...Then there were
the gray area between
What does a mechanical engineer need to know about control theory? More than he did ten years ago. The same is true for electrical engineers and their understanding of mechanical dynamics. More and more, engineers are having to cross disciplinary boundaries, finding themselves in a place uncharted by traditional college engineering curricula. This middle ground formed by confluence of mechanical, electrical, and computer engineering is commonly called “mechatronics” and is changing the complexion of motion system design.
One person quite familiar with the concept, Robert Bigler, President of Animatics, Santa Clara, Calif., gives his take on mechatronics: “It is when you physically meld electronics with mechanics. And, I would say most anyone would agree that’s the future.” The continual advancement of this interdisciplinary approach to machine design has mechanical, electrical, and industrial engineers treading new waters.
Mechatronics involves a deeper and broader melding of both the intelligence and energy coursing through a machine than an “electromechanical” route system, which is based on converting electrical signals into mechanical displacement, or vice versa. Fueled by advances in semiconductor technology, electronic signals and artificial intelligence can now regulate, often strictly, output position and power, while at the same time control any number of machine function parameters along the drive line.
In traditional machines, the intelligence (logic signals) and power (motor current) meet in front of the drive, usually at the controller. In a mechatronic machine, intelligence and power meet in the drive. If the intelligence is in the drive, then the drive controller is a window into the machine. There you can see how the mechanical structure is responding not only to the drive, but also to the control signal, explains Mahito Ando, Venture- Com, Cambridge, Mass.
Little by little, electronic intelligence is taking over functions in interactive toys, semi-active automobile suspensions, point-and-shoot cameras, and other adaptive machines. In other words, electronics are being absorbed into everyday machines and everyday life.
“For three or four decades we have been entering the computer’s world,” says Dr. Victor Zue, associate director of the MIT Laboratory for Computer Science for MIT. In systems now under development, the computer adapts to our world. Cars, houses, and rooms will be instrumented rather than us having to carry data back to external computers, explains Zue.
In five to 10 years, smart technology “will percolate into our homes,” predicts Chris Luebkeman, director of research development at Ove Arup Engineering and a founder of MIT’s Intelligent Homes of the Future project. He envisions buildings that are “aware and adapt to us.”
Engineers involved in the design of such “smart” systems will be required to learn about and apply mechatronic technology. To help them, universities across the U.S. are beginning to offer cross-disciplinary courses and programs in mechatronics.
Before the age of microprocessors, PLCs, and DSPs, state-of-the-art motion systems were heavily reliant on mechanical positioning devices. Complex (and often extremely clever) systems of cams, pulleys, indexers, and gears ran many machines with seemingly clocklike precision.
As applications demanded higher accuracy and electronic instrumentation began to take deeper root, the once-subtle shortcomings of many motion conveyance mechanisms became more imposing. Greater degrees of electronic control and monitoring made it clear that problems with accuracy often resided in the mechanical linkages. No matter how high the resolution on a controller or regulator, a servo could only position as accurately as the sloppiest portion of the system.
So it often was (and is) the case that, where high-grade servo positioning is required, meticulously processed gears, couplings, cams, and the like were used to supply high torsional stiffness and geometric tightness. These “servo class” components remain a huge factor in the scheme of many precision motion equipment designs.
But, even the most finely crafted mechanical part adds complications. As the driving component is linked to the load more and more directly, unwanted dynamics are eliminated and the load moves truer to plan. In the minds of many engineers, for certain designs direct-drive is the hands-down only option.
A direct-driven system may be one of several forms. A simple case is a motor directly fastened to the output shaft. The need for accuracy may preclude gearing or belting, even though the motor is liable to waste much of its energy running at relatively low speed and torque. A speed reducer harnesses the motor’s high-speed power to produce high torque at lower speeds. Thus, for simple rotary actuators, a direct-driven system is generally used only in rare cases where just the slightest backlash and loss of torsional rigidity are intolerable.
Nevertheless, these situations do come up, and they tend to be low-speed and high inertia. Therefore, large-diameter servomotors are often found on direct drives to supply the low-end torque needed. Specialized construction can make a motor better suited for direct drive. One design is an outer-rotor motor, in which the outer portion of the motor turns. The larger rotor radius converts a greater proportion of the electrical energy into torque; less of it goes into speed.
