Plant Disease Prediction System Using Deep Learning and Image Analysis

Abstract

failure modes are worth naming, because they are mundane and therefore easy to overlook. Goods receipt notes are raised but not always dated accurately, particularly where material arrives outside working hours. Partial deliveries are recorded inconsistently, and a truck carrying forty per cent of an ordered quantity may be logged as a delivery, a partial delivery or nothing at all depending on who is at the gate. Supplier master data accumulates duplicates as the same vendor is entered under slightly different names by different projects, which quietly destroys any attempt to aggregate performance. None of these is difficult to fix. All of them require sustained attention over months, which is a different and scarcer commodity than technical capability. The second risk is that measurement changes behaviour in ways the measurement did not anticipate. This follows directly from the adaptive systems argument in Section 2. A supplier evaluation scheme that affects the award of work gives suppliers two ways to improve their scores, and quoting longer lead times is considerably easier than delivering faster. If the firm then plans against the padded quotations, the scorecard improves while the programme deteriorates. Guarding against this requires monitoring quoted lead times over time as a signal in their own right, and treating a supplier whose quotations are lengthening as a finding rather than as a supplier who has become more honest.