Fuzzy inference system for failure strength estimation of plain and notched 3D‐printed polylactide components

Abstract

A fuzzy sets based computational fuzzy inference system has been used to estimate the failure strength of 3D‐printed polylactide components. The research has confirmed and validated the accuracy and reliability of this approach with a satisfying level of reliability. As far as failure strength is concerned, the following two types of input parameters have been considered: (i) manufacturing variables (i.e., manufacturing angle, infill density, and size of manufacturing voids) and (ii) geometrical features (i.e., notch root radius). The individual significance of the various parameters under investigation has been identified together with the influence on the estimation accuracy of the number of specimens being used. The fuzzy inference system has shown an accuracy improvement compared to the failure strength estimation, obtained as a result of an existing analytical method. The fuzzy inference system approach has also been shown to have a good potential as a decision‐making tool in design problems.

The application of the Theory of Critical Distances to nonhomogeneous materials

“…In light of the unique features of 3D printed materials, over the last decade, the Sheffield Structural Integrity Research Group has run a number of experimental/ theoretical projects to assess whether the TCD is successful in assessing the static [38][39][40][41][42][43] and fatigue [44][45][46] strength of notched/flawed 3D printed materials. In what follows, some specific outcomes from this body of systematic research work will be reviewed and revisited by focusing attention specifically on polymers and concrete.…”

Fuzzy set‐based methodology for manufacturing parameter determination of 3D‐printed polylactide components with user‐specified strength, geometrical design, and cost requirements

“…Furthermore, when it comes to objects printed flat on the build plate, the mechanical behavior depends on the manufacturing raster angle 19,23,24 . This is because, by changing the manufacturing raster angle, the strength performance of the printed objects vary based on the average layer adhesion and the ultimate tensile strength of the 3D‐printed material 25,26 . Specifically, Weake et al 27 found that the tensile strength of an acrylonitrile butadiene styrene (ABS) part could be up to 150% higher when the applied force is parallel to the material filaments (manufacturing angle equals to 0°) than perpendicular (manufacturing angle equals to 90°).…”

“…Typical data‐driven methods, including artificial neural networks (ANN), fuzzy inference system (FIS), and other techniques, are widely applied to study the mechanical behavior of 3D‐printed object 27,29 . The complexity caused by the joint effect of multiple parameters is seen to be effectively solved by the application of nonlinear regression solutions included in the aforementioned techniques 26 . Specifically, compared with ANN—a “black box” structure, FIS framework has a “grey box” structure which benefits from its fuzzy rules, allowing more user control through the relatively transparent structure.…”

“…Previous studies have proven that FIS has the capability to estimate the tensile strength with given manufacturing and geometrical parameters 26 . However, the current industrial problem is that with both strength and geometrical design requirements set for a 3D‐printed object, the optimal combination of manufacturing parameters can be impossible to be determined a priori.…”

“…Besides, adjustments are conducted to identify the most appropriate combination of both manufacturing parameters in order to meet specific industrial needs, for example, setting material‐saving as a top priority. Due to the fact that the estimation in the present work seeks for manufacturing parameters which lead to required strength and geometry, it will be referred to as inverse estimation in the following sections, contrary to direct estimation , resulting in strength based on the given manufacturing parameters reported in our previous work 26 …”