Spring 2015 (Intern)

I tested the integrated system(with the DC and the FT together) on a second order emulated process and the results were fine.
I solved the spikes problem totally. Now, no spikes appear in the system response.

I solved the spikes problem. They still appear but this is much less frequent than before.
I tested the integrated system ( with the DC and the FT together) on a first order emulated process (instead of the unity gain process I used before). The results were fine and as expected.

I integrated the Fault Tolerance and the Delay Compensator. I tested the system after the integration on the cloud; the primary controller was at a VM in Singapore, the secondary controller was at a VM in Sao Paulo, Brazil, and the emulated plant was on my laptop at SFU. The results were fine for delay value up to 900 ms. The remaining issue now is that there are some occasional small spikes that appear in the system response and I am now working to figure out how to eliminate them.

I fixed the last week's issue. Also, I tested the whole system on the cloud with its two modes; the fault tolerance and the delay compensator modes. The results were fine when I tested the system on the cloud on near locations (e.g. at Oregon and California). However, for the further places (e.g. US east coast and Tokyo) and for large delays (i.e. >= 30 sampling periods), the system response in the delay compensator mode oscillates and it does not settle. I talked with Dr. Tamir about this and he suggested some directions to fix this issue and this is what I will investigate in my next steps isA.

I worked on the integration between the delay compensator and the rest of the system. The goal was to have one system with two separate modes; the Fault tolerance mode and the the Delay Compensator mode. I tested the delay compensator mode using our machines (through the internet) and the results were fine. However, when I tested the code using Amazon cloud services, the results were not the same! For example, when disabling the delay compensator and the estimated delay was around 25 sampling periods, the system response, when tested on the cloud, was almost smooth without any overshoots. So, right now I am working to figure out how to solve this.

I solved the remaining issue in my last implementation. I have also added a delay estimator to the code that can deal with variable delays. I tested this with a unity gain process emulated in LabView.
Worked with Khaled and Maram on elaborating the system design.

I implemented the delay compensator assuming a unity gain process (to avoid any inaccuracies in the process model block in the delay compensator). My next steps are; solving one main remaining issue with the current implementation, then implementing an internet delay estimator(I am now using a predefined delay in LabView) and testing my delay compensator with a more complex process (that I can find a good model for).

We made a list of the physical systems that can suit our work. We are also in contact with the manufacturers of these systems to know about their prices(the prices are not available online) to find a system that fits our estimated budget.
We made a final list of 3 open source SCADA/HMI software that can be used for our work. We will discuss with Dr. Tamir to find the most suitable one.
We started working on the implementation of the PID controller with the bumpless transfer. After finishing this part, we should start implementing the fault tolerance algorithm.

We had a brief meeting with Dr. Tamir and showed him the results of the delay compensator and the fault tolerance implementations. They were fine. The only drawback is the ripples that appears in the response of the system with the delay compensator for the high delay values.
We are now working on finding a new physical system for our work. We discussed the criteria of choosing such system with Dr. Tamir. Besides, we also search for an open source SCADA/HMI for our system.

We arrived in Canada on 3rd, March and it took us about a week to settle in, thanks God.
We started our work full load on 11th, March. We had two remaining things to do; first, to implement the smooth handover part of the fault tolerance algorithm. We have done it. Second, to fine tune the settling time of the system with the delay compensator. We have done as much as we can in this, but still the settling time is longer than that in the transaction paper.

Private:progress-magdy