Intelligent optimization system for powder bed fusion of processable thermoplastics

Highlights

• An intelligent optimization system is proposed to establish quantitative relationships between process parameters and multiple optimization objectives, including mechanical properties, productivity, energy efficiency, etc.

• Contour maps of operation window, productivity and energy efficiency can be developed to predict optimal parameters by considering the constraints of mechanical properties and material degradation.

• Using a facile data-driven approach, the relationships between process parameters and optimization objectives can be utilized in the process optimization and material selection.

Abstract

Powder bed fusion (PBF) represents a class of additive manufacturing processes with the unique advantage of being able to fabricate functional products with complex three-dimensional geometries. PBF has been broadly applied in highly value-added industries, including the biomedical device and aerospace industries. However, it is challenging to construct a comprehensive knowledgebase to guide material selection and process optimization decisions to satisfy the product standards of various industries based on a poor understanding of process-structure-property/performance relationships for each type of thermoplastic. In this paper, an intelligent optimization system is proposed to establish quantitative relationships between process parameters and multiple optimization objectives, including mechanical properties, productivity, energy efficiency, and degree of material degradation. Polyurethane is considered as a representative thermoplastic because it is sensitive to thermal-induced degradation and has a relatively narrow process window. Material and powder properties as functions of temperature are investigated using systematic material screening. Numerical models are created to analyze the interactions between laser beams and polymeric powders by considering the effects of chamber thermal conditions, laser parameters, temperature-dependent properties, and phase transitions of polymers, as well as laser beam characteristics. The theoretically predicted features of melting pools are validated experimentally and then utilized to develop quantitative relationships between process parameters and multiple optimization objectives. The established relationships can guide process parameter optimization and material selection decisions for polymer PBF.

Introduction

In additive manufacturing (AM), powder bed fusion (PBF) processes, such as selective laser-sintering (SLS), can be employed to fabricate three-dimensional (3D) objects with complex geometric features based on the digitally programmed layer-by-layer fusion of particles [1,2]. Multi-scale and multi-physics numerical models have been developed to describe fusion kinetics and defect generation during PBF processes [3,4]. Furthermore, it is necessary and crucial to establish quantitative relationships between process parameters and the properties/performance of printed components, including dimensional accuracy, surface toughness, anisotropic properties, and porosity [4,5]. Thus far, AM-driven intelligent optimization has focused on utilizing data-driven approaches to establish such quantitative relationships. The intelligent optimization of PBF processes typically relies on predictive models for material forming behaviors and the properties/performance of printed components [[6], [7], [8]].

Conventionally, numerical models for AM processes have been applied to the evaluation of material wastage, energy consumption, volumetric and geometric errors, scanning strategies, and building orientations [9]. For example, Verma et al. [10] developed a novel adaptive slicing approach based on heuristic optimization that could systematically predict material wastage and process energy efficiency. Shen et al. [7] derived a process efficiency map for SLS based on numerical modeling and experimental evaluation results. The process efficiency map was used to guide parameter selection for polyamide-based composites. Wang et al. [11] applied a neural fuzzy system to study the influence of process parameters on part shrinkage. Recently, novel approaches using data-driven modeling and the internet of things have been employed to analyze the energy consumption of AM technology based on different combinations of process factors [12,13]. Gusarov et al. [14,15] proposed a volumetric heat source model to simulate the interactions between laser radiation and powder beds. They predicted the local profiles of temperature fields associated with laser beams. A simple model for energy density or heat flux in PBF can be expressed as a function of laser power p, scanning speed v, hatching space h, and layer thickness L. Such a function can estimate the effective range of a laser energy input corresponding to the stable sintering range (SSR) of a specific thermoplastic [16]. However, thus far, few numerical models have been able to correlate process factors with the performance/properties of printed parts quantitatively with consideration for underlying material solidification and microstructure formation mechanisms. The interdependency of process parameters has yet to be fully defined in terms of their net effect on the micro/macrostructures and properties of printed parts. Therefore, no fully established system exists for parametric process simulation and experimental validation to analyze the quantitative relationships of the micro/mesostructures and properties/performance of macroscale components [17].

An intelligent optimization system should be able to consider multiple objectives simultaneously, including sustainability, energy efficiency, productivity, process repeatability, dimensional accuracy, and the properties/performance of printed products. The processability of materials, efficiency of processing, and performance of printed products depend heavily on the processing of different feedstock materials. Currently, an expanding material catalog of thermoplastics is available as PBF feedstock material, including polyamides (PA), polyether ether ketone (PEEK) [18], polypropylene [19], and polyurethane (PU) [20]. These polymers have proven suitable for SLS processes and can potentially be used to construct functional components for the aerospace, automotive, and medical industries. New categories of thermoplastics, such as PEEK, PU, and poly (phenylene sulfide), require very specific process conditions. The interdependencies of their process parameters have not been fully revealed using either experiments or simulation models. Furthermore, a lack of understanding regarding process-structure-property relationships restricts the improvement of process repeatability and sustainability of printed components. For example, the high processing temperature and low recyclability of PEEK are critical obstacles to process control [21]. Thermoplastic PU (TPU) is susceptive to laser-induced rapid heating and cooling treatment based on the relatively low decomposition temperature of its short-chain diols or other additives [20,22]. Therefore, the narrow SSR of TPU makes it difficult to identify a suitable process window and desirable energy density range [23,24]. In other words, it is difficult to predict its mechanical properties quantitatively while optimizing process productivity and energy utility simultaneously. Therefore, an intelligent approach to PBF process optimization is essential for future material development and product quality control. Such an optimization approach can be derived systematically based on both numerical studies and experimental validation.

In this paper, an intelligent optimization system for PBF for thermoplastics is proposed to establish quantitative relationships between process parameters and multiple objectives, including mechanical properties, productivity, and energy consumption. Material/powder behaviors, process parameters, and process conditions are critical factors of intelligent process optimization. Melting pool features are predicted based on numerical analysis and validated based on experimental characterization. Quantitative functions of melting depth/width with respect to process parameters are also derived. Therefore, contour maps of operation windows and desirable energy density ranges can be identified for specific thermoplastics. Based on an improved understanding of the relationships between PBF processes, microstructures, and mechanical performance, a data-driven approach requiring a small set of experimental and simulation data is introduced to establish correlations between process parameters and the mechanical tensile strength of printed materials. Next, mechanical properties, productivity, and energy efficiency are described as functions of melting depth/width, which can be correlated with process parameters. Inversely, the newly derived relationships between process parameters and mechanical property/energy efficiency maps can serve as effective guidelines for material selection and process parameter determination. The proposed approach for systematic optimization is widely applicable to different types of thermoplastics and can facilitate the establishment of a reliable knowledgebase for AM technologies.