statistics in manufacturing industry
of data descriptors reported for statistical process control. BevScan is a smaller, quieter device
from Australian food technology company Jeffress Engineering which uses the Camo Unscrambler in conjunction with laser spectroscopy to provide a black-box end- user device for analysing and classifying wines in the bottle. In reducing complex multivariate metric data to a simple categorisation, this has much in common with the dress size example although the analysis here is focused entirely into that reduction rather than in the ecosystem around it. In heavier industry, one route ahead
BevScan uses Camo Unscrambler software to run multivariate analysis on laser spectrographs to classify wine by quality after bottling
From Georgia Tech Research Institute (GTRI), amongst a range of other manufacturing programmes, the Intelligent Deboning System brings several physical technologies and image processing together with statistical modelling to (as the name suggests) automate the process of removing bones during poultry processing. Also developed at GTRI, and straying back towards quality control, is a production line system that automatically inspects the quality of sandwich buns as they leave a bakery oven. Adjustments are made automatically where needed and a range
In sickness or in health
In the last issue, environmental pollution was mentioned[3]
as an example of an externality: a
benefit or cost of production for which the price is not paid by the producer. Another example (often closely linked to pollution) is health – especially damage to it as a result of working in, or living near, a manufacturing centre. The cost, here, is typically paid to affected individuals by insurance companies, or by communal health budgets. Some such cases are open and shut. Little data analysis is required, for example, to link multiple mortality from pulmonary edema, emphysema and bronchial pneumonia with demonstrable exposure to a high level ethyl isocyanate leakage. Others, while less clear cut, are amenable to fairly simple, straightforward, unsophisticated methods. Many, however, such as apparent leukaemia clusters in proximity to nuclear facilities, are subject to genuine uncertainty which require much closer analytic attention.
20 SCIENTIFIC COMPUTING WORLD
Statistical examination of these apparent associations is working in highly sensitive territory. On the one hand, type II errors which wrongly reject linkage expose human beings to avoidable risk and may well also then cut them off from subsequent assistance or compensation. On the other, false positives damage both manufacture and those wider social aspects (perhaps also affecting health) dependent upon it. Intimidatingly large strategic manufacturing decisions of national or global economic significance can hang on finely balanced and contradictory data analytic outcomes. An example of the grey areas which surround such questions is the suspected linkage between pesticides or herbicides and incidence of Parkinson’s Disease (PD). There is a great deal of scientific literature around this suspected linkage, but also a lot of emotion: such chemicals are on the one hand central to many food and public health strategies, and on the other hand viewed
with deep public suspicion.
One of the most widely used weed killers is methyl viologen dichloride (MVD), better known by various trade names. It is extremely toxic and has strong chemical similarities to MPTP (a contaminant of certain recreational drugs) which is known to induce permanent Parkinsonism. In the words of one 2011 study[4]
of farm workers
supported by the US National Institutes of Health: ‘No pesticide has been definitively associated with PD in humans.’
The same study nevertheless goes on to conclude that there is an association between ‘mechanisms implicated experimentally... supporting a role for these mechanisms in PD pathophysiology’. Another, however, found[5]
for an automotive shiſt away from fossil fuels is the electric car, which in turn means development of high-energy- density batteries. A number of promising possibilities have been identified using nanomaterials, prominent amongst which is the use of lithium and silicon nanopowders in lithium ion cells. Scaling up any technology from lab bench prototype to stable full-scale manufacture is always a demanding task and nanopowders are certainly no exception. Nanotechnology company nGimat, one
of several centres around the world funded for the purpose, works in conjunction with Professor Jianjun Shi (once again of GTRI; and for other examples of their work see box: Keeping the wheels turning) who seeks to work from system data to process variability reduction models.
More generally, centralised storage and
management systems for collected data are as crucial to manufacturing as to laboratory science, possibly even more so. Quite apart from profit and loss considerations, we are in a time when the maximum efficiency in the use and reuse of resources is becoming ever more clearly a survival issue, and the synergies of aggregated data from which analyses can be run in increasing combination are central to improvement programmes and future planning as well as to operation. Data mining operations which used to be
the preserve of large corporations are now accessible from every desktop, and their reach grows exponentially with data set size. Electronic laboratory notebooks (ELNs) and laboratory information management systems (LIMS) have moved over the past decade and a half from innovative curiosity to mission critical core, and manufacturing has its own equivalents which have made
that
although workers engaged in MVD production had a higher exposure than control groups, there was no evidence of increased mortality from Parkinson’s Disease. The story continues.
www.scientific-computing.com
Jeffress Engineering
Camo Software
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48