In the current world scenario, BLDC (Brushless DC) motors are highly used in current days and it is having several applications as per to its terribly high speed with an extremely compact size compared to the opposite motors with brushes, furthermore^{ 1}. Brushes are not available in BLDC motors, therefore electronically arrangements provide for commutation. BLDC motor is in fact a PMSM (Permanent Magnet Synchronous Motor) motor with trapezoidal back electromotive force ^{2}.

BLDC motors comprise many several striking properties like as easy speed control and torque –speed characteristics ^{1, 3}. Furthermore, the control of DC motor is also easy and no need of complex Hardware ^{4}. But, DC motors have main disadvantages relating to lifespan of brushes are the limited. A lower reliability happens caused by the brushes and therefore, the operation requires time to time maintenance of replacement ^{5}. ANN (artificial neural network) is a highly interconnected processing element which processes the information using their dynamic characteristics to the external input^{ 6}^{. }ANN can efficiently approximate dynamics without requiring detailed knowledge of the plant. Another advantage of ANN is their possibility of learning, which can reduce the human effort during the design of the controllers and allows discovering more effective control structures ^{7}.

The operating rule of a BLDC and conventional DC motor is same; i.e. based on internal shaft position feedback. In conventional DC motor, brushes and mechanical commutator maintains the feedback, while BLDC motor use Hall Effect sensors and optical encoders for feedback. Hall sensors works on the principle that when a conductor that is carrying current is placed in a magnetic field, a force is experienced by charge carriers depending upon the potential difference between both sides of the conductor. The developed voltage will be reversed if magnetic field direction is changed. Whenever rotor passes a hall sensor, it develops either a HIGH or a LOW level signal to indicate which rotor pole (N or S) has passed and hall sensors also monitor the position of the shaft.

The three phase brushless dc motor, the back- EMF and its phase current waveforms are shown in ^{8}.

Dynamic equations BLDC motor is described as following:

In Eq. (1), three phases voltages, three phases currents, three phases back emfs are represented by v_{a}, v_{b}, v_{c}, i_{a},i_{b}, i_{c}_{ }and_{ }e_{a}, e_{b}, e_{c}_{ }respectively._{ }The phase resistances, self-inductance and mutual inductance between phases are considered to be same in each phase and represented by R, L and M respectively.

The value of electromagnetic torque is with these parameters is

Here, parameter ω_{r} is used for mechanical speed of the rotor.

The motion equation is given by:

The relation between electrical speed ω_{e} and mechanical speed ω_{r} of a motor with P poles is given as ^{6}:

^{9}.

The w_{ij} is the link of weight parameter among j_{th}_{ }and i_{th} neuron at m_{th} layer, b_{mi}_{ }is the bias parameter of that layer at i_{th} neuron. Transfer function of the system at t_{th} neuron in m_{th} layer is represented as ^{10}.

The resulting activation function of neuron at m_{th} layer is represented by

MATLAB/SIMULINK software is used to simulate the BLDC motor as discussed in previous sections. PWM (Pulse Width Modulated) inverter is used for supply of BLDC motor. Hall Effect signals are decoded to develop gate signals of inverter. Various parameters for BLDC motor is shown in

Parameter |
Value |

Phase resistance of stator |
2.8750 ohm |

Phase inductance of stator |
8.5 mH |

Number of poles |
4 |

Number of phases |
3 |

Input voltage |
24 V |

Torque Constant |
1.4 N-m/A |

Voltage constant |
40 V/krpm |

Inertia Constant |
0.8 X 10-3 Kg-m2 |

Proportional constant |
0.0015 |

Integral constant |
0.25 |

To understand the system, a MATLAB designed simulink block, a schematic block diagram is shown in

ANN training block trains the system and its training window and validating the result is shown in

Data has been collected for training ANN for speed control of BLDC motor. After collecting data, Correlation between speed and voltage is 0.99999255 and Correlation between speed at (t-1) and voltage is 0.999992578. Correlation refers to the connection of two variables or more. Correlation is statistical tool that measures the strength of association between two variables and the direction of the relationship. Scatter diagram is a type of mathematical diagram using Cartesian coordinates to display both variables graphically ^{11}. In this process, the value of Karl Pearson correlation coefficient is more than 0.9 in each case, which advise a mathematical method for computing the magnitude of linear relation between the both variables. It means that data is correct to train the ANN for speed control of BLDC motor.

Neural networks will have lots of parameters and learning the optimum value of all parameters from large datasets in a serial implementation can be a very time-consuming task ^{12, 13}. For ANN learning process, multilayer feed forward back-propagation with gradient descent method is used. Multilayer feed forward networks are good for approximating any continuous function. Back propagation algorithm is simply iterative gradient descent on the empirical risk under squared error loss. This enables efficient calculation of the gradient.

In ref^{5}^{5}

Parameter |
PI Controller |
PID Controller |
Fuzzy Controller |
Proposed Neural Network |

Rise Time |
14.580 ms |
12.288 ms |
11.850 ms |
5.7 ms |

Overshoot (%) |
4.737% |
No overshoot |
0.274% |
No overshoot |

The study outcome provides dynamic execution of proposed controller based on neural network. The study has presented a neural network controller of a BLDC motor drive with closed loop control. The controller efficiency and sensitivity has been checked by MATLAB-Simulink software. Simulation outcomes show that torque ripple and current ripple are decreased which increase the drive performance. It is concluded that applying the load torque to the motor with neural network controller, motor speed will not be decreased. Speed control is done by using soft control technique ANN. Results obtained in this paper are justifying the selection of techniques for motor control as with step load motor speed is not decreased with the use of proposed neural controller. The proposed controller’s results are validated with the reference results obtained mathematically for the system. For ANN learning process, multilayer feed forward back-propagation with gradient descent method is used. The main philosophy of this paper is to investigate the parameters such as rise time, settling time and overshoot. With step load motor speed is not decreased. In conclusion, already mention that Simulation outcomes show that torque ripple and current ripple are decreased which increase the drive performance. With neural controller, BLDC motor has no steady state error. Neural network controller response offers high efficiency. The results give satisfactory outcomes of the dynamic execution of the BLDC motor under different load situations.