I have recently come up with a possible explanation of what may have caused my fire. It started with a discussion on the Thundersky group about bms systems in general and I chimed in so the discussion worked around to my fire. One possibility discussed was that the fire started from my slave board. I initially dismissed this because the load resistor only dissipates 1 watt of power, and even if it shorted on, that is nowhere near enough to cause a fire. But then I started looking at some other cell failures that have been reported and a cell failing with a high internal resistance is fairly common. Peter had one of his cells fail this way and Hipower sent him a replacement under warranty. Failures like this have also been reported with Thundersky and I am sure others as well.
So what effect does a high internal resistance have on a cell under charge? When you connect a charger to a battery the voltage seen across each cell is based on its resistance. If all the cells are of the same resistance then the voltage is equal for all cells. If one cells resistance is higher then it will have more voltage across it, if it goes open circuit then it has full charger voltage across it!
So lets assume I had a high resistance at one cell, and this could also be caused by a poor connection at the cell so not necessarily a internal cell fault. My setup had the slaves switching off the main power supply to the charger and then switching it back on after the cells voltage had dropped a bit. The idea is that the load resistors on the high cells would bleed off some power and let the rest of the cells catch up. If we look at what might of happened with this setup and a faulty cell, the charger switches on and the voltage across the cell starts to climb. The load resistor gets switched on, and as the voltage climbs further the charger gets switched off, problem is when you "pull the plug" on many electronic devices the internal capacitors keep them going briefly. Add to this the fact that the charger when it was switched on had to figure out the state of charge of the pack. The charge put out a brief 10amp pulse when first switched on. These 2 factors could have caused a voltage spike on the faulty cell that burned out the picaxe controller.
Now we have lost bms function on our faulty cell and possible the load resistor has shorted on. Now if we use ohms law we find that 7 volts across a 15 ohm resistor is around 3 watts of power. Any voltage above that and the resistor is overloaded and becomes a fire hazard. From past experience I have seen a front bumper that caught fire on a car, the cause was traced to a wiring fault that connected the exterior temp sensor straight to the 12 volt battery. Since a temp sensor is just a resistor, it clearly showed what happens if you apply too much voltage to a resistor.
So that is one possible cause of my fire, and one that I think is quite probable. Once the fire got started the cells probably "fueled" it and made it worse.
Hi-power have made good on their promise to supply me a pack at a good price, so I have ordered 140x 40ah cells for my newest project, a Matiz conversion which I will detail in the conversions section. Regarding my pack and what lessons I have learned from the fire they are as follows;
Slaves will be mounted away from cells in separate enclosure with 1 amp fuses on each cell connection.
I am going to use Peters full digital bms, and am working on slave pcb layout.
No scrimping on a charger, a Zivan with the proper charge profile.
Temp sensors for each battery box.
Cells will be clamped as per Thunderskys recommendation. Hipower says this isn't required as long as cells are operated within their specs, but I think this could help if there is a problem.
A suggestion of Cedric's, use a timer at the mains and set it to switch off when the pack should be charged. Sometimes the simple ideas are the best!