**1. Introduction**

By combining the attractive features of different constituents, composite materials can be properly designed for specific applications, thus replacing traditional materials in an ever-increasing variety of products and applications in aerospace, automotive, marine, nanocomposite, construction, etc. [1–3]. More recently, the emerging trend towards lightweight, high performance, high functionality and high sustainability components is driving the research towards nanocomposites, multi-material hybrid structures and composites derived from renewable resources or from reusing approaches [4–9]. The growing demand of polymer matrix composites in several fields has also driven the research and development of new manufacturing processes. The challenge of manufacturing high quality and complex geometry components at relatively low cost and fast cycle times has pushed toward the development of liquid composite molding (LCM) processes [10,11]. These technologies usually involve the use of dry reinforcement preforms, an assembly of dry fiber layers that have been pre-shaped to the form of the desired product and bonded together using a binder resin, which are impregnated by resin injection or infusion into a mold [12,13]. Autoclave curing is usually used to limit the final void content, which remains a significant issue, related to the filling process, fiber wetting and tows impregnation. To achieve the required mechanical performance of the composite parts, multi-layered preforms with different architecture are generally used, such as unidirectional non crimped fabric and plain or twill woven fabric [14]. Furthermore, LCM requires low viscosity resins, which upon crosslinking would be brittle, being made of low molecular weight oligomers. Therefore, the higher

molecular weight oligomers of the resin formulation, solid at room temperature, are added to the reinforcement as fiber coating or powders that are soluble in the matrix. These oligomers play also a role in preform fabrication, acting as low melting tackifying agent able to bind a ply to another upon heating [15]. The required performances are achieved in the final component only if a fast and complete filling of the preform with liquid resin is obtained thanks to the optimization of LCM process [16,17]. A suitable process design is necessary, which requires knowledge about different parameters affecting the filling behavior, such as mold geometry, resin viscosity, number and position of inlet and vent ports, etc. [18]. Besides the previous parameters, one of the most critical is the permeability of the reinforcement, defined as the resistance exhibited by the fibrous reinforcement against the resin flow [19]. This property determines the impregnation of the textile reinforcement with liquid resin and strongly depends on the reinforcement architecture, fiber orientation and volume fraction and on the procedure adopted to assemble and stabilize the plies into the preform [20]. The variability of textiles and the preforming operations affect the permeability of the fibrous reinforcements, resulting in possible unexpected or unwanted resin flow patterns [21]. Generally, permeability data are not available from the fabric manufacturer and must be determined depending on the fiber volume fraction. An accurate characterization of the reinforcement permeability is therefore fundamental to estimate the optimum process parameters for manufacturing high-quality components.

The permeability, K, of porous media is defined by Darcy's law [22] correlating the flow velocity V, pressure drop, ∆P, and viscosity, η, in a unidirectional flow over the length of the porous medium L:

$$\mathbf{V} = -\frac{\mathbf{K}}{\eta} \frac{\Delta \mathbf{P}}{\mathbf{L}} \tag{1}$$

The permeability of unidirectional preforms is an anisotropic property and is described by a second-order tensor [20]. Generally, by accounting for symmetry and assuming no coupling between inand out-of-plane flow, only three components of the permeability tensor are not zero: (i) longitudinal in-plane permeability K1, where in-plane refers to the reinforcement ply; (ii) transversal in-plane permeability K2, oriented perpendicular to K<sup>1</sup> and lower than K1; (iii) out-of-plane permeability K3, oriented perpendicular to K<sup>1</sup> and K2, lower than K<sup>1</sup> and of the same order of K<sup>2</sup> [21].

