1. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Presenter: Prof. Wei Chen, Northwestern
PIs: Wei Chen, Jian Cao, Northwestern
Brad Kinsey, UNH

2. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Industrial Needs and Relevance:
Identify metal forming uncertainties
Parameter variations in materials, tooling and processes
Imperfect characterization
Compensate sheet metal springback
Increase of the material formability
Control material property / microstructure
Increase the process flexibility
Variations in
product
properties
Problems in sheet metal forming
Lim et al., 2008
NSF I/UCRC Planning Meeting

3. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Industrial Needs and Relevance:
Novel sensors measure forces, material draw-in, and strain history
Control algorithms to follow prescribed critical trajectory and achieve accurate feedforward and feedback controls
Efficient and flexible actuators to achieve controlled motion
Allwood et al, CIRP, 2016
NSF I/UCRC Planning Meeting

4. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Project Objectives:
Identifying and characterizing uncertainties in metal forming system:
Model errors
Disturbances
Identifying effective actuators, sensors and control algorithms
Integrating data analytics with sensing data for improved product properties
NSF I/UCRC Planning Meeting

5. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Approach/Methodologies:
Identify various actuators and sensors for implementation of control process
Transducer to detect draw-in distance
Mahayotsanun, et al, 2005
Segmented die with local adaptive controller
Cao et al, 2001
NSF I/UCRC Planning Meeting

6. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Approach/Methodologies:
Develop a test component with critical features
Collect data on material, process, tooling and products for analysis;
Use data analytics tools for developing effective process control algorithms.
Actuators for control in deep drawing
Allwood et al, 2016
NSF I/UCRC Planning Meeting

7. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Deep Models
Recurrent Neural Networks
Reinforcement Learning
Approach/Methodologies:
Use state-of-the-art machine learning techniques to extract features based on prior manufacturing knowledge
Train (deep) Machine Learning models
Iteratively adjust feature space to achieve high predictive accuracy
Verify on validation dataset x1 h1 y x2 h2
Prediction x3
Hidden
States
Variables
Shallow Models
ARIMA (Autoregressive integrated moving average
ARIMA – Autoregressive integrated moving average
NSF I/UCRC Planning Meeting

8. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses𝐱 𝐲
Experiment 𝛉
Simulator
Approach/Methodologies:
Employ Bayesian model calibration and uncertainty quantification approach.
Identify unknown model parameters based on physical data
Correct model bias (model error) in predicting product properties
Experiment
Model
Experimental Error
Bias Function
Inputs
Low-fidelity model
Tests / High-fidelity model
Prediction mean
95% PI (prediction interval)
NSF I/UCRC Planning Meeting

9. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Approach/Methodologies:
Develop an adaptive Bayesian interface framework
Integrate data from off-line physics-based models, machine learning models, and on-line sensors
Design sensor locations and adaptive sampling strategies
Adaptive Bayesian interface framework for additive manufacturing process (DMDII sponsored project)
NSF I/UCRC Planning Meeting

10. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Deliverables:
Robust (validated) uncertainty prediction models for critical properties
Scientific characterization of non-linear features of forming process
Comprehensive analysis of the effect of different actuations and control algorithms
Potential control example of deep drawing
Polyblank et al, 2008
NSF I/UCRC Planning Meeting

11. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Budget and Timeline: Estimated cost of project is $500K for three years.
Task / Milestone
Year 1
Year 2
Year 3 Q1 Q2 Q3 Q4
Test Specimen Standardization
Novel Sensor Identification
Flexible Actuator Identification
Access Control Algorithm
Process Uncertainty Modeling
Achieve Desired Process Control
NSF I/UCRC Planning Meeting

12. 2. Industry 4.0: novel sensors, control algorithms, and servo-presses
Discussions:
Are the industrial needs and relevance accurately captured?
Are the objectives realistic and complete?
Are the approaches technically sound and appropriate?
Are there alternative implementation paths or better approaches?
Are the deliverables impactful to industrial partners?
Are the budget and timeline reasonable?
Are there conflicts with intellectual property or trade secrets?
NSF I/UCRC Planning Meeting