Towards Track Height Control in Additive Manufacturing on Porous Substrates: An Approach for System Identification and Controller Selection
summary
The gist
Directed Energy Deposition (DED-LB) on porous substrates presents a challenge in achieving consistent track heights during deposition…
In short
The work addresses inconsistent track heights during Directed Energy Deposition (DED) on porous substrates, which affects final part geometry. It proposes a conceptual approach using system identification to model process dynamics and compares three control strategies: feedforward, feedback, and lookup table methods. The initial study favors feedforward control as a starting point.
Key concepts
- System Identification
- This technique adapts a mathematical model to the specific DED process being studied. Instead of testing many single tracks, the method varies process parameters along one track to capture both its stable values and how it reacts dynamically to changes in those parameters. This creates a richer dataset for building accurate predictive models.
- Feedforward Control
- This controller adjusts the scanning speed based on pre-measured pore volumes. It compensates for expected disturbances offline by pre-computing the machine's G-Code before deposition begins. This approach is advantageous because it avoids needing online sensors or heat-resistant laser line triangulation during the actual manufacturing process.
- Feedback Control
- This method allows for real-time adjustments by comparing a desired track height reference to the actual measured track height. The difference (error) is fed into the controller, which then calculates new process parameters to correct the deviation. This enables online reaction to changes in track height, width, or temperature.
- Lookup Table Method
- This strategy uses pre-measured data points and linear interpolation to quickly determine the best process parameters needed to achieve a target track height. While fast and real-time capable, it requires a Design of Experiment (DoE) phase to generate enough accurate data points for reliable interpolation.
Terminology used across episodes
This episode discusses
- Towards Track Height Control in Additive Manufacturing on Porous Substrates: An Approach for System Identification and Controller Selection · Paper Radio
The paper
Towards Track Height Control in Additive Manufacturing on Porous Substrates: An Approach for System Identification and Controller Selection · Read on arXiv
Institute for Measurement and Sensor Technology, RPTU University · Institute of Assembly Technology and Robotics, Leibniz University Hannover
Additive manufacturing is transforming the creation of materials and parts, enabling the design and manufacturing of multifunctional parts with tailored properties. One compelling example is the combination of a porous core and a non-porous surface, which offers significant advantages in terms of weight reduction, material efficiency, strength and wear resistance. Directed Energy Deposition (DED) processes enable the transition from porous core to non-porous part surface by closing the pores. However, the surface-layer of porous substrates presents a substantial challenge to achieving consistent track heights during DED, directly impacting the final part geometry and performance. Therefore, precise control of the deposition process is crucial for realising the full potential of the multifunctional parts. In order to address this challenge, this work presents a conceptual approach for track height control during Laser-based Directed Energy Deposition (DED-LB) on porous substrates. First, the DED-LB process and the associated measurement setup are explained, along with a discussion of their limitations. The focus then shifts to the control problem, beginning with system identification to characterise the process dynamics. Subsequently, a selection of control strategies is investigated, as well as their compatibility with the process. This research provides a foundational step toward reliable additive manufacturing on porous substrates, and thus toward producing multifunctional parts.
Transcript
Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.
Rosa: We're looking at the paper: "Towards Track Height Control in Additive Manufacturing on Porous Substrates".
Dev: Here's the gist of the paper. Directed Energy Deposition (DED-LB) on porous substrates presents a challenge in achieving consistent track heights during deposition, which directly impacts final part geometry and performance The work presents a conceptual approach for track height control during Laser-based Directed Energy Deposition (DED-LB) on porous substrates.
Rosa: Process Description and Challenges Directed Energy Deposition (DED) processes enable the transition from porous core to non-porous part surface by closing the pores However, manufacturing parts that meet quality standards remains challenging for DED, particularly when building on porous surfaces A key challenge when working with porous substrates are geometrical distortions along the deposited tracks as they traverse multiple layers.
Dev: When the melt pool and track manufactured by the DED-LB process meet surface pores, a controller must compensate for the disturbances in height.
