The input layer of BPANN includes five neurons determined by the above discussion: sound pressure RMS amplitudes of noise samples in five different Critical Band Rates.
Based on the above definitions, the BPANN is able to be trained to project the input of Bark scaled sound pressure amplitude vector to the output of predicted Distraction Level.
The BPANN training procedure is performed by using Matlab program.
OPTIMIZATION OF DISTRACTION LEVEL WITH ANE SYSTEM AND BPANN
The modeled BPANN predicts and verifies the optimized Distraction Level before ANE system driving the loudspeaker.
Equation (1) is the mathematical expression of modeled BPANN which inputs sound pressure amplitudes and outputs Distraction
Training data is used to train the BPANN to capture the relationship between input and output.
And when constructing a BPANN, many network parameters should be set.
Under each test condition, the training of the BPANN is carried out with ten different sets of initial values of weights and bias produced randomly by computer.
In the past, the prediction performance of a trained BPANN was usually conducted by choosing randomly and manually several testing pairs from the experimental data that was never seen by the BPANN.
Comparison of Prediction Performance of BPANN and Taguchi's Method in Determining Quality Factors (Shrinkage and Warpage) and Optimal Process Condition
In Figure 8, red line represents the actual measured value ofDO and the blue line and green line represent the predicted value of dissolved oxygen content using BPANN
prediction model and BPANN
optimized by PSO prediction model, respectively.