These direct-driven systems obviously rely on sophisticated controls – you’re not going to go to create a seamless motor-toload connection and then manually turn a knob or throw a switch to hit the targeted position. Closed-loop digital control is used. Often, a motion function is input into a microprocessor-based controller, keeping the motor output in strict compliance with the required load movement.
This programmability enables use of such mechatronic functions as electronic cams.
A mechanical cam, to begin with, is a device with eccentric contours traced by a follower. The follower translates to describe a nonlinear path of motion, transferring simple rotary or linear motion into a more complex movement pattern. A mechanical cam is often designed through numerous iterations leading to the finalized shape. Once machined, it’s pretty unlikely you’ll want to change it. These cams are made to provide a specific – and usually permanent – motion profile.
Electronic cams are a different story. With a motor driving an axis, a controller can be programmed in a matter off minutes to produce any number of motion curves. Reprogramming for a different profile takes little time. By driving multiple axes, each with their own independedent motor, you can provide complex synchronized motion that can be changed as required.
Furthermore, even if the motion profile is never expected to change, it’s not unheard of to use an electronic cam; sometimes the cost of machining a mechanical cam justifies the electronic alternative.
Setting up “electronic gearing” is another method of simplifying the mechanical composition of a drive system. Actually, electronic gearing is merely a multiple-axis synchronized motion setup in which a controller is programmed to run separate motors (and axes) at various ratios as required – the master-follower relationship. The mechanical version is a driving axis branching off into several drive paths geared to different ratios, all of which work together. With electronic gearing, there are multiple drives. This can get expensive, but eliminating mechanical linkages will increase accuracy, and the ratios are reprogrammable. They can even be programmed to vary during different phases of motion, becoming position- dependent gear ratios.
And electronic linkages facilitate inprocess adjustability. Jacob Tal, President and co-founder of Galil Motion Control, Rocklin, Calif., uses the example of packaging. “Sometimes in packaging, if you’re pulling paper, plastic wrap, and so on, they can stretch a little,” he says. “Let’s say you’re cutting labels to put on bottles, and they’re made out of a plastic material that keeps stretching. If you always cut at 12 inches and they stretch one mil each time, after a few hundred labels you’re cutting not at the edge of the label but in the middle. This can happen with fixed ratios. With electronic gearing, you can read the mark indicating the edge, electronically compare it to where it should be, and superimpose corrections on top of the gearing.”
Nevertheless, for simpler and less-strict positioning schemes, mechanical linkages are a reliable economical solution, allowing fewer motors in multi-axis systems. For very high torque and power, mechanical linkages are practically indispensable. Most applications require speeds quite a bit less than what a motor is capable of, and much of the motor power goes unused. Designs such as the outer-rotor configuration address this issue to a degree. For very high torque, however, the common (and economical) solution is to couple a smaller motor to a gear, belt, or chain drive.
In linear systems, a linear motor is the direct-drive solution. Although very expensive compared to a rotary motor driving a power screw, belt, or rack, linear motors, when linked with the proper controls, can produce uncompromising accuracy down to the sub-micron level.
Continue on Page 2
A tight rein
Whether directly driven or not, in a closed-loop system a digitally controlled motor helps keep the system in tune. A typical servo system includes an instrumented load, amplifier, controller, and motor. The inertia and dynamic response of the load are fed into a servo algorithm at the amplifier stage to control the dynamic response. To develop such a system, mechatronic engineering talents are key. Someone has to be able to assess the mechanics of the system and develop code to adjust the servo algorithm accordingly. Programming on-site, “onthe- fly,” is sometimes necessary to finetune a system.
“We had a situation where two engineers spent a week at a customer’s site developing code,” says John Guite, Manager of Controls Engineering with Compumotor, Rohnert Park, Calif. “It required somebody to go down there who had an understanding of mechanical systems....If I increase or decrease the inertia or the load on a particular axis, what will that do back at the servoalgorithm and to the dynamic response? How do I adjust for that in my servoalgorithm? An understanding of mechatronics is invaluable in enabling engineers to go on site and ‘prototype’ almost, and develop features as they go along.”
A servo loop controls position, velocity, and current to the motor. As a machine runs, the servo algorithm is essential in keeping the motor adjusted according to the load response or output.
