Ultrasonic Welding of PEEK Plates with CF Fabric Reinforcement—The Optimization of the Process by Neural Network Simulation

Abstract

The optimal mode for ultrasonic welding (USW) of the “PEEK–ED (PEEK)–prepreg (PEI impregnated CF fabric)–ED (PEEK)–PEEK” lap joint was determined by artificial neural network (ANN) simulation, based on the sample of the experimental data expanded with the expert data set. The experimental verification of the simulation results showed that mode 10 (t = 900 ms, P = 1.7 atm, τ = 2000 ms) ensured the high strength properties and preservation of the structural integrity of the carbon fiber fabric (CFF). Additionally, it showed that the “PEEK–CFF prepreg–PEEK” USW lap joint could be fabricated by the “multi-spot” USW method with the optimal mode 10, which can resist the load per cycle of 50 MPa (the bottom HCF level). The USW mode, determined by ANN simulation for the neat PEEK adherends, did not provide joining both particulate and laminated composite adherends with the CFF prepreg reinforcement. The USW lap joints could be formed when the USW durations (t) were significantly increased up to 1200 and 1600 ms, respectively. In this case, the elastic energy is transferred more efficiently to the welding zone through the upper adherend.

Application of Neural Network Models with Ultra-Small Samples to Optimize the Ultrasonic Consolidation Parameters for ‘PEI Adherend/Prepreg (CF-PEI Fabric)/PEI Adherend’ Lap Joints

“…To achieve this goal, the optimal USC parameters were applied, which were determined through ANN simulation. The method for manufacturing the PEI-impregnated PEEK-based prepreg based on the CF fabric (Toray Cetex TC1200, Toray Industries, Japan) was described in a previous paper by the authors [ 38 ].…”

“…The following mechanical characteristics were used as the required properties of USC lap joints [ 38 ]: tensile strength ( σ ), MPa; elongation at break ( ε ), %; …”

“…Since it was not possible to simultaneously achieve extreme (maximum or minimum) values of all the functional characteristics of the USC lap joints, their acceptable ranges were specified, i.e., the optimality conditions ( Table 1 ). The quantitative values of the range boundaries were based on the previous experience of the authors [ 38 ] and the results of an analysis of data reported by other researchers. Accordingly, further search for the optimal combination of the USC parameters was carried out taking into account the values given in Table 1 .…”

“…At this stage of the analysis using a priori knowledge, only models with ‘spread’ values below 0.5 were recognized as inadequate. In this case, it was accepted that the τ parameter was not dominant [ 38 ] and the SOP areas could not be discontinuous. Obviously, such a gap was associated with a great relative distance between the points of the predicted optimal USC parameters and could be eliminated by supplementing the sample with intermediate data.…”

“…A previous paper by the authors [ 38 ] was devoted to the optimization of the US-consolidation (USC) parameters for the formation of USC lap joints of polyetheretherketone (PEEK) adherends, a carbon fiber (CF) fabric prepreg impregnated with polyetherimide (PEI), and two PEEK EDs [ 39 ]. Impregnation of the CF fabric with the polymer, characterized by a low melting point and a melt flow index (MFI) greater than that for PEEK, determined the specific development of the structure formation process.…”