The Art and Science Behind CNC Motion Control
Article From: EDM Today, Spring 2012 Issue
Editor’s Note: Motion control is a little discussed, often ignored, essential ingredient to successful CNC technology. The motion control components and logic incorporated into a CNC control determine how the machine gets from point A to point B. That “how” is a major factor in determining the contouring accuracy, response time, and surface finish capabilities of any CNC machine. The following technical article has been condensed by Don Miller from a technical presentation by Yuji Kaneko (previously president of Sodick America) now president of Sodick Co., Ltd. Don is the distributor for Sodick in Northern California, has been involved with EDM for 27 years, and has known the Sodick America folks since the division was founded 1991. As has been the case with previous Expert’s Corner articles, this in-depth discussion is substantially more detailed than a typical magazine article. However, your reward for wading through the detail will be a valuable insight into a technology that, in addition to generator technology, is critically important to the successful application of EDM technology to our industry.
At the southern end of the San Francisco bay area lies a valley not long ago fi lled with orchards and blessed with gentle seasons. History will someday defi ne this area as the center of a technological universe unlikely to be surpassed and its infl uence on mankind, unmatched. Major businesses that will forever change human history such as Intel, Facebook, Google, Apple, Hewlett-Packard and many others share this valley. More than a geographic area it really represents a mentality celebrating the genius of innovation. This is Silicon Valley.
San Jose, CA the capital of Silicon Valley is home to Sodick America for the last dozen years. Sodick America is one of three groups of Sodick Co., Ltd’s research and development efforts. In the United States and the Americas the better known Sodick Inc. is responsible for the sales and support of Sodick products. In 1999 Sodick Co. Ltd. encouraged by the successful introduction of its linear motor servo system, sought out the area and the people to continue the refinement of this technology. Sodick America’s primary mission is to further develop the motion controller and after four generations of the K-SMC™ (Sodick Motion Controller) over 25,000 Sodick machines share this technology. Practically all Sodicks produced since 1999 use linear motors and since that time no other EDM manufacturer has produced as many machines. In 2011 over 2800 Sodick’s were produced in one of four Sodick factories. No doubt the linear motors launched an incredibly successful series of machines all controlled by a version of the K-SMC motion controller developed here at Sodick America in the U.S.
Since the start of Sodick America, Yuji Kaneko has been its CEO. He is an electrical engineer working for Sodick for the last thirty years and until recently the Senior Managing Director of research and development for the entire Sodick organization. Mr. Kaneko has now been promoted to CEO of Sodick Co. Ltd. which will only reinforce Sodick’s commitment to leading edge core technologies. Mr. Kaneko lived with his family in California for six years and has a unique appreciation of Silicon Valley and the American culture. The new leader of Sodick America, an engineer living in California since 2000 is Koji Yoneda. Mr. Yoneda has worked for Sodick for 16 years and now manages a staff of American and Japanese engineers.
Currently Sodick America’s main product is a motion controller called the K-SMC-SiLink and it is capable of controlling up to eight axis and is installed on all Sodick wire EDM, sinker EDM and machining centers. Basically Sodick America has re-invented the motion controller incorporating Sliding Mode Control technology into the latest products. The very latest product, the K-SMCM4 Link will debut at the upcoming IMTS and JIMTOF shows can control up to 32 axes simultaneously and with this increased capability a wider variety of automation applications. The new motion controller will allow processor upgrades as the technology further evolves keeping Sodick products state-of-art for many years to come. As the following in-depth technical article will show, truly revolutionary motion control innovations have taken place to further the speed and accuracy of linear motion controlled equipment.
INSIDE THE LATEST SODICK MOTION CONTROLLER:
For those who may appreciate the finer points of machine tool technology, the development of the Linear Motor on a machine tool is truly a revolutionary concept. No serious observer would deny the simplicity of the linear motor as it eliminates the ball screw and the inherent backlash, reduces friction and increases acceleration. Despite the fact that the concept of linear motors have been around for decades, it wasn’t until 1999 that Sodick could control a linear motor to the degree necessary on a modern machine tool. The very first machine tool product with linear motors was the Sodick CNC Diesinker which set the stage for Sodick to become the world’s largest producer of EDM equipment. All modern Sodick products use the Sodick developed motion controller known as the K-SMC and it is designed to optimize the unique characteristics of the linear motor. This article will highlight the key components and compare the performance of Sodick’s Slide Mode design and the conventional PID (Proportional- Integral-Derivative) design. The linear motor is a very different animal and therefore needs a very different way to control it. By utilizing Modern Control Theory such as Sliding Mode and Disturbance Observer as well as PID control, Sodick is able to apply this science to linear motors. At the same time Sodick sought to simplify the control hardware by using a compact CNC controller, reducing wiring and increasing speed and reliability while keeping the costs comparable to conventional motion control technology. By using the one board CNC by COM Express PC, a well know processor within the industry, cost could be held in line while fiber optics allowed fewer wires and even faster communication.
THE SLIDING MODE
Briefly explaining the features of position and velocity by Sliding Mode you’ll see that Fig. #1 shows the position on the horizontal line and velocity on the vertical line. The two paths chart out the movement of a PID trajectory (black line) vs. Sliding Mode trajectory (green line). In a PID control, movement slows down as it gets closer to the targeted point and reverses its movement once it goes over its targeted point and repeats this motion until it actually converges at its targeted point (black line). Sliding Mode approach is different, as the movement is directed to the Hyper Plane (shown as dotted line). Once movement is on the Hyper Plane, it goes to the targeted point with the linear gain K1 by changing the non-linear gain K2. The key feature of Sliding Mode is by switching over Gain (K2) movement is constrained on Switching Plane (Hyperplane). This would cause excessive vibration for an AC rotary motor with a ball screw and likely increase backlash. There is no backlash on a linear motor therefore Sliding Mode Control concept is ideal linear motor control. There are however Pro’s and Con’s in designing a motion controller incorporating the Sliding Mode concept. In favor of Sliding Mode controller is that it is relatively easy to design, high resistance to outside disturbance as it has superior robustness, better image by designing in the time domain and non-oscillatory transient response. On the other hand it is not suitable for a rotary mechanism as it cannot be designed by frequency domain. Also it cannot define the frequency characteristic response and it is very complex to design by frequency shaping.
