Any material removal process utilising a laser beam as a cutting tool presents many benefits such as fast cutting speed (and thus higher productivity rates), high accuracy, good control of the heat affected zone and minimal deformation, especially when low power CO2 lasers are used [1, 2]. Low power lasers are employed to process polymer based materials in a wide range of applications [3, 4]. During the laser cutting process, the material melts and evaporates quickly, mainly due to the favorable material thermal properties and the beam density [5–7]. Laser processing of the PMMA thermoplastic is used in various applications like automotive and transportation (bumpers, panels, fenders), electronics (screens made from PMMA, covers for PC or 3D printers, etc.), furniture (door canopies, functional balustrades), welfare (cabinets), lighting (transparency and brilliance optical applications), etc. [8–13].
The cutting quality is characterised through a number of features, such as the kerf width, the kerf angle, the roughness of the edge, the formation of burrs, and the depth of the Heat Affected Zone(HAZ) to name few [1]. For assessing the state of the art, a thorough literature review was undertaken. A number of academic libraries and databases were queried using relevant search strings, such as “laser” AND “cut” AND “PMMA”. Scopus indicated that over the last 20 years more than one hundred papers have been presented on this topic. However, very few researches have been focused on PMMA kerf width and kerf angle optimisation and modeling (nine was found using “laser” AND “cut” AND “PMMA” AND “kerf”). In the following paragraphs, the more relevant and significant studies are being reviewed and discussed.
Kerf width is one of the most studied factors for the laser cutting quality performance [14–16] of thin thermoplastic materials and is affected by a number of process parameters such as the focusing distance, spot diameter, cutting speed and laser power [2]. A wide experimental range of laser power and cutting speed values has been investigated in the literature [17–19]. The kerf characteristics (upper and bottom width, melted transverse area, melted volume per unit time and mean roughness values on cut edges) were measured by Caiazzo et al. [19]. The laser power (200-1400W) and the cutting speed (18-150mm/s) were optimised for the CO2 laser cutting of polymeric materials (PE, PC, PP) of different thicknesses (2-10mm) using the ‘trial and error’ experimental method. The focal length was 127mm and the focal spot diameter was adjusted to 0.2–0.3 mm. Such process parameters resulted in kerf widths (upper and bottom) between 0.1 to 0.5mm and average roughness (Ra) of up to 5µm. The researchers however did not propose any model for predicting the process performance. They concluded that lower power values are suitable for cutting polymeric materials and that the laser cutting speed is a very important parameter.
A preliminary evaluation of the results of the cutting parameters in a range of polymeric materials (PMMA, PC, PP, etc.) was carried out by Davim et al. [1]. They studied the quality of cut (burr length and heat affected zone depth) without trying to model the quality metrics. Thin (6mm) PMMA plates were cut using CO2 laser (high frequency, continuous mode, Gaussian profile) with power in the range between 350 and 650 Watt and cutting speed between 25 to 58 mm/s [17]. The focal spot size was about 0.3mm and the focal depth about 1.5mm. Surface roughness parameters (Ra, Rt, Rz, Rp), dimensional precision and the depth of the heat affected zone were studied and optimised using evolution diagrams of each metric according to the cutting speed and laser power. They have measured the depth of the HAZ between 0.12 to 0.37 mm, without burr and low average surface roughness (Ra < 1µm). It was concluded that when the laser power increases, the depth of the HAZ also increases and the roughness decreases while the opposite effects are being observed when the cutting speed increases. No mathematical model (analytical or experimental) was developed for the above studied kerf characteristics.
The effects of laser power (0.275-2.5W) and cutting speed (7-64mm/s) of CO2 laser on the width, and depth of micro-channels manufactured from (1mm) PMMA of different molecular weights have been investigated experimentally by Nayak et al. [18]. The laser had a wavelength of 10.6µm, lens focal length of 50.8mm, spot size up to 0.13mm and a maximum power of 25W. They found that the depth increases with an increase in power or a decrease in beam speed or a decrease in molecular weight. The width increases with a laser power surge. As with the previous studies, no predictive models for depth and width were proposed.
The impact of the spot diameter (focal length 200-210mm), laser beam speed and incident power during PMMA miniaturised structures processed by a low power CO2 laser (50W, Gaussian continuous wave, minimum spot radius about 0.25mm) have been investigated experimentally by Romoli et al. [3]. By incorporating an analytical model for the depth of cut and an empirical model for the width, the laser power and speed were optimized, achieving depths between 0.05 and 0.6mm as well as various widths from 0.15 to 0.4mm
Varsi and Shaikh [16] used a low power CO2 laser (25W, Gaussian, continuous) for cutting 8mm thick PMMA plates. A three-parameter, five-level full factorial experimental design has been applied to investigate the impact of the power (13-23W), speed (202-586mm/s), and a number of passes (1–5) on kerf angle. They developed regression predictive models for controlling the process and concluded that higher number of passes, lower speed and higher power resulted in a lower kerf angle.
Empirical or soft computing (neural networks or genetic algorithms) models based on experimental design methods, were also developed in the literature and they are employed more and more the recent years [20]. The most characteristic ones are presented in the following paragraphs.
Polymeric materials (PMMA, PC and PP) were investigated when continuous CO2 laser cut by Choudhury and Shirley [21]. Laser power (200-400W), speed (3.3-6.6mm/sec) and pressure (2.5-3.5Bar) were tested and the kerf characteristics (HAZ depth, Ra, dimensional deviation) were investigated using a linear ANOVA analysis. All plates had a 3mm thickness. The central composite experimental design was applied to reduce the experiments. They measured average surface roughness for PMMA plates up to 9µm with a standard deviation of 1.1µm and HAZ depth between 130–210µm. They extracted predictive models based on the response surface methodology (RSM). It was concluded that the HAZ depth for all polymers increases with the increase of the power or decrease of the beam velocity and that the surface roughness was most affected by the laser speed and the air pressure.
Nukman et al. [22] studied the kerf width during CO2 laser cutting of Perspex glass thin plates (3-5mm). Laser power (100-500W), cutting speed (3.3-20mm/s), stand-off distance (1-10mm) and gas pressure (0.5–4.5 bar) were tested. A four-parameter three-level Taguchi design based on L9 orthogonal array was used and two models: an FFBP and an optimized GA-Taguchi NN were developed for the kerf width predictions. They measured kerf widths between 0.5-1.5mm and concluded that the developed FFBP-NN was not appropriate for accurate predictions. Consequently, they used a hybrid Neural Network-Genetic Algorithm (NN-GA) model to improve predictions. With the hybrid model, the error was reduced to below 10%.
Hossain et al. [15] studied the stand-off distance (1-10mm), gas pressure (0.5–4.5 bar), velocity of cut (3.3-20mm/s) and laser power (100-500W), according to the minimum kerf width. The focal length was kept constant at 127 mm. They developed an intelligent fuzzy expert system (FES) model to predict the kerf width in CO2 laser cutting of (3mm) PMMA thin plates. The FES model was completed by eighty-one (81) experiments based on a four-parameter three-level full factorial design. They measured kerf widths between 0.5 and 1.5mm and concluded that the predictions made by the FES model were satisfactory in terms of relative error and goodness of fit and that can be used in PMMA laser machining simulation.
Moradi et al. [8] studied the laser power (20-40W), the cutting speed (2-18mm/s) and the focal plane position (0 -4mm) during PMMA CO2 laser cutting (3.2mm). They used a central composite design (three-parameter five-levels, 17 experiments), quadratic polynomial functions, and response surface methodology (RSM) to model the kerf characteristics (top and bottom kerf, kerfs ratio, HAZ and Ra). They measured the HAZ depths between 0.15 to 0.45mm, kerf widths (top and bottom) between 0.15 to 0.6mm, and average surface roughness (Ra) between 1–15µm. They found that increasing the cutting speed or reducing the focal position level or reducing the laser power resulted in a reduction for the bottom kerf. They also concluded that when the focal plane was lower than the upper part surface the kerf characteristics had been improved.
Morady et al. [23] developed a finite element model to simulate the CO2 laser cutting of (3.2mm) PMMA plates. They concluded that increasing the laser cutting speed from 4 to 20 mm/s and decreasing the laser power from 50 to 20 W, results in the reduction of the heat-affected zone. Also, it is mentioned that the depth of the kerf for different laser power values, speed and laser focal plane can be predicted by applying the proposed FEM model.
The effects of power (120-150W), velocity (1.66-5mm/s), gas pressure (1-3bar) and plate thickness (4-12mm) on the upper and down kerf width as well as on kerf angle have been experimentally investigated by Elsheikh et al. [24]. A continuous CO2 laser was used (wavelength, 10.64 µm; constant stand-off distance, 6mm; focal length, 50 mm). An L18 Taguchi table (one parameter with two levels and three parameters with three-levels) was implemented and the kerf characteristics were statistically analysed using the ANOVA analysis. The kerf widths (top and bottom) were measured between 0.2 and 0.9mm and the kerf angles between 0.01 to 2 degrees (o). They determined that the most dominant parameter was the thickness of the sheet and when it increases, all the measurements of the 'kerf' decrease. The second most influential parameter was the cutting speed and found that when it increases the kerf widths decreases. Finally, they developed regression models and applied a genetic algorithm to optimize the process.
Concluding, all the above investigations studied theoretically or experimentally the kerf characteristics of the thermoplastic PMMA thin plates (3-12mm) within the following laser power and velocity ranges: (i) miniaturised structures: power bellow 2.5W and velocity up to 64mm/s [18], or power between 13-23W and velocities between 202-586mm/s [16]; (ii) Thin plate cutting with power higher than 100W and velocities up to 150mm/s [19, 1, 17, 21, 22, 15, 24], or power between 20-40W and velocities between 2-18mm/s [8, 23]. The kerf characteristics that measured were: (i) average surface roughness between 1–15µm, (ii) kerf widths (top and bottom) between 0.1 to 1.5mm, (iii) kerf angle between 0–2(o), and (iv) HAZ depth between 0.1-0.33mm.
The above literature review demonstrates that there is a gap of experimental work for laser power between 50-100W. The velocity and focal plane should be considered, too, when studying the kerf characteristics, as they are very predominant parameters [8]. Thus, after preliminary work in order to have a thorough cut for all experiments, it was decided to include as tested parameters: the laser power (82.5-97.5W), cutting speed (8-18mm/s) and stand-off distance (7-9mm). The thickness of the PMMA plates was selected to be 4mm as proposed by the literature review.
A three-parameter three-level full factorial experimental design was selected for investigating experimentally the kerf width characteristics (upper, middle and down and angle). Taguchi’s full factorial experimental design is proposed and analyzed in the following paragraphs [25]. In this work, a continuous CO2 laser is utilised. The kerf width in three levels (upper, middle and down) and kerf angle are measured and the effects of process parameters and their interactions are analysed using statistical analysis tools. The methodology that is proposed in [26, 27] regarding the implementation of statistical methods for the post-processing of experimental data, guides also the present work. Linear regression models gave lower accuracy predictions (ANOVA, R-sp < 0.90) and after interaction study between the parameters, FFBP-NN modelling was used to improve predictions, achieving R-sq bigger than 0.98.