“If, for instance, the motor is overshooting in one direction, you can apply current in the opposite direction to pull it in faster.” Guite explains. “But obviously, you have to know that it’s overshooting, you have to be able to model that and understand that. If it’s overshooting by ‘this’ amount, that means I have to add ‘this’ amount of current in a particular direction or polarity. But then you also have to know the dynamic effect of doing that. It’s a continuous process of controlling the motor and monitoring feedback, and if anything gets out of line from what you set, (the control) makes adjustments accordingly.”
As you add mechanical linkages between load and motor, the question of where to close the loop arises. The true load position can only be obtained by placing a sensor at the end of the line. However, the farther out from the motor the feedback device is located, the more lost motion and system dynamics are brought into the loop.
These mechanical errors contribute to delay and instability in the control process.
It is sometimes useful, therefore, to place multiple feedback loops, coming from, say, the motor output as well as the load. Taking several stages into account is a way to boost stability (by feeding back motor position) while still accounting for the load position. Positioning errors are easiest to correct within the inner (motor feedback) loop, and the burden of error correction in the outer (load feedback) loop can be minimized. (See “Course Audit,” PT Design October 2000.)
Computational input can further refine a system to reduce errors caused by slop in the mechanical linkages, eliminating the effects before they occur. Positional variance in a lead screw or gearbox, for example, can be recognized through testing and analysis, and linearized through electronic means. You can arrange a table to account for backlash, rigidity, and so on. By writing it into the control algorithm or feeding it through an open loop on top of the primary closed loop, you essentially predict position error and correct it ahead of the feedback process.
According to Tal, the key to such methods is software and programming. The hardware – the microprocessor – is extremely powerful and remains largely untapped. The trick is in developing electronic algorithms that bring out some of the hardware’s potential. For instance, today there are algorithms devised to cancel the effects of backlash, but the problem was often accepted in the past as largely unavoidable.
Motor control can improve system slop caused by mechanical linkages, but, Guite points out, there are other complications from adding mechanical components, such as vibration and resonance. Gearboxes, belt drives, any such mechanical linkage can contribute to vibrational disturbance.
Torsional vibrations can obviously lead to positioning errors at the output. As Guite explains it, motor control can help alleviate this some of the time, but there are limits. The problem is, many times the frequency will range on the order of kHz. “A control system has a very difficult time reacting that fast,” says Guite. “It’s above the bandwidth of the actual system.”
And the vibrations don’t have to be strictly torsional to cause a disruption in the positional control loop. A resonating gearbox, for example, may send high-frequency ripples down the drivetrain that are picked up by the sensing device (such as the encoder) at the output. If the control can’t correct for these signals, there’s not much to be gained by monitoring them – in fact they can be quite disruptive and contribute to the aforementioned instability. Therefore, it’s often desirable to eliminate them from the loop. This may be done electrically using notch filters that block specific frequencies so they won’t affect the control process.
In a servo loop, there is a need for highresolution position-sensing devices. More-than-adequate devices are available. “Feedback devices are particularly moving out,” says Guite, “driving some of the things that amplifier and control developers are working on – being able to handle these high-resolution feedback devices.” A position-detection system called SINCOS, for example, sends an analog signal consisting of sine and cosine waves that are decoded into exact resolutions. This references the absolute load position at power-up as well as during operation. Extremely high-res devices like this are often used with linear direct drives, with their near-definite positioning capabilities.
Puttin’ on the brakes
There are many situations where other components besides the motor impart corrections to the system. Web tensioning is a good example. Here, adjustments are made with brakes, and load tension is measured and regulated rather than output shaft position.
In a typical setup, brakes on an unwind roll resist the motorized take-up roll to maintain tension throughout the winding process. Sensors monitor parameters such as material roll diameter, weight of the roll, and web tension, and feed the information into the control loop to keep the braking torque properly tuned. While still an analog process in many cases, tensioning has become increasingly digital – increasingly mechatronic – depending on how delicate the application. Tom Oborn, a Control Engineer with Eaton Airflex, Cleveland, Ohio, explained the various reasons for using digital and microprocessor- based controls.
Coarse, heavy operations like steel coiling are not necessarily very precise – often a manually set tension controller will suffice. But some materials demand finely controlled tension, and achieving this requires sampling the response and correcting for it at a millisecond rate; usually a job for digital controls. Furthermore, a high level of correction management – minimizing overshoot and so forth – is inherent to digital control schemes. The resolution of the brake itself limits precision, with lower-torque brakes tending to have a finer resolution. Therefore, lighter applications are also likely candidates for achieving very high precision.