SLIDING MODE CONTROL SYSTEM
In referring to Fig. # 2, you’ll see that the K-SMC motion controller consists of the Target Position Generator, Sliding Mode Controller, Disturbance Observer, State Observer and State Variable Detector. The Motor Controller is connected to the controlled systems that consist of a Power Amplifier, Servo Motor and a Linear Position Scale.
The CNC interprets a NC program and calculates desired position data (P), velocity data (V) and acceleration (A) to the Target Position Generator on the Motion Controller board. The CNC can also provide data indicative of other state values such as pitch error.
The Motion Controller determines a target position (r) based on desired position data (P), velocity data (V) and acceleration (A) and revises the target position (r) so as to provide a control input (u) to the Controlled System through the Power Amplifier.
The control input (u) is a controlled current for a servomotor as the q-axis current (Iq). So, the Controlled System as shown on in Fig. #2 might actually be a movable member of a machine such as a work table, driven by a rotary servomotor with a ball screw or as in the case for a Sodick ma
chine, a linear servomotor. A position detector on the machine such as a rotary encoder or linear scale is used to measure the position of either the ball screw’s rotational position or with a linear scale, the actual position of that movable member. (Most Sodick machines use a Heidenhan linear glass scales of .0000004”resolution or .01 micron).
The State Variable Detector receives the target position (r) and the control input (u) from the Motion Controller and the measured position () from the position detector. At that point the State Variable Detector will provide the variable (U) to the State Observer.
A major sub-component of the Sliding Mode Control System is the Disturbance Observer (Fig.#3). The job of the Disturbance Observer is to estimate the disturbance torque based on the difference between the torque without disturbance calculated from the command of q-axis current (Lq) and the torque including the disturbance calculated from estimated velocity. The multiplier multiplies the estimated disturbance torque by 1/Kt to generate the estimated disturbance. The estimated disturbance is input into Sliding Mode Control.
PID CONTROLLER VS. SLIDING MODE CONTROLLER: IS THERE A DIFFERENCE?
The whole point of making a motion controller to optimize linear motors is that there would be a significant performance difference to justify the effort
. The following graphs and information will show the superiority of the Sliding Mode controller vs. a conventional PID controller for linear motors. Two performance parameters were measured to define an actual performance difference, Step Response and Parabolic (Sine wave) Response. The Step Response performance was given a condition of a Step Size of 2000 [counts] and Step Time of 500[msec]. Parabolic Response performance was given a condition of a Step Size (amplitude) 40000 [counts] and a Step Time of 300 [msec]. (One count is being defined as 0.03125 um in the following test).
Referring to Fig. #4.1 (PID Controller – Step Response) and Fig. #4.2 (Sliding Mode – Step Response) some significant performance differences are immediately apparent. When command position and actual position are shown, the transient response of “Overshoot” showed a difference 11.5% (PID) and >1.25% (Sliding Mode). In other words the PID controller overshot by factor of nearly 10:1. The transient response of “Setting Time” was 43.8 msec (PID) vs. 11.0 msec (Sliding Mode). When factoring in the other transient response
times of “Delay Time” and “Rise Time” the Sliding Mode Controller’s step response time was 24.5 msec vs. 53.8 msec. for the PID controller. In other words the Sliding Mode Controller is able to return the actual position to the command position is less than half the time.
Parabolic response is significantly improved as well (Fig. #5). The blue line is the targeted Sine wave, the red line the following error by PID control and the yellow line is the following error by Sliding Mode.
For a machine tool a well established test is for roundness; the ability to generate a circle as accurately as possible. For the most part this is a test of the mechanical properties of a machine like the amount of ball screw backlash or the accuracy of the bearing surfaces. These two Figures # 6 & 7 shows the performance of a PID controller vs. Slide Mode controller on the same machine tool. The circle generated had a radius of 10mm and the velocity was 3m/min. The results are fairly predictable particularly when we have seen the results of the earlier graphs. The PID controlled test measured a deviation of about +/- 2.5 um the Sliding Mode controlled deviation was +/- 1 um. So in terms of speed and accuracy the Sliding Mode controller shows superior results compared to PID.
The challenge for machine tool builders has always been and will always be to go faster with greater accuracy. In September of this year in Chicago the IMTS will be held displaying products from virtually all machine tool builders. No doubt the relatively new innovation of the linear servo motor will be available on a wide variety of machines including milling and turning centers. Linear Motor equipped machines will likely come at a premium relative to their standard models. As of this writing, four major EDM manufacturers have offered linear motor equipped EDM machines (either wire or sinker or both) and the first of course was Sodick way back in 1999. As this article suggests, it is not simply a matter of bolting the linear motors on a machine and expecting exceptional results as it will require a re-thinking of how a motion controller functions with this new servomotor technology. Sodick has a substantial head start in this area with well over 25,000 EDM machines using linear motors thus far produced. In the future many more machine builders will adopt linear motor technology as its advantages are simply too great. The choice of Sodick to create its own controller years ago to optimize the linear motor performance seems very timely now.
Editor’s Note: I’d like to express our profound appreciation to Yuji Kaneko for contributing this material and to Don Miller for his diligent efforts to condense Yuji Kaneko’s presentation into article format.