Additive Manufacturing (AM) is a fast developing technology which started in 1987 with a stereolithography (SL, SLA) from 3D Systems, Inc. First non-SL systems, such as Fused Deposition Modeling (FDM) from Stratasys, Laminated Object Manufacturing (LOM) from Helisys, were presented in 1991. The first machine based on curing acrylate photopolymer that used the Digital Light Processing (DLP) technology from Texas Instruments was presented at EuroMold 2001 by Envisiontec . Different companies use different AM techniques for their products. NASA, for example, use AM to customize spacecraft and instrument components. It makes time, expenses and mass to be saved. But for such companies, other problems appear. These problems are related to a spaceflight certification and the validation of printed components by a Non-Destructive Evaluation (NDE) .
In the field of biomedical engineering, SL printers have become widespread. The main advantage of them is the possibility to use biocompatible UV curable photopolymers with special additives. And at the same time, printed models have a high precision. It makes production of scaffolds for live cells or drug delivery systems possible [3, 4]. On the other hand, there are a lot of unsolved problems in AM technology. One of the main of them is a long printing cycle. It depends on model dimensions and type of the printer. In the case of the DLP printers some solutions already exist, which can decrease build time in hundreds times . But it is still required to use special fast curing aggressive materials. It can happen that after a long time printing cycle, a poor quality product or even a totally unusable product is arrived at. It decreases the value of the main advantage of the 3D printing technology which is high speed prototyping. Embedded systems, which can control a lot of parameters during the printing process, can help to receive a product, which has as few differences with CAD model, as possible . That’s why implementation of process monitoring system during the printing can increase the production speed.
Most of the NDE methods are suitable only for the printers which use metal powders as a building material. For example, fluorescent penetrant inspection (FPI), eddy current testing (ET), magnetic particle testing (MT) are experimentally used in selective laser melting (SLM) and direct metal laser sintering (DLMS) technologies [2, 7–9].
The problem of implementating NDE methods in DLP and other stereolithography printers is based on a consistency of the photopolymer solutions, which these printers use as a building material. It makes using of common NDE methods impossible.
The problem, which is typical for DLP printers, is shrinkage during polymerization. It can cause cracks and deformation in printed layers  and, as consequence, low quality products or interrupted building process. Residual stress, which appeared in printed part, can cause even dropout from a building platform.
A method, which is described below, is focused on the bottom-up DLP printers, where the liquid resin bath has a UV transparent bottom side. It is shown on Figure 1. A DLP unit generates an image that is projected on the bottom side of the bath. Under influence of the UV radiation, a thin layer of photopolymer cures. Then platform goes up, and light source generates an image of the next layer. Process continues until the last layer is printed.
One of the advantages of these machines is that the printed object is pulled out from the resin. It makes these printers more cost-effective, than the top-down analogues , because a liquid resin bath should not be full during the printing process. Proposed solution is suitable for all known methods of separation of solid photopolymer from the bottom of the tank, such as displacement tank with a photopolymer [12, 13], tilting tank, passive tilting  and passive self-peeling technology . It makes measurement and analysis of adhesion force during the printing process possible. This force consists of the van der Waals force, chemical bonding force and suction force . The adhesion force appears between the printed surface and bottom part of the tank with a photopolymer when Z axe starts to move up.
Some types of DLP printers have a silicone-based bottom surface cover. It causes poor adhesion properties and it can reduce adhesion force during the unsticking of printed layer from the bottom surface. But on the other hand, the disadvantage of this cover is poor mechanical properties that can cause some damages during an inappropriate printing process. For example, when the printed model falls down from building platform, and printer continues printing, the last layer will be radiated for a long time and can stick to the silicon surface.
2 Experimental setup
We placed a strain gauge force sensor (load cell, LC) between a holder and a platform (Figure 2) and connected it through Phidget Bridge to a computer. Program for collecting the data from a sensor was made in LabView with using Phidget‘s libraries. Data were gathered from a direction output of a motor driver of Z axis and LC at the same time. It helped to reduce amount of data and to create proper time axis for resulting graph. For the next processing we used data which were collected when the platform moved up. Then for each movement of the platform, we found a maximal value of the force. This data are presented as results of next experiments. Frequency of measurement was set up to 1000 samples/sec.
The first experiment was designed to simulate a situation when the printed model, which is shown on Figure 3, falls from the platform and lies on the printed surface. We had set printer properties in such a way that the first layers of the model did not solidify properly. That caused unsticking of the building model after some time. In this case we tried to detect a moment when the building process faulted.
The next experiment was designed for finding more detailed relations between the area of printed surface and the force which appears during the unsticking process .