Acceleration and deceleration of the winding process may demand specialized adjustments. The ability to automatically vary the control parameters during these phases lies with digital control. And in some situations multiple webs of different materials may be processed together, a condition usually requiring a more sophisticated controller with several channels. Programmable digital controls also enable use of presets that correspond to different applications, a valuable option if numerous lines of product are processed at different times.
Continue on Pag 3
Actively controlled bearings are another prime example of “smart” components. Magnetic bearings are one variety of active-control bearings. They are not dissimilar from electric motors, albeit ones that actuate radially. In active bearings, coil current induces an attractive magnetic force that suspends the rotor, enabling noncontact shaft support. While operating according to a relatively simple principle, such bearings rely on a control loop to maintain stability and allow radial position adjustment. Without the controls, even a small radial displacement of the rotor could compound the influence of the magnetic field, leading to collapse of the air gap between rotor and stator.
During operation, the position of the rotor is continually monitored and corrected by the controls, which may be analog but are often digital. Sensors located radially around the rotor detect position. Position feedback is compared to the control program and the calculated correction signal is sent through an amplifier to the actuator. This occurs on a millisecond schedule, allowing magnetic bearings to accommodate shaft speeds higher than 100,000 rpm.
For a given application, the operating dynamics can be analyzed and used to develop the control algorithm or transfer function. As with many systems, this “smart” component’s control function can be implemented through hard-wired circuitry for analog control, but the hardware is essentially tailored to the specific application, and altering this type of scheme is often impractical. In the digital realm, though, the transfer function can be programmed and changed with little difficulty. Digital systems also allow openloop control, where external signals can be incorporated along with the rudimentary closed-loop sensor-based control.
There are magnetic bearings for thrust loading as well, similar in principle to the radial versions except for the geometry and physical orientation of the elements. Some hydrostatic fluid bearings are also candidates for active control.
Magnet in the middle
Mechanical components can often handle a positioning or braking job just fine, but electronics can add another level of control precision. Lord Corp., Cary, N.C. offers an adaptive interface between electronic control and mechanical components with its magnetorheological (MR) fluids. These are somewhat freeflowing fluids, such as oil or water, containing small iron particles. With a magnetic field in the fluid’s proximity, the particles are magnetized and the viscosity of the mixture changes, creating a semi-solid.
This technology was discovered in the 1940s, but was left undeveloped because there were no suitable computer controls at the time. Lord picked up research and then began commercializing the technology in the mid-nineties.
The company recently introduced an MR brake for pneumatic linear actuators. One such component, a servo pneumatic system, can stop at multiple programmed points. Normally, a pneumatic cylinder is not highly precise – it just bangs back and forth. MR brakes, however, change everything. Because air pressure in the cylinder is not great enough to shear two plates held together by activated MR fluid, you can instantly stop the cylinder anywhere as programmed into the electronic controls. Plus, you can vary the degree of braking power for strict velocity control.
MR fluids can also be used in suspension dampers to dissipate energy. In one of its first commercial applications, Lord designed a complete mechatronic truck seat damper to prevent back injuries to drivers.
“The passive shock is designed so for most conditions, it’s a smooth ride. But, if you were to design for the extreme conditions, you wouldn’t have a good ride at most conditions. It’s a compromise because you never know what your inputs are going to be,” says Lynn Yanyo, manager, Lord sales and marketing.
With the truck seat damper and in automotive suspension semi-active dampers, a sensor knows the seat position and velocity relative to the cab, or, where the car body is relative to ground. If a shock absorber is accelerating to the top of its limit, the MR damper will quickly slow it down and prevent it from topping out. The damper remains inactive until it is needed. It automatically recalculates the damping need at a rate of 500 times a second.
Incidentally, there are mechanically controlled semi-active dampers on the market. By adding a stepper motor, the hole in the piston through which fluid travels can be adjusted to influence the degree of damping. But, Yanyo points out, moving parts and electronics embedded in oil add complexity to the design and compromise robustness.
