Build Orientation Optimization Based on Weighted Analysis of Local Surface Region Curvature

Round 1

Reviewer 1 Report

In this paper, the authors proposed a method to optimize 3D printing build orientation for improving the surface quality of a part using surface feature weight. However, this paper put too much effort in describing the proposed method using the case study, while theoretical contributions are not well identified. The paper has several shortcomings that must be addressed before it is suitable for technical publication.

1. The abstract should be updated to reflect the substantive content of the paper.
2. In the introduction, the authors need to provide contextualized information about the study; the main objective and novelty of the manuscript are unclear. Also, the authors need to add a graphical explanation of “the stair effect” for better understanding.
3. There has been increasing interest in applying machine learning algorithms for optimization of AM process design including build orientation. The authors should systematically provide more literature review on this topic in Section 2. In particular, an overview of so far developed methods for single- and multi-objective optimization needs to be described.
4. The authors need to describe the details of the proposed optimization approach using a step-by-step approach.
5. The authors need to quantitatively compare the accuracy of different clustering algorithms for identifying geometry features.
6. It is difficult to understand the clear linkage between build orientation and surface roughness shown in Figure 10. The relevant mechanism should also be discussed.
7. The conclusion in this paper summarizes the research content, but does not reflect the deficiencies, limitations, the future prospects or the research direction. Please confirm and modify accordingly.

Author Response

Response to Reviewer 1 Comments

Comments Proposed：

In this paper, the authors proposed a method to optimize 3D printing build orientation for improving the surface quality of a part using surface feature weight. However, this paper put too much effort in describing the proposed method using the case study, while theoretical contributions are not well identified. The paper has several shortcomings that must be addressed before it is suitable for technical publication.

Point 1:The abstract should be updated to reflect the substantive content of the paper.

 

Response 1: The abstract is updated already

 

Point 2:In the introduction, the authors need to provide contextualized information about the study; the main objective and novelty of the manuscript are unclear. Also, the authors need to add a graphical explanation of “the stair effect” for better understanding.

 

Response 2: I add more information in the third paragraph in Introduction and add Figure 1 to explain the staircase effect. The supplementary content further clarifies the research purpose and limitations of this article.

 

Point 3:There has been increasing interest in applying machine learning algorithms for optimization of AM process design including build orientation. The authors should systematically provide more literature review on this topic in Section 2. In particular, an overview of so far developed methods for single- and multi-objective optimization needs to be described.

 

Response 3: I add one paragraph(line90-line100) in section Literature to provide machine learning algorithms for optimization of AM process design including build orientation.

 

Point 4:The authors need to describe the details of the proposed optimization approach using a step-by-step approach.

 

Response 4: I add a flowchart to describe the details of the proposed optimization approach shown in Figure 7.

 

Point 5:The authors need to quantitatively compare the accuracy of different clustering algorithms for identifying geometry features.

 

Response 5: I add two comparison graphics(shown in Figure 5 and Figure 6) to illustrate that the algorithm proposed is more effective in extracting regions with large changes in the surface curvature of the model.

 

Point 6:It is difficult to understand the clear linkage between build orientation and surface roughness shown in Figure 10. The relevant mechanism should also be discussed.

 

Response 6: I add more information(line 320-line325) to discuss the mechanism.

 

Point 7:The conclusion in this paper summarizes the research content, but does not reflect the deficiencies, limitations, the future prospects or the research direction. Please confirm and modify accordingly.

 

Response 7: I add one paragraph(line343-line348) about the deficiencies, limitations and the future prospects in the end of this paper.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper poposes a mono-objective evaluation for build orientation determination problem. A decomposition method, based on, curvature shift, is applied to generate facet clusters and a special weight assignment scheme is adopted. Some numerical cases are presented for method demonstration. It seems complete, but there are still a couple of problems that should be solved.

