Presentation 20120324 - ziqi yang
- 1. Ziqi Yang 24 Match, 2012 Iterative Learning Control of the Injection Stretch-Blow Moulding Process 1 Intelligent System and Control School of Electronics, Electrical Engineering and Computer Science Queen’s University Belfast Email: zyang06@qub.ac.uk
- 2. 2 Outline Introduction Simulation Identification Control Plan
- 3. 3 1. Introduction – Blow molding Extrusion blow molding Injection blow molding Stretch blow molding
- 4. 4 1. Introduction – preform reheat Temperature distribution inside and outside each part from base to shoulder
- 5. 5 2. Introduction – Stretch blow moulding
- 6. 2. Simulation - Abaqus and Python Abaqus – finite element analysis Python – Abaqus based on Python
- 7. 7 2. Simulation - Abaqus and Python
- 8. 8 2. Simulation - Minitab and Main/Interaction effect analysis 4029175 4000 3000 2000 1000 1086 11010510095 4000 3000 2000 1000 100500 Massflow rate Mean Pressure Temperature Timing Main Effects Plot for Base Volume Data Means 1086 11010510095 100500 5000 3000 1000 5000 3000 1000 5000 3000 1000 Massflow rate Pressure Temperature Timing 5 17 29 40 rate flow Mass 6 8 10 Pressure 95 100 105 110 Temperature Interaction Plot for Base Volume Data Means
- 9. 9 2. Simulation - Minitab and Main effect analysis 4029175 11400 11250 11100 10950 10800 1086 11010510095 11400 11250 11100 10950 10800 100500 Massflow rate Mean Pressure Temperature Timing Main Effects Plot for Sidewall Volume Data Means 1086 11010510095 100500 11500 11000 10500 11500 11000 10500 11500 11000 10500 Mass flow rate Pressure Temperature Timing 5 17 29 40 rate flow Mass 6 8 10 Pressure 95 100 105 110 Temperature Interaction Plot for Sidewall Volume Data Means
- 10. 10 2. Simulation - Minitab and Main effect analysis 4029175 7000 6000 5000 1086 11010510095 7000 6000 5000 100500 Massflow rate Mean Pressure Temperature Timing Main Effects Plot for Shoulder Volume Data Means 1086 11010510095 100500 8000 6000 4000 8000 6000 4000 8000 6000 4000 Massflow rate Pressure Temperature Timing 5 17 29 40 rate flow Mass 6 8 10 Pressure 95 100 105 110 Temperature Interaction Plot for Shoulder Volume Data Means
- 11. 11 3. Molde RBF 0 20 40 60 80 100 120 140 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 Number of simulation normalizationdataofBaseVolume RBF BaseVolume model figure measured model prediction
- 12. 12 4. Control - Iterative Learning Control ILC can fast achieve perfect tracking in a repetitive mode process ILC have good performance in non-linear systems ILC is a mode-free control method which is low degree of dependence on model accuracy
- 13. 13 4. Control - Iterative Learning Control
- 14. 14 5. Plan for next step Build model by Gaussian process Combine fuzzy logic control with ILC
Editor's Notes
- #2: Hello everyone, I am pretty glad to show you my recently work. Cause I just research this area and methods for 2 month, this presentation will be a briefly introduction. I wish you would like.
- #3: I am going to begin by giving you some background to the blow moulding of polymer industry. I am then going to describe to you the software and result of simulation. After that, the identification and control method will be shown. And last is my plan for next step.
- #4: What is blow molding? It is pretty simple, in brief, it is a industry process for making bottles. In general, there are three main types of blow molding: extrusion blow molding, injection blow molding, and stretch blow molding. The blow molding process begins with melting plastic and forming it into a preform, like this….. In stretch blow molding, The preforms are then loaded and conveyed to reheat in an infrared oven. Finally, the preforms are simultaneously stretched with a stretch rod and blown up by high pressure air to produce to the desire shape. After a cooling period in the mold the bottles are ejected out.
- #5: In preform reheating process, temperature distribution is a problem, in order to guarantee the temperature distribution uniformity, the position and power of the lamps, rotating and transport speed need to be designed and controlled carefully. In order to avoid overheating, continuous forced flow of air is blown against the exterior surface of the preform.
- #6: The whole process of stretch blow moulding is just a few seconds. I will show you one of our experiment video by high speed camera. The process is free blow, it means blowing without mould. And in fact, this process is just one second.
- #7: Abaqus FEA[5][6] (formerly ABAQUS) is a suite of software applications for finite element analysis andcomputer-aided engineering. This software is used widely in blow moulding simulation. And it is pretty powerful, simulation results by abaqus is almost same with actual experiment. We design 144 group simulations Meanwhile, python is a programming language, Leo said it is the most easy programming language, just like English. Because Abaqus is developed base on Python, we use Python to write a few batch programming to write the input file of Abaqus, batch processing and read results from Abaqus.
- #8: This is one of our simulation video by Abaqus
- #9: We divide wall volume into three parts to analysis the thickness distribution. Mass flow rate is depend on air pressure and value of die. Pressure is blowing air pressure. Temperature is the preform temperature and we assume it is uniform. Timing is the time difference between the start time of stretch rod and blowing. For base Volume, mass flow rate is main factor whose line indicates a larger slope,. It means higher mass flow rate would yield a low base volume. Interaction plot means relationship of every two values.
- #11: Through the analysis, we should know mass flow rate is the main factor of ISBM process and each factor has different effect on different part volume. It will help us to build mode and design controller.
- #12: We used RBF to build the system modes. However we need more data to make the mode better. Because we are not sure if the input setting changed, whether the model can forecast or not. This is the mode of base volume.
- #13: Iterative learning control is a method of tracking control for systems that work in a repetitive mode. In each of these tasks the system is required to perform the same action over and over again with high precision. Repetition allows the system to improve tracking accuracy from repetition to repetition, in effect learning the required input needed to track the reference exactly. The learning process uses information from previous repetitions to improve the control sign ultimately enabling a suitable control action can be found iteratively. The internal model principle yields conditions under which perfect trackingcan be achieved. where L is the learning function. ILC can fast achieve perfect tracking in a repetitive mode process. ILC have good performance in non-linear systems. ILC is a mode-free control method which is low degree of dependence on model accuracy
- #14: where L is the learning function. Algorithm flow chart of ILC