Hierarchical Terminal Sliding Mode Contouring Control for Five-Axis CNC Machine Tools With Combined Contour Error Estimation

Imagine you are trying to paint a perfect, swirling masterpiece on a curved wall using a robotic arm. If the robot just tries to move its arm in a straight line, the paint will look jagged and messy. To get a smooth curve, the robot needs to know exactly where it is relative to the line it's supposed to follow, not just how far it is from its starting point. This is the challenge of "contouring control" in the world of high-tech manufacturing. Machines that carve complex shapes—like airplane wings or medical implants—use five different motors working together: three to move the tool left, right, up, and down, and two to tilt and twist the tool so it hits the surface at the perfect angle. The problem is that these five motors are like a five-person rowing team; if one person lags even a tiny bit, the whole boat drifts off course. The goal is to keep the cutting tool glued to an invisible, perfect 3D path, correcting for any wobble instantly, even when the path twists sharply or the machine is moving very fast.

This paper introduces a new "brain" for these five-axis machines, called a Hierarchical Terminal Sliding Mode Contouring Controller. Think of it as a two-tiered coaching system for the robot. The lower level is a tough, no-nonsense coach for each of the five motors individually, ensuring they stop moving exactly when they are supposed to, without shaking or jittering (a problem known as "chattering"). The upper level is the master strategist. It constantly calculates the "contour error"—the tiny distance between where the tool actually is and where the perfect path is. The tricky part is that this error is hard to measure because the tool often lags behind the command, like a runner who is still catching their breath after a sprint. The authors created a clever new way to estimate this lag, combining two different math tricks to get a super-accurate reading. In their computer simulations, this new system proved to be the most accurate, reducing the tool's deviation from the perfect path by about 14% to 19% compared to older methods, especially when the machine is racing along a curved path.

The Problem: The Lagging Dancer

Imagine a dancer trying to follow a glowing line on the floor while spinning. If the dancer is fast and the line curves sharply, their feet might not land exactly on the line; they might overshoot or lag behind. In a five-axis machine, the "dancer" is the cutting tool, and the "line" is the path the computer wants it to follow. The machine has five "legs" (motors) to move this tool. If the legs don't coordinate perfectly, the tool creates a bumpy, inaccurate cut.

The biggest headache for engineers is that the tool often lags behind the command. If you just look at the tool's position and draw a straight line to the target, you might think the tool is way off course. But in reality, the tool is just a little bit behind in time, not necessarily far away in space. A simple measurement would mistake this "time lag" for a "distance error," causing the machine to over-correct and make the cut even worse. The paper argues that you cannot fix the path if your ruler is lying to you.

The Solution: A Two-Level Coaching System

The authors propose a solution that acts like a two-level coaching system, designed to handle both the individual muscles and the overall dance routine.

Level 1: The Muscle Coach (Lower Level)
First, the system needs to make sure each of the five motors stops exactly where it should without shaking. The authors use a method called "Non-Singular Terminal Sliding Mode" (NTSM). Imagine a car braking. A normal brake might let the car roll a little bit before stopping, or it might jerk the car to a halt. This new "brake" is designed to bring the car to a stop in a guaranteed, short amount of time, and it does it smoothly so the car doesn't shudder. They also add an "integral action," which is like a coach who remembers if the car has been drifting slightly to the left for a while and gently nudges it back to the center to fix that drift permanently. This level ensures that every single motor is under tight control and stops exactly on time.

Level 2: The Choreographer (Upper Level)
While the muscle coach handles the individual motors, the choreographer looks at the big picture. Its job is to manage the "contour error"—the distance between the tool and the perfect path. To do this, the choreographer needs a perfect estimate of where the tool is relative to the path.

The paper introduces a "Combined Contour Error Estimator." This is the paper's secret sauce. Previous methods tried to guess the error in two different ways:

1. The Delayed Frame: This method looks at where the tool was a split second ago to correct for the lag. It's good at handling the "time delay" but can get confused on very sharp curves.
2. The Curvature Circle: This method looks at the curve itself and tries to fit a circle to it. It's great for sharp turns but can get confused if the tool is lagging behind.

The authors' innovation is to fuse these two methods. They take the "Delayed Frame" to fix the time lag and then use the "Curvature Circle" to refine the position on the curve. They found that doing this just two times (two iterations) is enough to get the error down to less than 1 micrometer (which is thinner than a human hair). This combined estimator is so accurate that even when the tool is lagging significantly, it doesn't get fooled into thinking the error is huge.

The Results: Smoother, Faster, and More Accurate

The authors tested their new controller in a computer simulation against three other methods: a standard linear controller, a version without the "muscle coach" refinements, and a version without the "choreographer" (the contour hierarchy).

Here is what they found:

• The Winner: The full hierarchical system (HTSMC) was the most accurate. It kept the tool-tip contour error at a tiny 0.65 micrometers on average.
• The Comparison: The standard linear controller had an error of 1.74 micrometers, and the version without the contour hierarchy had 0.74 micrometers.
• The Speed Factor: The real magic happened when they increased the speed. As the machine moved faster, the "lag" got worse. The new system's advantage grew with speed. At three times the normal speed, the hierarchical system reduced the error by 19% compared to the version without the upper-level choreographer. This proves that the system is specifically designed to handle the difficult, high-speed, high-curvature scenarios where other machines struggle.
• No Shaking: A common problem with these types of controllers is "chattering," where the machine vibrates violently. The authors used a smooth, continuous control law that kept the machine movements smooth, avoiding the vibration that usually comes with this type of math.

What This Means

The paper demonstrates that by separating the problem into "fixing the motors" and "fixing the path," and by using a smart, combined way to measure the error, you can get much better results. The system is robust, meaning if you change the settings by 20%, it still works well.

However, it is important to note that these results come from computer simulations, not a physical machine in a factory. The authors acknowledge that the next step is to test this on a real five-axis machine to see if it holds up in the real world with actual metal chips and vibrations. But for now, the math suggests that this two-level, combined-estimator approach is a powerful way to make five-axis machines carve perfect shapes, even when they are moving fast and turning tight corners.