1. The application background of the mono-optimization method is not well introduced. Why only surface quality is enough? Which process and what kind of application would be in this case?
2. Some key references concerning feature or facet cluster based build orientation optimization are missing. There are already similar proposals in literature. Here are some representative papers: (a). Zhang, Y., Bernard, A., Gupta, R. K., & Harik, R. (2016). Feature based building orientation optimization for additive manufacturing. Rapid Prototyping Journal. (b). Zhang, Y., Harik, R., Fadel, G., & Bernard, A. (2019). A statistical method for build orientation determination in additive manufacturing. Rapid Prototyping Journal. (c). Shi, Y., Zhang, Y., Baek, S., De Backer, W., & Harik, R. (2018). Manufacturability analysis for additive manufacturing using a novel feature recognition technique. Computer-Aided Design and Applications, 15(6), 941-952, etc.
3. The 'feature' in the paper is not clear and different to that in traditional CNC or newly defined AM feature/benchmarking feature in AM domain. The method mentioned in the paper is more about surface decomposition and there is no defitniion/description on 'feature'. The 'feature' here is more about geometric property intself but has no manufacturing/printing meaning associated.
4. In the proposed method, curvature similarity is used as clustering similarity rule. However, the impact of the so called similar facets to printing quality is not well proved. The authors should explain clearly why these facets in the same cluster have similar manufacturing results;
5. Surface decomposition is quite complicated. Curvature similarity is used for decomposition, but for build orientation determination problem, concerning the computation cost and global optimamity, an optimal number of clusters should be found. How to find a reasonable number of clusters is not well explained in the paper. Clustering analysis is necessary.
6. For the build orientation determination problem, there are usually two main steps: generate candidate orientations and evaluate them. In the paper, it is not clear that how to generate alternatives after decomposition of the surface model;
7. Manufacturability issue is not considered in the paper. Pure geometric modeling sometimes is not enough to ensure the optimization result. Authors should be aware of this. For example, two cylinders having similar curvatures may have different printing results if their sizes are different (such as one is normal and the other is a long shaft).
8. The decision weight assignment scheme in this paper is based on the assumption of automatic orientation optimization without input of user preference. However, as said in the paper, different regions of a CAD model may have different importance or function in application. Hence, how to deal with given decision preference is another issue.
9. The case study is imcomplete since the full procedure of the optimization is not well illustrated and the comparison is not well presented. More information should be provided, e.g. how to generate the candidate orientations? What is the printer printing parameter setting in the comparison? ...

In general, the paper is not suitable for publication in the current version. The comments above could be used for authors' consideration in revision for resubmission.

Author Response

Response to Reviewer 2 Comments

Comments Proposed:

The paper poposes a mono-objective evaluation for build orientation determination problem. A decomposition method, based on, curvature shift, is applied to generate facet clusters and a special weight assignment scheme is adopted. Some numerical cases are presented for method demonstration. It seems complete, but there are still a couple of problems that should be solved.

Point 1:The application background of the mono-optimization method is not well introduced. Why only surface quality is enough? Which process and what kind of application would be in this case?

 

Response 1: I added supplementary information to the Introduction. Both the Abstract and the Introduction mentioned that this article is indeed a theoretical analysis and verification from a geometric perspective, and does not involve process parameters and manufacturing issues. Methods proposed belongs to 3d printing pre-processing. The subsequent research will expand to complicated issues such as process parameters and manufacturing problems

 

Point 2: Some key references concerning feature or facet cluster based build orientation optimization are missing. There are already similar proposals in literature. Here are some representative papers: (a). Zhang, Y., Bernard, A., Gupta, R. K., & Harik, R. (2016). Feature based building orientation optimization for additive manufacturing. Rapid Prototyping Journal. (b). Zhang, Y., Harik, R., Fadel, G., & Bernard, A. (2019). A statistical method for build orientation determination in additive manufacturing. Rapid Prototyping Journal. (c). Shi, Y., Zhang, Y., Baek, S., De Backer, W., & Harik, R. (2018). Manufacturability analysis for additive manufacturing using a novel feature recognition technique. Computer-Aided Design and Applications, 15(6), 941-952, etc. 

 

Response 2: I have added these references to theLiterature. These articles are indeed very good and have great reference value for me

 

Point 3: The 'feature' in the paper is not clear and different to that in traditional CNC or newly defined AM feature/benchmarking feature in AM domain. The method mentioned in the paper is more about surface decomposition and there is no defitniion/description on 'feature'. The 'feature' here is more about geometric property intself but has no manufacturing/printing meaning associated.

 

Response 3: After reading the articles on features mentioned above, I admit that I used the wrong word“feature” in the article. The feature area I want to express refers to the region where the curvature of the model surface has a relatively large change such thin edges and corners rather than specific geometric features such as cylindrical holes. In order to avoid causing ambiguity, I removed the word “feature” from the text and used “key regions” and “key points” instead

 

Point 4: In the proposed method, curvature similarity is used as clustering similarity rule. However, the impact of the so called similar facets to printing quality is not well proved. The authors should explain clearly why these facets in the same cluster have similar manufacturing results;

 

Response 4: I add two comparison graphics(shown in Figure 5 and Figure 6) to illustrate that the algorithm proposed is more effective in extracting regions with large changes in the surface curvature of the model

 

Point 5: Surface decomposition is quite complicated. Curvature similarity is used for decomposition, but for build orientation determination problem, concerning the computation cost and global optimamity, an optimal number of clusters should be found. How to find a reasonable number of clusters is not well explained in the paper. Clustering analysis is necessary.