First of all, we tried to find relations between the shape of the printed object and the adhesion force. We printed two items with same parameters except the layer form. 3D models of printed objects are shown on Figure 4. Both consist of 4 steps. The area of every next step is two times smaller than the previous one and is the same for both models. The thickness of each step is 1.5 mm. Thickness of each layer was set up to 50 μm. Therefore, every step consists of 30 layers. Dimensions of printed objects are shown in the Table 1.
Evaluation of the maximal force was provided by taking the maximal detected force and by calculation of all area under the measured graph. There was no big difference between the results of these two approaches. For simplification of the calculation, we choose the first method and took just a maximal value of the force.
3.1 Experiment for detection of the moment of model fall
We performed the first experiment four times. Data from these measurements are shown on Figure 5. Conditions for all tests were the same. We measured the adhesion force during the movement of the platform in the up direction. Measurements were provided 1000 times per second. Then we found the maximal force that appeared during unsticking process for each layer and got the results.
In all experiments, printing job was not finished. We can see that all graphs have similar shape. They consist of three stages that are marked on graphs on Figure 5 as areas I, II and III. The first stage of printing process describes a state, in which the printing platform is immersed in photopolymer solution. In this case the platform is stuck to the bottom surface of the tank over the entire area. Consequently, the force that appears during the unsticking process in this phase has the highest values. Length of the first stage is approximately the same for all 4 experiments. It is gained by mechanical properties of the printer (platform thickness, shape of the photopolymer tank) and by the level of the photopolymer solution in the tank. During the first stage, a gap between the platform and the surface of the tank increases and the force gradually starts being proportional to the area of the printed layer.
The second stage of the printing process (marked as II on all four graphs in Figure 5) shows normal printing process, in which the platform sticks to the tank surface only through the printed model. The force depends mostly on the area of the printed layer. Layer by layer weight of the printed model rises and internal forces start to bend it. This bending causes an unsticking of the printed object from the platform. Hence, at the end of the second stage a fault of printing process appears, the model falls into the tank.
The third stage shows the situation, in which the printed model falls from the building platform and the printing process should be stopped to prevent damages. In each experiment, the model fell in during printing of different layers. It was difficult to set parameters in a such way that the fault during printing appeared at the same layer. In each experiment, we can see significant decrease of the adhesion force. Therefore, the fault appearance can be detected using the measurement of this force. During the described experiments, fault detection algorithms were not used, printing job was stopped manually.
Non-automatic fault detection is not a trivial task in this type of 3D printers because the printed part is immersed in a photopolymer solution during relatively long time. This time depends on layer thickness and amount of building material in the tank. That’s why we can observe that in the first test (Figure 5a) printer continued working approximately 20 layers after fault had appeared. During this extra time in the fault state, the model had been adhered to the bottom of the tank and further printing was not possible without full cleaning of the printer. This problem can be eliminated using the automatic fault detection algorithms.
3.2 Experiment for examining of dependency between the model shape and the unsticking force
Results of the second experiment represent the values of the adhesion force during printing of four-step models (Figure 4). Data for the pyramid are shown on the upper part of Figure 6, data for the cylinder are presented on the lower part of Figure 6. Each step of both models consists of 30 layers that have the same area. In each layer, we measured the force during unsticking process with a frequency 1000 samples per second. Then we choose maximums for every layer, results are shown as dots on graphs on Figure 6.
Graphs on Figure 6 represents, that the value of the force during unsticking of the last layer from the bottom of the tank depends on the area of printed layer and doesn‘t depend on its shape (both measurements provide us with almost the same results). Also, we can conclude, that the value of the force is positively correlated with area of the layer. Even though the variance of the force is relatively high, it is still possible to apply appropriate algorithm of fault detection.
On presented graphs (Figure 6) we can also see, that observations are noisy. Uncertainties of measurement are caused by the error of a sensor, the nonuniformity of the photopolymer solution and other factors. But also, it is clear, that there is a dependence between the printed area and the force. That is why we can use force measurement for in-situ process monitoring. An implementation of the continuous adhesion force monitoring can allow us to detect such faults during a printing process as non-properly solidification of some areas of printed layer.
In each layer adhesion forces, which appear due to the solidification of the photopolymer resin in the bottom part of the bath, are presented. Simple evaluation shows that if this force is not presented, it means that no layer was formed. That can exclude the possibility of continuing the printing when the printed part has already fell. Job should be stopped for decreasing the amount of used material and to save the prototyping time. This solution can also help to manage the clog problem of photopolymer resin. It leads to material and time savings that can be achieved by reducing of quantity of required services.
For in-situ process monitoring an algorithm, which is shown on a Figure 7, was designed.
