Fully Parallel Learning Neural Network Chip for Real-time Control

1
Fully Parallel Learning Neural Network Chip for
Real-time Control

• Students (Dr. Jin Liu), Borte Terlemez
• Advisor Dr. Martin Brooke

2
Combustion Instability Control -Simulation
Results Review

• Simulated Neural Net and Combustion
• One-frequency Results
• Multi-frequency Results
• Parameter Variation Results
• Added Noise Results

3
Simulation Setup
4
One Frequency Plant without Control
5
One Frequency Result
f 400Hz b ?
6
Two-Frequency Results
f 400Hz 700Hz b ?
7
Parameter Variation Results
f 400-600Hz z 0-0.008 b 1-100
8
10 Added Noise Results
f400Hz z0.005 b1
9
Neural Network Chip Control of Combustion
Instability
10
Experimental Setup
11
Test Box
12
Experimental Result
f 400Hz z 0.0 b 0.1
13
More Results
f 400Hz z 0.0 b 0.1
14
More Results
f 400Hz z 0.0 b 0.1
15
Details of Initial Oscillation Suppression
Error Decreases
f 400Hz z 0.0 b 0.1
16
Details of the Continuously Adjusting Process
Error Decreases
f 400Hz z 0.0 b 0.1
Error Increases
17
Experiments with Run Time
f 400Hz z 0.0 b 0.1
18
Experiments with Damping Factor z0.001
f 400Hz z 0.001 b 0.1
19
Experiments with Damping Factor z0.002
f 400Hz z 0.002 b 0.1
20
Summary of NN Chip Control of Simulated
Combustion Instability

• The NN chip can successfully suppress the
  combustion instabilities within around 1 sec.
• The NN chip continuously adjusts on-line to limit
  the engine output to be within a small magnitude.
• I/O card delay and engine simulation delay
• 30 times longer than real time
• Weight leakage
• Fixed learning step size

21
Improved Neural Network Chip in 0.35- mm Process

• Seven Time More Neuron Cells
• Two layers
• Each layer has 30 inputs instead of 10
• Totally 720 neurons instead of 100
• Adaptive Learning Step Size
• Capacitor charge sharing scheme
• Current charging and discharging scheme
• Partitioned Error Feedback
• Synchronized Learning, without stopping the clocks

22
New Chip
23
Chip Architecture - Block Diagram
24
Cell Schematics
Cell
Cell
25
Full Chip Spice Simulation after Parasitic
Extraction

• Shift Register
• Weight Updating
• Current Outputs at Pads
• Clocking Scheme

26
Shift Register
X1ms First 0 to 1 at sh_in
X15.4ms First 0 to 1 at sh_out_end 720 cycles of
delay
X1.48ms First 0 to 1 at sh_out_1r 24 cycles of
delay
27
Weight Updating
Shifted in voltage
Weights
28
Output Currents at Pads
29
Clocking Scheme for Learning
One clocking cycle is 20 ms
30
Conclusion

• Extensive software simulations to provide a
  solution for real-time control using the RWC
  algorithm, with direct feedback scheme
• Successful application of the analog neural
  network chip to control simulated dynamic,
  nonlinear system
• Improved chip resulted from the extensive
  hardware experiments
• Automated test method and system

31
Future Works

• Acoustic Oscillation Suppression
• Test of the New Chip
• Real Combustion System Control
• Third Generation Chip (10,000 Weights)

32
Acoustic Oscillation Setup
33
The Two Layer Board