Different measurement methods of permeability are available in the literature [20,23,24]; each is capable to measure saturated or unsaturated permeability. Saturated permeability is a steady state permeability measured under a constant flow rate when the fiber reinforcement is fully saturated by the test fluid. Unsaturated permeability, also named unsteady-state or transient permeability, is measured when the reinforcement in the mold is progressively impregnated by a test fluid injected under a constant pressure [25]. However, there is a complete lack of standardization in the methods and experimental set-up, and the data, obtained using different methods and different fabric architecture, are not consistent [25]. Some benchmark studies demonstrated that when the same measuring procedure is used, the results are very different if different parameters are used, such as in terms of injection pressure and test fluid [26]. Most studies have been focused only on the in-plane permeability [27–29], disregarding the out-of-plane permeability, which is more difficult to measure accurately [30]. However, the out-of-plane (or through thickness) permeability is the dominant property in the infusion of large and flat panels with a high thickness when a resin distribution net is used. Therefore, its determination is of fundamental importance for the efficiency and robustness of LCM processes. The traditional video recording methods used for monitor flow front in RTM processes and determining in-plane permeability, cannot work well in the case of through-thickness permeability [31].

Several nondestructive methods have been investigated based on thermocouples [32], on the reflection at the flow front of electrical signals from thin metallic wires inside the fibers [33] guided mechanical waves [34], piezoelectric sensors [35] dielectric sensors [36] embedded in the fibers and pressure sensors positioned in the mold [37,38]. All of these measurement methods require both direct contact with the resin in order to detect the flow-front position or complete mold-filling and sensor embedded in the preform or into the mold [39,40]. The sensor embedding is not always allowed and is hence limited in application. Mounting sensors into the mold can affect the vacuum tightness of the mold and the sensors usually leave marks on the part surface so they can be only located at the border but not in the areas of interest [39].

Ultrasonic wave propagation has been widely recognized as a nondestructive testing (NDT) method applied to for the estimation of the physical and mechanical properties or the damage composite materials [41–45] and for process monitoring of composite materials [46,47]. Even the application of high intensity ultrasound in composite processing and joining present the potential for online monitoring [48–50]. Some research works proved the feasibility of using ultrasonic imaging systems for flow front monitoring even if the technique was limited to a single ply or thin preforms due to the severe attenuation of the ultrasound waves [31,51]. Stoven et al. [52] developed a set-up based on two ultrasonic probes operating in through transmission mode for flow monitoring and permeability determination in thickness direction. The ultrasonic based method presents several advantages: i) absence of any direct contact between the transducer and the composite and of any embedded sensor as the sound waves can be send through the mold wall; ii) low cost; iii) no disturbance to the fiber stack and liquid flow; iv) measurement of unsaturated permeability; v) reduced effort for preparing the experimental set-up; vi) no need of transparent tools; vii) compatible with all the fiber typologies [10,39,53].

Despite the numerous advantages, these ultrasonic based methods are still at a laboratory level due to the experience required to manage weak acoustic signals and the limitation of the resolution depth in thick preforms [53,54]. Moreover, the ultrasonic set-up described in the literature works in through transmission method with two ultrasonic transducers acting one as emitter and the other as receiver of ultrasonic waves [52], but sometimes the two-side access to the composite part is very difficult to achieve.

The aim of this work is to present a new experimental set-up for the measurement of unsaturated through thickness permeability based on the ultrasonic wave propagation in pulse echo mode, i.e., by using only one ultrasonic transducer working both as emitter and receiver of ultrasonic waves. Compared to the ultrasonic based systems previously reported in the literature, the proposed set-up enables one-side access, which is of great importance in composite manufacturing. This work stands out of the few previous related papers [10,52,53] by using the pulse echo technique with a single transducer, working both as emitter and receiver of ultrasonic waves, positioned at one side of the preform. The literature studies, in fact, use the ultrasonic transmission method, which is not always feasible when the composite is not accessible from both sides. The single ultrasonic probe, coupled to a vacuum assisted resin infusion (VARI) system, allows in-situ monitoring of unsteady preform impregnation. The system has been applied to multi-layered carbon fiber preforms. The robustness of the system has been tested on three different carbon fiber preforms used in aerospace field, each obtained by stacking balanced or unidirectional plies and then tested in vacuum infusion experiments. Saturated out-of-plane permeability has also been measured by a traditional gravimetric method and the results obtained by the two techniques have been compared. Finally, another novelty element is the validation of the permeability data, calculated from ultrasonic measurements, by some analytical models.
