Artificial intelligence saves lives in hazardous environments

Applying artificial intelligence to a gas compressor controller is necessary to improve safety, because there are ways that accidents can happen with gas, that blindly relying on Atex and similar regulations will not prevent. While developed for the only country in the world that runs much of its vehicle fleet on compressed natural gas, the lessons embodied in it are relevant to everyone who designs products for safety critical applications. The Artificially Intelligent Digital Gas Compressor Controller has been developed by Pakistan company Tesla Industries to service the supply of compressed natural gas (CNG) used to power the 1.2million vehicles that currently run on CNG in that country. Tesla is the largest supplier of CNG dispensers in Pakistan. Key to safety, in the opinion of ceo Aamir Hussain, is monitoring absolutely everything very closely, looking ‘for any abnormality’. For example, the controller senses the temperature of both oil and water in all four compression stages, and detects any current perturbations down to 2%. Hussain explained: “If even a slight current flicker starts to persist, this is likely to indicate some kind of drag force acting on the motor.” He explained there had been ‘some 20 accidents’ where conduits had helped gas to seep into what was supposed to be a safe area. This is often caused because gas pipe trenches should be ‘breathing trenches’, but are often sealed. If there is a gas leak, this can find its way into the gas compressor enclosure. It is therefore necessary to have gas detectors even in areas where there should in theory be no possibility of gas, and also to think of safety. Hussain gave an example of an accident which occurred when power failed – load shedding power cuts are a daily fact of life in most of Pakistan – and a contactor attached to a charged capacitor disengaged, causing a spark. With the new controller, detection of gas activates a backup power supply to keep the contactor locked up. The philosophy of the controller is fairly typical in that it logs any recorded events and sends messages about them. But it also suggests corrective measures. This does not follow the usual industry practice of discreetly indicating an error code, rather displaying prominently on a large LCD what the controlling computer has deduced and what the operator needs to do. Hussain suggests, as an example, that if a valve on the second stage compressor starts to leak, the small pressure increase will be detected and the system will suggest checking that valve. If more drastic measures are required, it will suggest these too, or at least explain in plain English why the machine has been shut down. Pointers * The controller monitors everything it can, including small changes and flicker in motor current, and the possible presence of gas where it should not theoretically be possible for it to be found. * As well as logging changes, the system uses artificial intelligence concepts to advise of remedial measures, and displays advice in plain English on a large LCD screen at eye level.