The Inspector Calls

Inspection is part of the delivery of quality in all manufactured products. John Anderson, Managing Director of Diverse, looks at the role of inspection in the glass industry from its practicalities to its commercial basis.

 Introduction

It may seem trite, but it is worthwhile defining inspection: it is a methodology aimed at measuring the output of the production process to determine either a specific metric, or measurement of a parameter. The parameter could be used as a surrogate to determine whether the object is generally similar to the ideal sample product.

Any measurement system must interfere with the object being tested. This interference may be due to the mechanical handling system, the feed system through the inspection station or direct transducer interface. Whatever the interference, it is essential that any inspection process does not itself make defects in the product. This issue is particularly important in the glass industry where mechanical handling is particularly undesirable.

What can be inspected and where ?

Almost anything can be inspected but the detail and finesse increases the time required to complete the inspection process. Thus a simple metric such as the height of a bottle can be found with an accuracy of say 200um with an on-line system, but to measure the height to 1um would require off-line measurement. Further, the definition of the measurement would be different in the 2 scenarios - for the on-line system the a measurement of average height with a probe system of 2mm x 200um might be appropriate, whereas the 1um measurement would only have meaning if taken at a specific point or points.

Where is the optimal place for measurement to take place?

To answer this type of question a management tool known as multi-attribute utility theory is required. The technique is a combination of statistical methods, costs matrices and penalty functions. The format of the mathematics depends very much on the precise process but a generalised scheme for glass object manufacture is shown in the box. This methodology can identify the best points in the manufacturing system for application of inspection investment, as well as giving a rational calculation basis for the supply of commercial ware.

Methods

There are many inspection methods from visible light to infared, from ultrasonics to electrostatics, but what ever the method several key issues must be considered:

1. Accuracy of measurement - whether this is a metric or a loose parameter.

2. Probe size - this is the ultimate resolution of the measurement system, for example in an optical system this is set by the MTF (modulation transfer function) of the system.

3. Orientation - Faults can occur with a preferential orientation within the body of the material. However to see the whole item, there is a clear trade between gripping the ware and presenting it in all orientations to the inspection system, or illuminating and imaging in a variety of different directions and not contacting the ware at all.

Limitations

There are a number of limitations which are common to the inspection of all products - size of defect and the speed at which the inspection takes place. The glass industry has some special limitations. These include extreme temperatures, transparency and the potential for product damage.

The transparency at visible wavelengths, together with the possibility of rogue light coupling make it difficult to accomplish both good illumination and good imaging. There are several ways of solving the illumination problem; these include:

1. Coupling light into the glass material and relying on light being reflected or refracted out.

2. Self illuminance at the hot end of the process.

3. Uniform illumination in transmission e.g. for side wall inspection.

4. Collimated (parallel light) in the object space to produce on-line shadow image.

Advantages

The advantages of inspection extend beyond simply reducing the unit cost. By correlating the detection of flaws with raw materials, press start up and annealing time etc. the opportunity exists for tuning the complete manufacturing process.

The cost per item of supplying faulty product can be a powerful penalty function. It is not a case of not being paid for a faulty container, it can be the return of a pallet load, a container load, even a ship load, and this could have the high added cost of the food or drink etc. with which it has been filled. Not only that, there may be consequential damages - such as passed on liability if a member of the public is injured by a shard of glass.

Indeed, even if the manufacturer picked up all of these costs there may be still further costs such as lost business with the customer of the flawed product, unfavourable publicity causing current or potential customers to retreat to alternative packaging. In this context, inspection is not an expensive luxury, but an essential part of product delivery. This consideration suggests that the old concept of commercial ware is not really practicable today and that the industry must strive for 100% fault free product.

This paper has shown that it is only by analysing the complete production process, that a proper justification for inspection can be made. Alternatively, it can be argued that inspection is an absolute requirement in the glass industry and that it is only by detailed mathematical analysis that the optimal application of funds to inspection can be determined.

John Anderson is Director of Diverse Technologies and Systems in Cambridge England. Diverse are active members of the UK Industrial Vision Association and advise clients on metrology and quality systems. The company has extensive experience in developing inspection systems for glass containers, glass plate and injection moulded products etc.

Inspection Rational

Developing an inspection rational can be done based on applying multi-attribute utility theory to glass container manufacture.

Process attributes

C₁= Cost of raw material delivered to the glass press. P₁= Probability of bad material
C₂= Cost of pressing. P₂= Probability of failed pressing.
C₃= Cost of annealing. P₃= Probability of failed annealing.
C₄= Cost of palleting. P₄= Probability of handing damage.

Inspection costs

 Cost of hot end inspection = I₂ Per item cost including amortisation.
Q₂= probability that raw material and pressing faults are found.
Cost of cold end inspection = I₃
Q₃= probability that annealing and any missed raw material and pressing faults are found.

Penalty cost

 C₅= cost per item of supplying faulty product:

 Simply calculated final cost (the perfect system cost) :

Final cost taking account of probability of flaws at each stage (the uninspected cost):

 Final cost, C_tot taking account of probability of flaws with hot and cold end inspection:

 The business aim is to minimise C_tot and this is done by minimising the optimisation penalty function:

 These expressions can be extended to take account of the further inspection stations, specific faults and time related quality for example process start up. Once all the data is collected, then the attributes can be combined to give a net cost and this can be optimised against a combination of costs and probabilities. An important observation is that return on investment for inspection is faster the earlier it occurs in the manufacturing process.

Clearly even in this simplistic model the expressions are building up in complexity and this can lead to difficulty with the optimisation. Computer based matrix methods can be used to optimise the data and deliver reaction curves for the sensitivity of flaw probability and inspection cost.