Continue on Page 4
In the past, Lord has investigated electrorheological (ER) fluids as well. But these require an electric field on the order of kilovolts, which creates safety concerns. And, any bit of water in the fluid turns it into a conductor, thereby diminishing the required potential. Also, the highest force that could ever be generated with ER fluids is a factor of ten less than what can be generated with the same volume of MR fluid, according to Yanyo.
The MR devices have proven reliable. “Usually other things mechanically fail before the fluid does in controllable brakes,” says Yanyo. This allows use in applications like prosthetics. Lord engineers have worked with German-based Biedermann Motech to apply MR damper technology in a prosthetic leg that lets amputees walk with a more natural gait. Often, amputees have to train their muscles to work in new ways or adjust the prosthesis to perform certain motions, but MR dampers and electronic controls eliminate this.
The leg has strain gauge sensors in the artificial shin that measure the direct axial force and the bending moment. There are also sensors at the pivot point in the knee structure that measure angle, speed, and direction of travel. Onboard electronics take sensor inputs and mimic the natural human thought process to determine what kind of action the person is performing, whether it be walking up stairs, jogging down a hill, or riding a bike. The electronics control a fluid damper in the knee to exert the necessary force for the appropriate amount of time so the leg has a natural swing. For example, if a person is going down stairs, it stiffens for about 50 millisec to support the person’s weight as the heel strikes the stair. Then it releases so the knee will bend as the person strides through. As the foot lifts and swings back, the damper releases even more for a natural swing-through.
The impact of mechatronics is evident all around us; from a simple 35-mm camera to multi-task machines in today’s factories. Thus, universities are grooming a new crop of students and researchers to develop the systems necessary for this mass integration of digital intelligence.
Traditionally mechanical research academia are tackling mechatronic projects these days. For example, Caltech’s mechanical engineering department is developing microthrusters for small spacecraft propulsion systems. Adjusting propulsion in small increments by lighting different numbers of thrusters is difficult, says David Lewis, a TRW engineer who invented the system with Caltech professor Eric Antonsson. Their digital propulsion rocket chip prototype contains 15 individual thrusters on a chip array.
Engineers, and soon to be engineers, taking on such projects need diverse knowledge in order to thrive as machine designers. Many colleges are developing mechatronics-specific courses and programs to ensure their students’ readiness.
On mechatronics, Dave Carlson, Lord Corp., says, “It’s less well-known or less generally accepted here in the United States.” He explains, “I think it’s more recognized in some of the educational structures in Germany, and in some senses it’s a buzzword, and it’s caught on a little more in Europe.” says Carlson. Nonetheless, the term and concept are creeping into U.S. curricula.
According to the Colorado State University Web site, more than 20 colleges offer Mechatronics courses or programs.
One such university, San Diego State, has a large lab devoted specifically to Mechatronics projects. The lab has custom made machinery, an array of sensors, and even a 10-ft. high roller coaster. Another, Ohio State, has developed a core mechatronics program, with courses in microprocessor based systems, logic design, circuit design, and machinery systems, as well as broad mechatronics courses.
Kettering University, Flint, Mich., has had a mechatronics program since 1998. It consists of an introductory course, a more elaborate mechatronics course, and a mech lab. Kettering’s program focuses on the use of microcontrollers, as well as sensors and actuators, which are important components in any mechatronic applications.
“It’s amazing how fast students can embrace the microcontroller and start making mechatronic devices,” says Jeff Hargrove, a professor at Kettering.
Kettering plans to add a mandatory mechatronics course to the curriculum that will be taken by all sophomores to help prepare them for the higher level courses as well as give them valuable knowledge about mechatronics.
“Examples of mechatronics can be found everywhere, in all aspects of our lives, and are certainly going to be a huge part of engineering from this point forward,” says Hargrove.
Rensselaer Polytechnic Institute offers two senior-level mechatronics courses. They look at design principles, modeling, analysis and control of dynamic motion systems, mechatronic component selection, electronics, and real-time control programming.
The university also helped found TCG, a company devoted to teaching practicing engineers about mechatronic design. TCG conducts customized, model-based workshops. Continuing education is important because, “A mechatronic system design will only be cost effective if the control aspect is designed in from the beginning. Adding controls as an afterthought will only add cost,” according to Dr. Kevin Craig, Rensselaer professor.
Even universities that do not have a specific mechatronics track are incorporating a cross-disciplinary approach. Mechanical engineering students may, for example, be required to take courses on digital signal processing and control systems.