Most process monitoring in additive manufacturing watches a build as it happens. A team at Rowan University has gone the other way. In a method announced by the university on October 13, engineers Islam Mantawy and Behrad Koohbor apply digital image correlation (DIC) "in reverse time." They start from the completed structure and work back through thousands of images to reconstruct how 3D printed concrete deformed, shrank and cured while it was being printed. The work appears in the journal Experimental Mechanics.
The idea is simple to state and hard to do. Image correlation tracks how a surface moves from one frame to the next. Rowan's version picks the finished wall as its starting frame and steps backward through the image sequence, recovering the deformation history of a structure whose material was still changing the whole time it was being built.
The Problem: A Material That Won't Hold Still
Printed concrete is an unusual build material because it keeps changing after it leaves the nozzle. Coverage of the work by VoxelMatters notes that temperature, humidity and the mix's chemical composition all affect how a printed structure cures. Newly deposited layers are still deforming and curing while they bear the weight of additional material, and the lower layers respond to the weight and pressure of what is placed on top.
Mantawy gives an example of what that means in practice: in an 8-foot-tall wall, the bottom layer can start to crack and collapse, and the entire structure can fail before the final layer is reached. A failure like that starts at the base, under the load of the layers above, during a process where nobody can easily see inside the structure. That is what the method is designed to address. The aim is to recover how the structure actually behaved as it rose, so that deformation, shrinkage and curing behavior show up in measurements and not only in a post-mortem.
Mantawy, who heads Rowan's Additive and Robotic Construction Laboratory, put the stakes plainly in the VoxelMatters piece: "Without knowing what's happening inside the structure, you won't be able to achieve self-operating 3D printing."
How the Reverse-Time Approach Works
The method has two parts, as the university describes it.
The tracking pattern is the concrete itself. DIC needs surface texture that it can follow from image to image. Rowan's method uses the natural surface roughness of the printed concrete as that pattern. For a construction-scale print this matters a great deal: a technique that relies on the material's own texture does not depend on adding a pattern to the structure.
The analysis runs backward. Conventional DIC compares camera images against a reference image of an existing object. Concrete printing offers no such reference, because the structure is still being created as the process unfolds. So the correlation starts at the completed structure and moves back through the image sequence, which the release says can be thousands of images long. Working back from the end state, the team reconstructs the deformation, shrinkage and curing behavior that came before it, even during the first moments after a layer is printed.
The university says the team constructed a mathematical approach to validate the method. That matters for a technique like this one. A deformation history rebuilt from images is only useful if there is a way to check that the rebuilt numbers reflect what the material actually did, and not artifacts of running the correlation in an unusual direction. The release does not go into the details of that validation, and readers who want the formulation should go to the Experimental Mechanics paper.
Why This Is a Software Story
FilamentFeed files this under software on purpose. Nothing described here is a new printer, a new nozzle or a new concrete mix. The contribution is an analysis method: a different way to process image data that cameras can already collect. The new part is the order in which the images are correlated and the choice of texture to track.
That has practical implications. A method like this could, in principle, be layered onto existing printing workflows without redesigning the machine, though the sources do not say so directly. The hard work sits in the processing pipeline and in validating what it outputs, and the Rowan team says it has addressed the second of those with a mathematical approach.
The release also points past concrete. The researchers say DIC "in reverse time" could be useful in other additive manufacturing processes where conventional quality control methods are limited because a structure does not yet exist when production begins. That is a hedged claim, and it should be read that way. Still, the underlying problem is familiar to anyone who prints: you can inspect a finished part, but reconstructing how it got that way, layer by layer and minute by minute, is much harder.
What It Means for Makers
Most FilamentFeed readers are not printing 8-foot concrete walls. The research still connects to problems that desktop and prosumer printers deal with every day.
- Failures often start low and out of sight. Mantawy's example of a bottom layer failing under an 8-foot wall is a large-scale case of a familiar pattern: the trouble begins in the early layers and only becomes visible later. Any method that recovers what happened at the base during the build tackles the right problem.
- Your part's texture is data. Using natural surface roughness as a tracking pattern turns something makers usually try to get rid of into a measurement signal. Whether that carries over to polymer extrusion is an open question, and the release does not claim it does. It is still a useful way to think about the material.
- Autonomy needs observability. Mantawy's point about self-operating 3D printing applies at every scale. A printer cannot correct what it cannot measure, and closed-loop control depends on knowing the material's state, not just the toolpath.
- Analysis can beat new hardware. The method's novelty is in how it processes images. For makers who already put cameras on their printers, the message is that how you analyze the footage can matter as much as having a camera at all.
Bottom Line
Rowan's reverse-time DIC is an early research result, not a product. The university itself calls it an early step toward smarter, more autonomous forms of additive manufacturing, and the public materials leave a lot to the paper: frame rates, structure sizes, and how the mathematical validation is set up. The core idea is clear and well matched to its target. Construction printing uses a material that keeps changing after it is placed, is built at a scale where you cannot see inside, and has limited conventional quality-control options because the structure does not yet exist when production begins. Starting from the finished structure, running the image record backward, and letting the concrete's own roughness serve as the tracking pattern is a neat way to get information out of a process that has been hard to observe. If the hedged suggestion that it could reach other AM processes holds up, it will be worth following beyond concrete.
