13 October 2004
Cash with drawing
'Sketch and Search' creator illustrates potential money savings
A SIMPLE SKETCH will make 'endless savings' for manufacturing companies wasting time and money on redesigning new components, said its creator.
Aston Business School's Dr Doug Love is the inventor of CADFind, a unique software package that allows engineers to sketch and search for existing parts.
'There is no need for people to retrain,' said Love, 'this package utilises their existing skill to draw the part they need. If the part is already available then CADFind will locate it.'
Love's claims for CADFind's potential are based on calculations made by the US Department of Defense's costings for redesigning new parts: $20,000 (£11,173)* for avoiding a new design and $33,000 (£18,436) where new manufacturing tooling is needed.
'One academic study found that 20 per cent of existing parts could be re-used,' said Love, 'another 18 per cent could be used with some modification.
'Using the lower 20 per cent figure as an example, a company that creates 5000 new parts a year is creating 1000 of them unnecessarily.
'Multiplying 1000 parts by the US Defense Department's lower costing for creating them ($20k or £11.173k) means the company can make a saving of £11.173m a year. It does not take a genius to work out that potential savings relative to the output of any company are huge.'
Recycling existing designs for parts also avoids the incidental costs involved in introducing new ones, including data management, testing, release management, production engineering, supplier selection and purchasing, tooling and planning, inventory and warehousing and after-sales support.
'Recycling also generates economies of scale,' Love continued. 'A study by Scania suggests that manufacturing costs fall by ten per cent with a doubling of batch size.
'The study also shows that distribution and development costs are proportional to the size of a company's parts range.
'So what originally appears to be a small decision to design a new part has potentially massive cost implications for the business,' said Love, a former engineer.
- ENDS -
For further information please call 0121 204 4549 or email: firstname.lastname@example.org
Notes for editors:
* Based on $1 = �0.558653 sterling
With the help of Aston University's business partnership unit, Dr Love has set up a company for the commercialisation of CADFind. For information visit the website: www.sketchandsearch.com
How CADFind Sketch & Search works
As Google searches for text similarities, CADFind does a similar job with images. From the user's sketch, CADFind automatically extracts part properties and builds a code representing that part. This code is compared to previously coded drawings stored in the database or 'catalogue' and as it scans the database it calculates the similarity of each part to the target and then displays the results in similarity order.
The coding and matching process is completely hidden from the user who simply sketches the part and clicks the search button.
To create the catalogue, existing drawings are coded in a similar way except there is additional processing of a) removal of non-geometrical data (drawing frame, dimensions, notes, etc) and detection of each projected view.
Part coding systems are a proven technology (often called group technology codes) and have been around for many years but they always needed a skilled engineer to interpret the drawing and produce the code. Manual coding rates rarely exceed 100 parts/day/engineer and that means it would take a skilled engineer over a year to code a modest database of 30,000 parts.
CADfind automates this process and, by coding over 1000 parts an hour, could process the same number of drawings in just a couple of days.
Find parts from a sketch
Existing retrieval systems, like those used by product data management packages, work by using text descriptions whose fundamental flaw is that one word (eg, 'bracket') can apply to thousands of parts.
There are also cases where descriptions are different but the parts are the same, so 'bush', 'spacer' and 'retainer' may all refer to similar parts. CADFind retrieves images rather than words. Existing drawings can also be used as the target.
CADFind's retrieval ability is excellent - hit rates of 100 per cent can be achieved with detailed drawings and over 95 per cent with sketches. In this case a 'sketch' is produced by the engineer in the same way as proper drawings but much more quickly as only the simplified, essential geometry of the part is required. Naturally the quality of the sketch and its level of detail affect the retrieval rate.
If a rough sketch is used, retrieval rates can be improved by selecting the best item of those returned as the basis of a second search. In this way very high retrieval rates can be achieved from very simple sketches.
A simple payback calculation shows that a period of less than one month is typical even if very pessimistic figures are used. For example, a company that has a parts range of 50,000 parts and uses CADFind to avoid creating 400 new parts per year, the payback period would be 2.6 weeks including all the costs of purchase, training and installation of the system. This calculation assumes a saving of only �500 for each part avoided instead of the $33,000 (�18,436) estimated by the US Department of Defense. Clearly the total savings could be very much larger.
This payback calculation ignores any on-going cost savings that could be expected from more economical production or distribution and are probably a significant underestimate of the real figures that a company could expect to realize.
Recycling parts is an important way of saving very substantial costs and is particularly important for companies that generate a large number of non-standard or specialist parts and assemblies.
CADFind is a unique system for finding parts easily and quickly. It works the way engineers do and makes it quicker for a designer to find an existing part than to design a new one.
The system is the product of many years of research at Aston University which ensures that excellent performance can be combined with very low implementation costs.