notes that markets are different around the world. "In Europe, there's a much broader acceptance of tool monitoring. There, you're forced by local labor laws to automate completely, whereas here, the driving factor is cost savings."

Isn't there also a reluctance here that this technology is too sophisticated for our labor force? "Well, we've found that once people actually use a monitor effectively, they never want to buy another machine without one. Getting people to try this technology is a bit of a barrier.

"For example, one customer, who makes brake components, is running very automated and at high cutting speeds. He quickly concluded that running at 3000 sfm with ceramic tooling on cast iron without a full-blown sensing system was a significant risk. A ceramic insert often fails here by breaking, and every time this happens, you wipe off the entire end of the toolholder. Whereas, with a good wear sensor, you can avoid a lot of this breakage, or when you do break the insert, you can stop the machine immediately, change the insert, and be back in production immediately."

How fast can you stop the machine? "We can typically output a stop-tool signal in about 10 msec, and the real issue is the total response time of the machine. We're seeing numbers there of 25 to 50 msec (including our 10 msec). A lot depends on the interface. You must go directly into the axis drives to stop them quickly, not through the entire CNC loop."

Power-monitor alternatives

Steve Beem, project manager, Government Systems Group, Cincinnati Milacron, Cincinnati, OH, is their expert on tool-wear monitoring. Milacron's machining monitor has been out for over five years and is based on spindle-motor load and slide-drive loads, without any force or acoustic sensor inputs. The system for lathes has been recently upgraded to a PC-based unit with graphic displays. Compared to the other more sophisticated technologies, it is much simpler: you run a signature and a tare for a particular cutting-cycle program, and the monitor can then detect tool wear or insert breakage.

Beem acknowledges that a lot of work is being done in tool-wear sensing. "Everybody has a different product, and Milacron is not adverse to putting other systems on our machines. We're doing a research project on untended machining, and this is where tool-wear sensing becomes much more critical. It's part of the critical mass of features we've determined you need to do untended machining.

"Yet, this study also found that very few people do untended machining," he concedes. "One of the things we're addressing in the MIAS program (Machining Initiative for Aerospace Subcontractors) is to make untended machining easier to do, take some of the overhead and up-front work out of it, so that more people can benefit from it. This involves things like probes to compensate for operator absence by measuring or sensing the part, and tool monitors tuned to the cutting process as well as sensing what's going on with machine loading/unloading, part queuing, etc. All of this information must be available."

What experiences have Milacron users had with tool-wear sensing? "Although our present tool monitor is not as sophisticated as others, there's a definite benefit, even when there's an operator standing at the machine. We have, though, identified a need that current untended technology isn't addressing, and that is the difficulty of detecting the breaking of one insert in a multitooth cutter. In one case, we could not catch this before the tool would weld to the workpiece.

"Our monitoring system's response time for the case of the single-point cutter on a lathe is quick--it gets the job done. But, on the multitooth cutter, we have problems because the torque sensor can't detect a single-insert break until it's too late. When we looked at the technology that was being offered last year, we found nothing that could address that situation. Although we felt that acoustic might work, at that time that technology worked best with single-point situations."

Beem agrees that tool-wear sensing has been very slow to catch on in this country. "I think people are equating this with buying insurance. They feel that by assigning their best people to that new machine and training them well, they won't have tool-breakage or wear problems. So, why pay for 'insurance' they probably won't need. If tool-wear sensing clearly made a machine run faster and make more parts, they might see the benefit more readily.

"Once tool-wear sensing becomes an integral part of the total machine package of software, controls, and hardware, it will become more readily adopted by industry. Any machine that's part of a flexible manufacturing system or a single machine running untended needs these features. But most people today still have operators standing there, and see no benefit of relieving them of their responsibility to supervise the operation of that standalone machine. This is particularly true for the smaller shops."

PHOTO : Montronix TBS tool-breakage sensor on a Warner & Swasey lathe.

PHOTO : Elements of Automation Intelligence's IntelliTool tool-monitoring system: feed-force sensor, left; plate sensor, center; and monitor, rear.

PHOTO : Comparison between axial feed-force and torque measurements using the same tool, first sharp and then worn, shows greater discretion between the two conditions for the force-sensing approach.

PHOTO : How the IntelliTool feed-force sensing system monitors cutting-edge conditions on a lathe, and how the sensor output signal increases with tool wear.

PHOTO : IntelliTool plate sensor is installed behind the turret of a vertical turning machine to monitor all turret operations.

PHOTO : Setting up the Kennametal tool-breakage acoustic-emission sensor: For initial calibration, a pulser is placed on the workpiece, emitting a simulated tool-breakage signal that travels through the workpiece, vise, and table to the AE sensor. Tool-monitor gain is then adjusted for the damping caused by this collective impedance, and the pulser removed. During subsequent machining, any tool break will be instantly sensed by the monitor.

PHOTO : The Milacron Machining Monitor for turning centers compares X- and Z-axis drive torques with an upper limit based on an initial "tare" noncutting cycle plus a factor for tool wear.

Eugene E Sprow Special Projects Editor
COPYRIGHT 1990 Nelson Publishing
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1990 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Sprow, Euhene E.
Publication:Tooling & Production
Date:Oct 1, 1990
Previous Article:EC-92: more opportunity, more competition.
Next Article:Systems integrator manages multiple-vendor system: systems integrator, working with General Motors Buick City plant, helped improve LeSabre quality.

Related Articles Now Features a Comprehensive Onsite Auction Area. Passes the `10,000th Request for Quote' Mark. Corrects and Replaces Previous Announcement, BW1156, IL-MACHINETOOLS.COM. Passes the '10,000th Request for Quote' Mark.
Distribution News. launches financing center.
The proliferation of the dot com.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters