A cognitive approach for laser milled PMMA surface characteristics forecasting

Leone, Claudio; Matarazzo, D.; Genna, Silvio; D’Addona, Doriana M.

doi:10.1016/j.optlastec.2018.12.025

Cited by 16 publications

(17 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This choice may introduce a limitation in the determination of higher-order interactions (as a matter of fact, no high order interaction can be checked), as well as in terms of model robustness. In addition, the replications adoption is not useful for ANNs modelling since it introduces errors during the trial phase and prevents the fitting [46,62]. However, also considering the number of tests carried out (108 tests or treatments), this choice is an acceptable compromise in terms of the resolution of the statistical model and the number of tests to perform.…”

Section: Methodsmentioning

confidence: 99%

“…More specifically, the type of network (Back Propagation, …) and the number of nodes in the input, hidden and output level strongly influence the predictive response. Based on bibliographic analysis [42,48,49,65,66] and relevant background [46,62,67], it was chosen to adopt feed-forward back propagation neural networks (FFBPNN) since the latter is particularly suitable to understand functional relationships between given inputs and outputs. In FFBPNN, two-or more layer cascade-network can learn by examples any finite input-output relationship arbitrarily well given enough hidden neurons.…”

Section: Artificial Neural Network Configuration Set-upmentioning

confidence: 99%

“…Since, as can be inferred by Section 3.1, all the control factors are involved in the definition of all the response variables (directly or by way of an interaction) according to [46,48,62], the input nodes were selected equal to the control factor numbers: 5 (i.e. each input node represents a factor numbers).…”

Section: Artificial Neural Network Configuration Set-upmentioning

confidence: 99%

“…The advantage of this approach is that, under the appropriate conditions, ANNs are able to model a complicated relationship for the modelling to adapt to the different problems [38,39], solving engineering issues in industrial applications, such as in-process control [40,41], and optimisation procedures [42][43][44], as well as in non-destructive evaluation [45]. Moreover, ANNs can allow robust models even in the presence of a high noise ratio [38,46,47]. However, ANNs required an adequate number of experimental trials to train the network (i.e.…”

Section: Introductionmentioning

confidence: 99%

“…However, ANNs required an adequate number of experimental trials to train the network (i.e. large experimental dataset), an appropriate selection of the network configuration [38,46,48,49] and a training phase that can be significantly long (several hours).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

An integrated approach for the modelling of silicon carbide components laser milling process

Leone

Genna

Tagliaferri

2021

Int J Adv Manuf Technol

View full text Add to dashboard Cite

The paper deals with characterisation and modelling of laser milling process on silicon carbide hard ceramic. To this end, a Yb:YAG pulsed fiber laser was adopted to mill silicon carbide bars. Square pockets, 5×5 mm2 in plane dimension, were machined at the maximum nominal average power (30W), under different laser process parameters: pulse frequency, scan speed, hatching distance, repetitions and scanning strategy. After machining, the achieved depth and the roughness parameters were measured by way of digital microscopy and 3D surface profiling, respectively. In addition, the material removal rate was calculated as the ratio between the removed volume/process time. Analysis of variance (ANOVA) was adopted to assess the effect of the process parameters on the achieved depth, the material removal rate (MRR) and roughness parameters, while response surface methodology (RSM) and artificial neuronal networks (ANNs) were adopted to model the process behaviours. Results show that both RSM and ANNs fault in MRR and RSm roughness parameters modelling. Thus, an integrated approach was developed to overcome the issue; the approach is based on the use of the RSM model to obtain a hybrid Training dataset for the ANNs. The results show that the approach can allow improvement in model accuracy.

show abstract

Section: Methodsmentioning

confidence: 99%