Some time ago I was called to a textile factory to investigate some fairly new water-jet weaving looms it had purchased secondhand, which had an intermittent fault that was driving the in-house technicians to distraction.
Each loom had hundreds of electronic sensor-switches that detected warp threads breaking while weaving cloth. When a thread snapped, the loom stopped automatically and awaited a weaver to tie the thread and restart the machine. This prevented the machine from just carrying on and making huge amounts of scrap cloth.
The problem annoying the maintenance technicians was that in four cases out of 10, when a thread snapped, the weaving loom wouldn’t restart after the thread was retied. The machine had to be completely powered down to reboot the control box, which took a good 10 minutes as various capacitors had to be discharged. Threads snapped, on average, five times a day on each machine, and there were 54 identical looms. Ten minutes is an awfully long time in a production schedule. Multiply this by the number of breakages times, the number of looms, and the fail-to-restart factor, and you get an idea of how much downtime the factory was experiencing. The local management was seriously considering scrapping the machines because of the losses they were incurring.
My first problem was that the machinery came without maintenance documentation. The local team had not expected serious technical differences to their existing equipment as it looked superficially similar, and they accepted the machines after they saw them running.
As the obvious causes were the sensors-switches and the control box, these were the first places to look. I hooked up a logic analyzer to a machine, started it up, and cut a thread to simulate a break. The weaver retied the thread and started the machine. We had to do this a few times to generate a fault.
The logic analyzer showed the fault was in the switch, which sent a signal to the control box and then stuck in the open mode. I reset the machine and 10 minutes later tried again. This time the switch stuck in the closed mode.
A few tests demonstrated that the switch always sent out a pulse when it sensed a break but needed an external signal to reset properly. A little delving into the control box soon showed that this had been forgotten in the design -- there was no reset signal generated at all. My guess is this was the reason the looms were sold on the secondhand market in such good condition. An oddity was that the switch needed a trailing then a leading edge 20ms or more apart to reset.
Working overnight with the technicians, I installed a circuit that sent a logic low of 50ms to all the sensor-switches and delayed the restart signal by 100ms when the restart button was pressed on all the machines. At the same time I installed a capacitor discharge circuit into every control box, consisting of a 10k 2W resistor, an LED, and a momentary contact switch. This reduced the maintenance reset time to less than 30 seconds.
About a month later I called the factory to ask how things were going, and was told that the problem had gone away.
This entry was submitted by Ian Proffitt and edited by Lauren Muskett.
Tell us your experience in solving a knotty engineering problem. Send stories to Jennifer Campbell for Sherlock Ohms.
Click here to access the Sherlock archives.