Rosa: System Identification Methodology System identification enables the adaption of a process model towards the specific process being investigated Conventionally, parameter identification in AM involves manufacturing single tracks, each with a different parameter configuration.
Title and authors: Rosa: So we're looking at this paper by Schatull et al. titled "Towards Track Height Control in Additive Manufacturing on Porous Substrates: An Approach for System Identification and Controller Selection." It tackles the problem of making sure those tracks stay consistent in Directed Energy Deposition, especially when you are building on a porous surface.
Dev: Yeah, it’s about that geometric distortion along the deposited tracks as they go up through multiple layers. It sounds like they are proposing a way to control the height based on what's happening right at the deposition point.
Taro: I'm curious if this control setup would actually work when you take it outside of a perfectly controlled lab environment, you know, in some real-world scenario.
Rosa: That’s exactly what I want to ask. The paper looks at different ways to handle that height variation, setting up a comparison between feedforward control and feedback control and a lookup table method.
Dev: Right, so they are looking at three distinct control approaches for this DED-LB process on porous substrates. Feedforward adjusts the scanning speed based on premeasured pore depths, while feedback reacts online to the current track height compared to a desired reference height.
Taro: And that sounds like it needs some kind of measurement happening in real time, which brings up my concern about the hardware interface and communication latency.
Rosa: That’s a fair point. For feedforward control, they mention it's limited because it relies on premeasured pore depths and can't react to unexpected changes during the process itself. But feedback control allows it to react to temperature too, as long as you have open communication interfaces with the DED-LB machine.
Dev: Exactly. The paper shows a desired reference height compared against the current track height, and that difference is what goes into the controller to calculate new process parameters, which means it’s an online control loop.
Taro: So if you have a disturbance, like a sudden change in powder flow or heat input, how fast does this feedback loop actually need to react? That's where I get worried about real-world failure modes.
Rosa: The paper dives into system identification to adapt a process model to the specific process they are looking at. They suggest that instead of just identifying parameters from single tracks, it's better to vary the process parameters along a single track in a step-wise manner.
Dev: That variation helps capture both the stationary track values and the dynamical behavior, which should give them a richer dataset for building their control model.
Taro: So they are trying to get this model right so that when things go sideways, the system doesn't just crash or produce garbage geometry? What kind of disturbances are they focusing on most?
Rosa: They focus on the disturbance caused by the melt pool and track meeting surface pores, and how a controller compensates for that height disturbance. It’s about keeping the final part geometry consistent.
Dev: The paper also looks at a lookup table method, which relies mainly on inherent data from their Design of Experiment process, using linear interpolation to find the optimal process parameters to hit that desired track height reference.
Taro: A lookup table sounds fast and real-time capable if you have enough data points, but it depends entirely on how good that initial experiment design was. If the DoE isn't thorough enough, the look-up table will just be wrong.
Rosa: That’s the caveat mentioned in the paper. So they are comparing these three ways—feedforward, feedback, and lookup table—to see which one is most robust for this specific DED-LB application on porous substrates.
Dev: The conclusion suggests that feedforward control is the initial approach because it doesn't require adapting the machine setup at all, which simplifies things initially. But they clearly see the potential for developing a feedback controller next, given this enhanced knowledge of the process dynamics.
Taro: If feedforward is just a starting point, what's your view on when you absolutely need that online feedback capability versus sticking with something simpler for faster manufacturing?
Rosa: I think it’s about balancing the complexity against the required precision. For reliable additive manufacturing on porous substrates, this paper suggests that moving toward a feedback controller is the next logical step to handle those online variations.
Dev: The overall implication for me is that understanding this system identification and control selection process helps us move closer to a more reliable manufacturing process where we don't have to constantly recalibrate based on every little surface variation.
Taro: For the end user, it means potentially better quality parts without needing constant manual intervention or expensive machine adjustments for every layer.