 

Response 5: In this paper, cluster analysis uses the maximum curvature of the gradient to obtain a cluster center, and then averages the curvatures of all clusters to obtain a sorted clusters in order to generate a classification of curvature differentiation clusters. The number of clusteres is determined by the bandwidth and threshold.

 

Point 6: For the build orientation determination problem, there are usually two main steps: generate candidate orientations and evaluate them. In the paper, it is not clear that how to generate alternatives after decomposition of the surface model;

 

Response 6: Curvature shift clustering in this paper is not meant to decompose the entire model, but to more reasonably assign weight factor to each triangle facet to construct the error function. For the calculation of candidate directions, please refer to the literature [13]: Luo, N.; Wang, Q. Fast slicing orientation determining and optimizing algorithm for least volumetric error in rapid prototyping. The International Journal of Advanced Manufacturing Technology 2016, 83, 1297–1313.

 

Point 7: Manufacturability issue is not considered in the paper. Pure geometric modeling sometimes is not enough to ensure the optimization result. Authors should be aware of this. For example, two cylinders having similar curvatures may have different printing results if their sizes are different (such as one is normal and the other is a long shaft).

 

Response 7: The article does not involve the related issues of process manufacturing just as respons 1 says. It is purely to analyze the principle of staircase error from a geometric perspective. From this principal, to optimize the optimal printing direction and reduce the impact of low surface accuracy caused by stair case error. The research scope is too wide if considering manufacturing. The method is universal which belongs to pre-processing and it does not contradict the subsequent process and manufacturing. If involving a specific part and its applications, of course, the production and process issues need to be considered, and the choice of build orientation must also be adjusted appropriately. Thank you for your reasonableadvice. It is suggested that this issue should also be one of the topics to be further studied and refined in the future.

 

Point 8: The decision weight assignment scheme in this paper is based on the assumption of automatic orientation optimization without input of user preference. However, as said in the paper, different regions of a CAD model may have different importance or function in application. Hence, how to deal with given decision preference is another issue.

 

Response 8: The weight distribution strategy is indeed automatic distribution. The strategy adopted in this article is to assign a large weight to the triangle facet with a large surface region curvature. If there is no specific requirement for the process of the part, this strategy is relatively reasonable. If there are special requirements such as positioning, the strategy needs to be re-adjusted to meet the requirements.

 

Point 9: The case study is imcomplete since the full procedure of the optimization is not well illustrated and the comparison is not well presented. More information should be provided, e.g. how to generate the candidate orientations? What is the printer printing parameter setting in the comparison? ...

 

Response 9: Same as Comment 6, the specific reference for determining the candidate printing direction can be found in [13]: Luo, N.; Wang, Q. Fast slicing orientation determining and optimizing algorithm for least volumetric error in rapid prototyping. The International Journal of Advanced Manufacturing Technology 2016, 83, 1297–1313. At the beginning of this article, the research goal is limited to optimizing the surface quality caused by staircase errors. The optimization strategy is more focused on the regions where the surface curvature changes obviously by assigning different weights. Therefore, in the final verification, the region with large curvature change will be compared, especially the sharper region of part. If the process requires the accuracy of the mating surface, the weight can be distributed more to the area with relatively small curvature change.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

After reviewing the revised version of the manuscript, the paper is now clear than before. It is recommended that the authors consider the following suggestions.

1. The authors need to add more descriptions of the flowchart in Figure 7.
2. In literature review, the authors need to provide more descriptions of the work of others than a critical appraisal of prior work. It would be good to see more of the author views and opinions reflected in the review.

Author Response

Comments Proposed：

After reviewing the revised version of the manuscript, the paper is now clear than before. It is recommended that the authors consider the following suggestions.

 

Response：We have made detailed improvements in accordance with the Comments given, and have marked in different colors in the comparison PDF（diff.pdf）. For grammatical problems in the text, we agree to use third-party editing services to improve and polish.

The following are the detailed revisions:

Point 1: The authors need to add more descriptions of the flowchart in Figure 7

 

Response 1: We add more descriptions in the paper: line 264 – line 283:

