A vision of the future

It is an unfortunate fact that manual inspection of manufactured parts is never 100% reliable. Often this is because production rates are very high and it is just impossible to inspect everything. An inspector's job is also monotonous, with the obvious effects of faulty product being missed by people leaving their stations to go on breaks. But in applications such as biscuit making, for instance, there is nothing better than the combination of the human eye and brain to determine, in a fraction of a second, whether a biscuit is faulty. The key to 100% quality inspection, therefore, is to marry this human quality to the reliability of a machine.

A biscuit producer in the North East of England recognised this and asked machine vision manufacturer Matsushita Electric Works to help with its quality control problems. Overcooked and undercooked biscuits are not to everyone's taste, but equally undesirable from a production point of view are broken or misshapen biscuits. These can foul the packaging machinery causing downtime and backlogs. So, based on stringent trials at both MEW's offices and on site at the biscuit company, a vision system was proven to meet the inspection requirements.

In its simplest form a vision system can act as a policeman at the end of a production line rejecting bad product. However, a more efficient way of using a vision system is to feed back information about the output to statistical packages. This allows the manufacturing process to be analysed and information on control variables such as conveyor speed or oven temperature to be fed back directly to the machine. Using a machine vision system in this way provides significant savings in time and money by ensuring the process is running as efficiently as possible.

The biscuit manufacturer's application comprises two AX40 colour systems, each with two cameras mounted over the conveyor system immediately after the ovens. So that the cameras are not affected by the temperature of the ovens, they are mounted at a height of 2m, with lenses chosen to fill each camera's view with nine biscuits in a 3x3 array - allowing a total of 36 biscuits to be inspected at one time.

Localised white LED light was used to provide a consistent environment for inspecting the biscuits without having to worry about making adjustments for changes in ambient light levels occurring during the day. A signal from an encoder is processed by a PLC and outputted to the vision system as a trigger to capture the images of the biscuits. Once this has been done, the images are written to memory and processed using the vision software against the manufacturer's criteria.

The image processing consists of an overall check to determine the centre point of each biscuit - a necessary step in the inspection process as the positioning of the biscuits on the conveyor is not precise and subsequent checkers need to be positioned accurately relative to the centre of each biscuit.

<b>Precise quality judgement</b>

Results from the dimensional checkers include biscuit diameter in the X and Y axes, perimeter, area, as well as a count of the number of objects for each biscuit - a count of more than one indicating a broken biscuit. These checks give a precise judgement regarding conformity to size and completeness of the biscuit when compared to preset parameters that represent 'good product'.

Further checking of the biscuits' colour is carried out by individual colour checkers that again compare actual inspection results with preset allowable levels. In all, the complete inspection takes a fraction of a second, leaving the vision system plenty of time before the next trigger to inspect the next 36 biscuits.Having a system such as this identify and reject faulty biscuits is obviously a great way of improving the quality of product sent to the customer. But using the data from inspections to improve the production process is a primary goal of most vision system users.

Simple statistics, such as the number of biscuits produced versus the number of misshapen or broken biscuits, provide a valuable overview of production efficiency. Furthermore, an increasing trend in faulty product may flag up the need for maintenance, or act as an early warning system for unpredictable faults that could cause significant downtime if they were allowed to occur.

At the North Eastern manufacturer, apart from rejecting over- or under-cooked biscuits, the checkers used for analysing the colour of each biscuit are also used in a much more desirable and direct way. Each biscuit's colour is compared to values in a look-up table in the controlling PLC and, as a result, the conveyor speed and oven temperature are varied. This leads to a more efficient production process that optimises energy usage and biscuit production.

Given that the loss of an own-brand supermarket contract would be catastrophic, the biscuit manufacturer considers the relatively low cost of machine vision to be money well spent. Not only does the customer appreciate the investment made in the biscuit maker's manufacturing process to supply current contracts, it is now also considering awarding new contracts.

Ian Purcell is with Matsushita Electric Works.