At the start of the printing cycle we have a 3D model of the printing object. Then a slicer cut up an object into layers accordingly to defined parameters. The next step is calculation of the area of each layer. Then calculation of minimal and maximal (A and B) theoretical adhesion forces with some permissible error (ξ) follows. At the same time, printing of the first layer starts and the first measurement during moving platform up is provided. The measured force is compared with calculated forces, and system tests if measured force is in allowed range. If it is not the case, then the printing process is stopped, otherwise system continues the printing (if the printed layer was not the last one in the model). For increasing the effectiveness of the algorithm, it is possible to add an additional option for reducing probable failure decisions. After detection of the fault, the control system will wait few more measuring cycles before stopping the printing process for verifying the results (the force must be out of range in few sequential cycles).
For increasing a range of used photopolymer solutions, the next idea was proposed. At the beginning of a printing cycle, program will move platform to the start position and a maximum value of force, which appears during the compression of the photopolymer, will be measured. This force depends on the viscosity of used solution and affects the adhesion force.
Implementation of this algorithm for in-situ process monitoring can reduce a printing time and allow to have an information about printing process for DLP 3D printers.
The research reported in this paper was supported by targeted support for specific university research within the student grant competition TUL (Project 21071 – Development and prototype production of compact DLP 3D printer).
Wohlers T., Gornet T., History of additive manufacturing. In: Wohlers T. (Ed.), Wohlers Report 2014: Additive Manufacturing and 3D Printing State of the Industry, Wohlers Associates Inc., USA, 2014, 1-34 Google Scholar
Waller J.M., Parker B.H., Hodges K.L., Burke E.R., Walker J.L., Generazio E.R., Nondestructive Evaluation of Additive Manufacturing, Langley Research Center, Hampton, Virginia, 2014 Google Scholar
Wang G., Goyanes A., Gaisford S., Abdul W.B., Stereolithographic (SLA) 3D Printing of Oral Modified-Release Dosage Forms, International Journal of Pharmaceutics, 2016, 503, 207-212 CrossrefWeb of ScienceGoogle Scholar
Tumbleston J.R., Shirvanyants D., Ermoshkin N., Janusziewicz R., Johnson A.R., Kelly D., et al., Continuous liquid interface production of 3D objects, Science, 2015, 347, 1349-1352 Web of ScienceCrossrefGoogle Scholar
Guessasma S., Zhang W., Zhu J., Belhabib S., Nouri H., Challenges of additive manufacturing technologies from an optimisation perspective, Simul. Multisci. Des. Optim., 2015, 6, A9 CrossrefGoogle Scholar
Volker C., Monitoring system for the quality assessment in Additive Manufacturing, In: Chimenti D.E., Bond L.J. (Ed.), 41st Annual Review of Progress in Quantitative Nondestructive Evaluation (20-25 July 2014, Boise, Idaho, USA), American Institute of Physics, Melville, NY, 2015, 1650, 171-176Google Scholar
Rieder H., Dillhöfer A., Spies M., Bamberg J., Hess T., Online monitoring of additive manufacturing processes using ultrasound, In: Prevorovsky Z. (Ed.), 11th European Conference on Nondestructive Testing (6-10 October 2014, Brno, Czech Republic), Brno University of Technology, Brno, 2014 Google Scholar
Shkolnik A., El-Siblani A., Hendrik J., Envisiontec Gmbh, United States Patent 7892474 B2, 22 February 2011 Google Scholar
Hendrik J., Envisiontec Gmbh, United States Patent 7195472 B2, 27 March 2007 Google Scholar
Joyce M., B9Creations LLC, United States Patent 20130292862 A1, 7 November 2013 Google Scholar
Holtrup R., Design and construction of a multi-material 3D DLP printer, Final Bachelor Assignment, University of Twente, Netherlands, 2015 Google Scholar
Syao K.C., United States Patent 20150064298 A1, 5 March 2015 Google Scholar
Kovalenko I., Garan M., Shynkarenko A., Zelený P., Šafka J., Examining the Relationship between Forces During Stereolithography 3D Printing and Geometric Parameters of the Model, In: Chang G.A., Ma M., Perumal S.A., Chen G. (Ed.), International Conference on Mechanical Engineering and Electrical Systems (16-18 December 2015, Singapore), EDP Sciences, France, 2015 Google Scholar
About the article
Published Online: 2017-04-26
Citation Information: Open Engineering, Volume 7, Issue 1, Pages 100–105, ISSN (Online) 2391-5439, DOI: https://doi.org/10.1515/eng-2017-0016.
© 2017 I. Kovalenko et al.. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. BY-NC-ND 3.0