The entire optimization process is shown in Figure~\ref{fig6}. The optimization process first loads the 3D CAD model format. The model adopts the STL format. The STL file format can easily obtain the points information and triangle facets information of the model. The curvature information of each point is calculated by fitting the NURB parameter surface to the local point cloud, as the initial reference value for the curvature shift calculation. Substitute the curvature weight coefficient $ w(c) $ and the distance weight coefficient $ w(d) $ of each point in the local point set into the formula to calculate the drift vector $ C_{sv} $, and use the set threshold as the condition for $ C_{sv} $ convergence. If the convergence condition is met, it is the same cluster, otherwise set a initial point for the new cluster to calculate. When the clustering of the entire model is completed, the curvature of the cluster center point is used as the cluster weight $ w_c $, and the curvature of each point in the cluster is used as the local weight $ w_l $ of each point. The weight of each point is determined by $ w_c $ and $ w_l $.  The weight $ w_f $ of each triangle facet is determined by the arithmetic average of its three vertices. The points of the model whose curvatures are greater than the average curvature $ c_m $ are selected as the key points for calculating the optimal build orientation. The triangle facet where each vertex is located is used as the key facet. Finally, $ w_f $ is introduced into the volume error function  $ V $ and three vectors can be obtained as candidate build orientations after eigenvalue decomposition of $ V $. The vector that minimizes  $ V $ is the optimal build orientation. This orientation ensures the surface quality of the regions where key points are located. Of course, different process and manufacturing requirements have different key points. In this article, only the points with greater curvature change is selected for method demonstration. As for special process and processing requirements, the mechanism of screening key points is more complicated, and the optimal build orientation is adaptively changed.

 

Point 2: In literature review, the authors need to provide more descriptions of the work of others than a critical appraisal of prior work. It would be good to see more of the author views and opinions reflected in the review.

 

Response 2: In literature review, we add more critical opinions and our own opinions(blue text):

 

[12] determines the optimal build orientation based on the position of normal vectors of the original CAD model. All the normal vectors are considered asthe candidate orientation. It is certainly feasible to adopt an exhaustive approach, but subject to the existing the surface normals. It is difficult to get the optimal build orientation. [13] used a new volume error model. The problem with the least absolute deviation. But the biggest problem is that all normal-weighted regions are weighted equally.The geometry features of the model itself are neglected and the weighted optimization process essentially belongs to normal weight, which means all facets contribute equally to the build orientation. [14] proposed a new method to determine the optimal build orientation using Gauss Map, which collects all facet normal in a unit sphere. The unit vector from the center of the sphere to the center of the bottom circle of the spherical crown is the optimal build direction. Although just considering the facet normal, it effectively decreases the volume of support structure. But it should be noted that the algorithm is not versatile especially for some more complex parts which contain open concave loops or non-sharp edges. [15] researched the effect of different build orientations on Ti-6Al-4V for microstructural and mechanical property evaluation. However, printing performance is more affected by material properties and cooling deformation after thermoforming. Relatively speaking, the printing direction has relatively little influence. [16] used particle swarm optimization [17] to search an optimal build orientation owing to its ease of programming, convergence speed and computational efficiency. If the initial particle is not good, it may be biased by a certain particle. The algorithm is easy to converge prematurely and fall into a local optimal solution. In addition, the complexity of the algorithm increases, resulting in reduced efficiency.

 

……[21]proposed a multi-object optimization of surface roughness, build time and support structures. Utility function approach is adopted to convert the optimization problem into a linear-weighted function. The optimal build orientation is determined by iteratively inputting the parameter. Although the multi-objective optimization algorithm uses multiple research objects and assigns a certain weight to each object to select the optimal printing direction to ensure that the final printing effect meets the expected settings, there are common shortcomings. The requirements of each research object on the printing direction are different or even opposite, which leads to a contradictory calculation process in the direction optimization process. If the weight of a research object is increased or decreased, in fact, the multi-objective optimization algorithm will reduce the dimensionality to a single-objective optimization algorithm. Back to the problem of studying single objective optimization.

       ……[23] proposed generalized machine learning based parameter optimization framework to determine optimal build orientation for FDM components. The mechanical properties of the test examples selected are tested and simulated from the three main directions of XYZ, and then the training data is obtained through orthogonal experiments. Obviously the algorithm does not have universal applicability and is not applicable to other complex models. The simulation data of complex models is definitely more than three main directions for testing. [24] proposed method applies a non-supervised machine learning method, K-Means Clustering with DaviesBouldin Criterion cluster measuring, to rapidly decompose a surface model into facet clusters and efficiently generate a set of meaningful alternative build orientations. Cluster analysis based on statistics takes a lot of time and it is difficult to converge for more complex models. CNN are most precise at estimating all three studied factors than the baseline linear regression model for the training and evaluation conditions [25]. Machine learning algorithms have a great advantage for studying a class of models with similar characteristics. Given enough training data, it fits very well. But when the model has anisotropic characteristics, either increase the training sample data or the new model needs to be re-learned to train the data. In addition, machine learning requires very high computer performance and it is more difficult to choose meta-parameters and network topology

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper has been improved. But there is still a need to check the writing and avoid to use 'I' in the test since you have more than one authors. 

Author Response

Comments Proposed:

The paper has been improved. But there is still a need to check the writing and avoid to use 'I' in the test since you have more than one authors.

Response：We have made detailed improvements in accordance with the Comments given, and have marked in different colors in the comparison PDF（diff.pdf）. For grammatical problems in the text, we agree to use third-party editing services to improve and polish.