Rosa: So, that’s the gist of "Towards Track Height Control in Additive Manufacturing on Porous Substrates: An Approach for System Identification and Controller Selection." They show how to use system identification to build a model, test different control strategies like feedforward and feedback, and conclude that it's a good first step toward reliable additive manufacturing on porous substrates.
Dev: It gives us a clearer path forward for designing the real-time loops needed in these complex processes.
Taro: And it sets up the groundwork for testing those online feedback controllers in more challenging, non-lab environments.
The paper's summary: Rosa: So, we just talked about how this paper looks at different ways to control track height during Directed Energy Deposition on those porous metal parts.
Dev: Yeah, and what they really did was compare three main control strategies: feedforward control, which adjusts the scanning speed based on pre-measured pore depths; feedback control, which reacts online by comparing a desired height to the current height; and a lookup table method that uses pre-measured data points for fast, real-time parameter selection.
Rosa: And they used system identification to build a model of the process first. They found it helpful to vary the process parameters along just one track, not just looking at single tracks in isolation. That way you get both what happens when things are steady and how the whole thing reacts when you change something.
Dev: That variation is key because it gives them a much richer dataset for building that model than if they only looked at one fixed setting per track. It helps them understand the dynamics of how the process itself responds to those parameter changes.
Rosa: The paper seems to conclude that feedforward control is a good starting point because it doesn't require you to change the machine setup right away, which saves effort upfront. But they see a path forward with feedback control next, because that’s where you get the ability to react when things go wrong during deposition.
Dev: I agree on that second step. The feedback loop lets you react to changes in track height, or even temperature, as long as you have the right hardware interfaces set up for real-time communication with the machine. That brings up how much computing power and communication stability is needed for that kind of online control.
Rosa: Right. And they also looked at that lookup table method which uses linear interpolation based on their experimental data to find the best process parameters quickly, but they pointed out it needs a solid Design of Experiment setup first to generate enough reliable data points for that interpolation to be accurate.
Dev: The limitation there is that if the initial experiment design isn't thorough enough, the lookup table controller won't give you what you need to hit your desired track height reference consistently. It’s fast and real-time capable, but it depends on the quality of that upfront data collection.
Rosa: So, to summarize, they showed how to identify the process dynamics by varying parameters on a single track, compared feedforward control which is setup-based and feedback control which is online reaction based, and a lookup table method that relies on pre-calculated data points.
Dev: The implication for anyone building these systems is that understanding this identification step helps you choose the right controller strategy for your specific DED process on porous substrates. It’s about picking the right tool for the job before you start building those complex control loops.
Rosa: And what this means for people just listening to this? It means that when you're trying to build something with additive manufacturing that needs high precision, especially on tricky surfaces like porous metal, you can't just rely on a single method. You have to understand the system dynamics first so you can choose between setting things up ahead of time or letting the machine react in real time.
Dev: Exactly. It’s not just about picking a controller; it’s about knowing what kind of information you need to feed into that controller and how fast your hardware can deliver it. Now, if we want to see how well these kinds of control systems perform when you take them out of the perfect lab setting, that's what we're looking at next.
The paper's improvements: Rosa: So, we just talked about how this paper compares feedforward control, feedback control, and a lookup table method for controlling track height during Directed Energy Deposition on porous substrates.
Dev: Yeah, and what they suggested as improvements was making the system identification step smarter. Instead of just looking at single tracks to figure out the process parameters, they propose varying those process parameters along a single track in a step-wise way.
Rosa: That variation is important because it captures both the steady track values and how the whole system responds when you change something, which gives them a much richer dataset for their model. It’s like testing different speeds while simultaneously changing your powder flow rate to see how they interact.
Dev: Right. That richer data helps them build a better model for when things actually happen in the real process, not just on the paper. And that improved model is what lets you design a more accurate control strategy for those online loops we talked about earlier.
Rosa: They also highlighted that feedforward control is a good initial step because it doesn't require changing the machine setup right away, but they strongly push for developing a feedback controller next because that’s where you get the ability to react when things go wrong during deposition.
