matteo macchinistudent meeting – june 2014 motion control design for the new bws matteo macchini...
TRANSCRIPT
Matteo Macchini Student meeting – June 2014
Motion control design for the new BWS
Matteo MacchiniTechnical student
BE-BI-BLSupervisor:Jonathan Emery
Matteo Macchini
Outline
Student meeting – June 2014
• PSO method overview
• Detailed description of the system
• SVPWM implementation and testing
• Tuning results from the simulations
• Robust stability/performance analysis
• Conclusions – what’s next?
Matteo Macchini
Particle Swarm Optimization
Student meeting – June 2014
How does it work?
Given a multi-dimensional function:
• Initialises random particles (parameter values)• Computes cost • Updates position and speed of the particles making them move towards the best result obtained so far
Matteo Macchini
Particle Swarm Optimization
Student meeting – June 2014
Iteration 1
Matteo Macchini
Particle Swarm Optimization
Student meeting – June 2014
Iteration 10
Matteo Macchini
Particle Swarm Optimization
Student meeting – June 2014
Iteration 20
Matteo Macchini
Particle Swarm Optimization
Student meeting – June 2014
Iteration 30
Matteo Macchini
Particle Swarm Optimization
Student meeting – June 2014
Cost function
Matteo Macchini
Previously achieved results
Student meeting – June 2014
Current loop
Speed loop
Iteration 1 Iteration 4 Iteration 7 Iteration 10
Iteration 1 Iteration 4 Iteration 7 Iteration 10
Matteo Macchini
“Old” model
Student meeting – June 2014
Motor
Current sensor
Control system
PWM
IGBT inverter
Analog filter
Cable
Matteo Macchini
Simplified model(s)
Student meeting – June 2014
Cascade version
Main differences from classic version
• Ideal amplifier (no PWM-inverter-filter blocks)• Fully digital implementation and simulation (no Simulink-simscape
switch)• Typical cascade control system with no AW feature• Control analysis for iterative optimization
Control system
Motor
Current sensor
CableAmplifierControl analysis
Matteo Macchini
Simplified model(s)
Student meeting – June 2014
Parallel version
Control system
Motor
Current sensorCable
Amplifier
Control analysis
Differences from cascade version
• Parallel control system with no AW feature (designed entirely from scratch)
Matteo Macchini
Space Vector PWM
Student meeting – June 2014
Advantages [1] :
• Lower THD (Total Harmonic Distortion)• Greater PF (Power Factor)• Less switching losses• Lower computational cost
Principle:
Differently from classic SPWM, SVPWM transforms a three-phase sinusoidal wave into its PWM, considering the combination of the three inputs at once.
Matteo Macchini
Space Vector PWM implementation
Student meeting – June 2014
Simulink implementation
Obtained waveformLow-pass filtersSVPWM block
Output coherent with expectations
Couldn’t appreciate advantages (SO FAR!)
Matteo Macchini
Tuning strategy
Student meeting – June 2014
• Create system model• Initialize parameters• Implement system control based on desired dynamics• Compute a COST FUNCTION based on the obtained results• Modify parameters in order to minimize the CF
Matteo Macchini
Cost function computation
Student meeting – June 2014
As a cost function, the integral absolute error (IAE) between the desired motion profile and the results has been used.To improve its quality, some weights were added on the critical zones of the dynamic.
Matteo Macchini
Simulations and results
Student meeting – June 2014
Cascade design, 100 particles, 10 iterations
Matteo Macchini
Simulations and results
Student meeting – June 2014
Cascade design, 100 particles, 10 iterations
Matteo Macchini
Simulations and results
Student meeting – June 2014
Parallel design, 100 particles, 10 iterations
Matteo Macchini
Simulations and results
Student meeting – June 2014
Parallel design, 100 particles, 10 iterations
Matteo Macchini
Simulations and results
Student meeting – June 2014
Cost function evolution
Matteo Macchini
Cable modeling
Student meeting – June 2014
For the last simulations, a cable model has been implemented.It takes into account:
• Cable self attenuation• Cable cross-talk• Cable length
Matteo Macchini
Performance/robustness
Student meeting – June 2014
ROBUST controller: results are “similar” into a given range of uncertainty.
PERFORMING controller: the reference profile is followed “properly”, i.e. the cost function has a “low” value.
A controller should be performing in order to guarantee the control quality for the tested device. A controller should be robust in order to guarantee the control quality for a family of devices working in different technical/environmental conditions.
Matteo Macchini
Robustness test
Student meeting – June 2014
Robust controller: results are “similar” into a given range of uncertainty(cable length variable between 1m and 300m)
Robust controller Non-robust controller
Matteo Macchini
Robustness test
Student meeting – June 2014
Robust controller: results are “similar” into a given range of uncertainty(cable length variable between 1m and 300m)
Robust controller Non-robust controller
Matteo Macchini
Robustness test
Student meeting – June 2014
Robust controller: results are “similar” into a given range of uncertainty(cable length variable between 1m and 300m)
Robust controller Non-robust controller
Matteo Macchini
Robust synthesis
Student meeting – June 2014
IDEA:
• Launch tuning algorithm several times
• Test robustness
• Check if robust controller have similar parameters and try to reproduce them
Non-robust controller
Matteo Macchini
Conclusions
Student meeting – June 2014
• Particle Swarm Optimization can be used to tune controller
parameters in the considered system
• SVPWM will help increasing the quality of the amplifier
• Cascade architecture: good performances, very good robustness
• Parallel architecture: very good performance, hard to make robust
• Robust controllers can be tuned using iterative methods
Matteo Macchini
In the future…
Student meeting - May 2014
NEXT GOALS
• Validating the previous system
• Implement control on the bench
• Make it work properly
• Study VHDL/hardware design in order to port it on FPGA
Matteo Macchini
References
Student meeting – June 2014
[1] Waheed Ahmed, Syed M Usman Ali, “Comparative study of SVPWM & SPWM three phase voltage source inverters for variable speed drive”