Complete simulation process chain for the manufacturing of braided composite parts

Abstract

A complete simulation process chain has been used to predict the production and subsequent injection of over-braided textile preforms. A range of mandrel geometries and braiding configurations were used to illustrate how these factors affect the resin injection of the part. Braiding simulations were first completed, predicting the geometry of the braided textile throughout the mandrel. Following this, a range of multi-layered unit-cells were modelled, capturing the variations in geometry. These virtual stacks were produced with both no and maximum in-plane ply shift so as to capture the range of stacking configurations possible. Following a compaction simulation of these stacks, their in-plane permeability tensor was predicted and used to identify the permeability of the braided preform at different regions. This was used to predict the propagation of the resin flow front, highlighting the effects that the mandrel geometry, braiding process parameters and stacking method have on the resulting resin injection.

Introduction

The use of fibre reinforced polymer composite parts has been rapidly increasing, taking advantage of their high weight specific mechanical properties. Highly automated production processes are essential in order to enable their use in products which are often mass-produced. The braiding process is one such manufacturing method. The process is highly repeatable, creates little raw material waste [1] and provides high levels of flexibility in terms of part geometry.

The braiding process may typically be classified as 2D or 3D [2]. The study presented here focuses on the 2D over-braiding process. An over-braiding machine is shown in Fig. 1. Two groups of carriers, which carry yarn spools, are moved by horn gears on the braiding machine in opposite directions as they cross each other. The yarns are winded up at the same time, led over a guide ring in the centre of the machine and placed on a mandrel. By crossing the carriers, a textile is generated. Depending on the sequence of this crossing, different braid patterns can be achieved. An overview is given in [3]. This study focuses on the regular triaxial braid pattern.

The dry braided preform is injected with resin in order to create the final, hollow, composite part. A variety of different methods from the Liquid Composite Moulding (LCM) family may be used for this [4]. In cases where a Resin Transfer Moulding (RTM) [5] process is used as the filling method, the outer contour of the composite part is determined by the mould geometry used.

The geometry of the braided part is determined by the mandrel geometry: the inner surface of the braided part is equivalent to the outer contour of the mandrel. The architecture of the braided textile is typically characterised by the braiding angle, $\alpha$ and spacing, $s$ (as shown in Fig. 2). These parameters may vary throughout the braided part due to changes in the mandrel geometry as well as variations in the braiding process parameters.

A crucial parameter influencing $\alpha$ and $s$ is the mandrel take-up speed $v_{\text{M}}$ relative to the angular velocity of the horn gears ${\omega }_{\text{HG}}$. For cylindrical mandrel geometries, this ratio can be described by Eq. (1), where $d$ is the diameter of the mandrel and $N_{\text{HG}}$ the number of horn gears. Other prediction methods have to be developed for complex mandrel geometries, e.g. spoiler [6] and A-pillars and Cashboxes [7], with the most promising being the use of Finite Element Analysis (FEA). This can be used to simulate the braiding process, providing highly detailed information of the preform properties [8], [9].

$$\text{tan}(\alpha )=\frac{d{\omega }_{\mathit{HG}}}{v_M{N}_{\mathit{HG}}}$$

The textile architecture of the braided part has a significant influence on both the mechanical and processing properties of the part. By carrying out FEA braiding simulations, data is obtained about the architecture and its variations and may be used as an input for further simulations. An example of this is LCM filling simulations.

The use of LCM filling simulations as process design tools is increasing in industry. These simulations are used to accurately predict fill time, flow front advancement and dry spot formation during the injection of fibre reinforced parts. These tools are increasingly utilised to enable the production of complex high quality parts using the most efficient conditions [10]. In these simulations, the resin flow during the LCM manufacturing process is commonly modelled using Darcy’s law [11]. The relation is given in Eq. (2), where q is the volume averaged Darcy velocity; μ, the fluid viscosity; P, the fluid pressure and K, the permeability tensor. These simulations therefore require knowledge of the reinforcing material’s permeability characteristics.

$$\mathbf{q}=-\frac{1}{\mu }\mathbf{K}\nabla \text{P}$$

Permeability, K, is a measure of the ability of a reinforcement material to transmit fluids. It is an important material characteristic that determines the flow propagation of the resin during LCM manufacturing processes and is an indispensable input into LCM process simulations [12]. The permeability behaviour of reinforcing textiles is a strong function of the textile’s complex architecture [13], the amount of compaction applied to it (and resulting fibre volume fraction) as well as any nesting interactions between individual layers in the preform.

Braided components may be created using a range of different mandrel shapes [14]. When the cross-section of the mandrel is not constant, the spacing between the axial yarns varies. Additionally, when a variable mandrel take-up speed is used, the resulting braiding angle changes [9]. These changes in the structure of the braided textile create regions with different permeabilities further complicating the prediction of the filling step in the manufacturing process.

This paper presents a complete simulation process chain which may be used to optimise the manufacturing process of braided parts. First, the braiding process is simulated, predicting the resulting textile architecture at different locations on the mandrel (local definition of $\alpha$ and $s$). The predicted architecture is used to generate a range of textile models of the reinforcement. These models are stacked so as to create four-layer preform structures with both no and maximum in-plane ply shift. Compaction simulations are applied to these models, reflecting the compaction that happens to the preform during the manufacturing process and the resulting models are then used to predict the in-plane permeability tensor. Focus is placed on capturing the effect that textile geometric variability (resulting from the braiding process) has on the permeability as well as assessing the influence that different in-ply shifts (and resulting nesting) have. Each of the simulation components have already been validated experimentally in the past, and this is already presented in literature [8], [22], [23], [31], [32].