Dev: I agree on that. The feedback loop lets you react to changes in track height or even temperature, which is crucial for real-time control, provided you have the necessary open communication interfaces with the DED-LB machine running at a high enough loop rate. That brings up the safety case for that kind of online reaction.
Rosa: And they also talked about a lookup table method, saying it’s fast and can be real-time capable if you have good initial data from a Design of Experiment setup, but the limitation is that if that initial experiment design isn't thorough enough, the interpolation to find the desired track height reference won't be accurate.
Dev: So they are essentially saying you need to invest in good upfront data collection for any model-based control approach, whether it’s a lookup table or a feedback controller. It’s not just about having the right software; it's about getting the ground truth measurements right first.
Rosa: What this changes for us is understanding that for reliable additive manufacturing on these complex surfaces, you can't just pick one control method and assume it works everywhere. You have to see which approach fits your specific process noise and required response time.
Dev: Exactly. It shifts the focus from just building a controller to first mastering the system identification—making sure you actually know what the process is doing before you try to tell it what to do. Now, if we want to see how well these kinds of control systems perform when you take them out of the perfect lab setting, that's where we need a testbed.
Conclusion: Rosa: So to wrap up, we looked at how to control track height during Directed Energy Deposition on porous substrates using feedforward, feedback, and lookup table methods.
Dev: We established that system identification by varying process parameters along a single track is key because it gives you the right data for the model before you try to build those online loops.
Rosa: And we saw that while feedforward control is a good starting point because it simplifies things initially, the next logical step for reliable manufacturing is developing that feedback controller.
Dev: That’s true. The paper shows how to transition from pre-programmed settings to an actual reaction when the world misbehaves during deposition, provided you have the right communication interfaces for real-time loops.
Taro: I just think it’s important to remember that even with these controllers, we still need robust testing outside of a perfect lab setup. If the machine gets bumped or if there's unexpected heat variation, how gracefully does this whole system recover?
Rosa: That’s what I always ask. We need to see how long these systems can hold up in a messy environment and what happens when the world doesn't behave exactly as the authors modeled it.
Dev: I agree. The safety case depends heavily on those online reaction capabilities and the latency involved in that feedback loop, so we need to keep pushing on optimizing that loop rate for real-world operation.
Taro: It sounds like this paper provides a solid foundation for the hardware design, but now the focus has to shift toward testing how it performs when things get noisy and unpredictable.
Rosa: Exactly. We’ve got this concept of "Towards Track Height Control in Additive Manufacturing on Porous Substrates" which lays out a clear path from identification to control selection.
Dev: It’s a solid piece of work because it shows the practical steps for building these systems, and it sets up the groundwork for us to design those more sophisticated state-feedback controllers next.
Taro: Next week, we need to look at how these kinds of control strategies handle more complex disturbances, maybe when you're dealing with multiple interacting variables instead of just track height.
More episodes
- 2608.13438-ContactGuard: Pre-Contact Execution Monitoring with Action-Conditioned Latent World Models
- 2610.34593-Future Prediction as Training-Only Video Supervision for Pretrained VLA Policies
- 2610.35385-EmbodiedLab: Beyond Manipulation, Benchmarking Embodied Agents That Act to Know
- 2610.36378-Data-Driven Hamiltonian for Neural Optimal Control
- 2610.37275-A Heuristic for Scheduling Autonomous Vehicles at Intersections
- 2610.37288-Dynamic Coordination of Autonomous Vehicles at Intersections
- 2610.37632-Real-Time Coordination of Autonomous Vehicles at Intersections
- 2611.13779-Softness is something you also need! Why compliance helps control: information metrics of a passive walker with compliant ankle
- 2611.13411-Residual-Corrected Physics-Informed Neural Network KKL Observers with Input-to-State Stability Guarantees
- 2611.14488-Is Your Robot's Memory Just a Clock? Temporal Shortcuts in Memory-Augmented VLA Policies