The present invention relates to the testing of packaging for air leaks. Many foodstuffs and drinks are sold in airtight containers such as bags or sealed trays made of plastics

(perhaps including aluminium foil covers) , aluminium or steel drinks cans, and the like (hereinafter referred to generically as "packages") . It is almost always desirable, and in many cases essential, that the packages be free from leaks due to holes, cracks, defective seals, and the like.

Current methods of detecting such leaks fall into two broad categories, namely visual inspection of the packs coming from the packaging system, and mechanical tests such as pressure testing, air flow measurements, and submersion in water. These methods however suffer from many disadvantages which include considerable human error, expense and consumption of time and, in the case of mechanical testing, the necessity of designing the test to be product-specific.

According to a first aspect of the present invention, there is provided a method of testing a package for leaks comprising the steps of: increasing the pressure within the package; detecting ultrasonic emissions from the package; and, identifying those ultrasonic emissions which are caused by gas leaking from the package.

According to a second aspect of the present invention, there is provided apparatus for testing a package for leaks, the apparatus comprising: means for increasing the pressure within a package; detection means for detecting ultrasonic emissions from the package; and, identifying means for identifying those ultrasonic emissions caused by gas leaking from the package.

An embodiment of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which:

Fig. 1 is a diagrammatic illustration of apparatus according to the present invention for testing packaging for air leaks;

Fig. 2 is a diagrammatic illustration of the application of the invention to a number of different types of packaging; Fig. 3 is a trace showing the output from a processor against frequency for a non-leaking pack;

Fig. 4 is a trace showing the output from the processor against frequency for a leaking pack;

Fig. 5 is a trace showing the processor output against time for a non-leaking pack; and,

Fig. 6 is a trace showing the processor output against time for a leaking pack.

A flexible package 10 is transported on a conveyor 12 from a packaging station (not shown) to a testing station shown generally in Figure 1. At the testing station are provided a pressure applicator 14 and a closely spaced acoustic transducer 16. The signal output by the transducer 16 is logged using a data logger such as a data acquisition card and subsequently processed in a computer or other processor 18.

The pressure applicator 14 shown in this example is a simple pneumatic cylinder 20 in which a piston 22 is moved down when compressed air is supplied to the cylinder 20 behind the piston 22. A piston rod 24 extends downwards from the piston 22 and carries at its lower end a transverse member 26, which in the example shown is a disc¬ like plate 26, which applies pressure to the package 10 as the piston moves downwards. A return means 28, for example a coil spring 28, may be provided to raise the piston 22 in the cylinder 20 when the compressed air is no longer applied to the cylinder 20.

The acoustic transducer 16 in the example shown is a phased array of piezoelectric transducers and generally can detect ultrasonic frequencies in the range 20-100 kHz. Signals from the transducer array are passed to the processor 18 where the signals are filtered and processed. In an illustrative example, the transducer array and processor equipment was an Ultraprobe 2000 manufactured by UE Systems Inc. of 12c West Main Street, Elmsford, New York 10523. In use, the conveyor 12 is stopped so that the flexible package 10 is positioned the beneath the cylinder 20. Compressed air is supplied to the cylinder 20 to move the piston 22, piston rod 24 and plate 26 downwards to apply pressure to and slightly compress the package 10. This compression increases the gas pressure within the package and, if there is any leak, gas is forced out. The escaping gas causes a noise to be emitted which has significant frequency components in the ultrasonic range, i.e. in excess of approximately 20 kHz. Thus, detection of additional ultrasonic vibrations when the package is compressed is indicative of a leak. Detection of components in the ultrasonic region is particularly preferred because low frequency ambient noises do not interfere with the measurement and it is relatively easy to filter out unwanted ultrasonic frequencies emitted by other sources.

The frequency range monitored can generally be in the range 20 to 100 kHz. In an illustrative example, the frequency range monitored is 38 to 42 kHz. The signal received from the transducer 16 may be heterodyned in the processor 18 so that an audible signal in the range 0 to 4 kHz is output. In this way, an operator will easily be able to judge when a package being tested is leaking. Alternatively or additionally, the processor 18 may provide an output which stops the conveyor 12 or otherwise causes the package 10 to be automatically rejected when it is judged by the processor 18 that the package 10 is leaking.

The precise nature and form of the pressure applicator may vary and it may be electrically, pneumatically or mechanically driven. The pressure applied, the form of the pressure member 26, and the direction in which the member 26 contacts the package, may be varied from that shown in order to suit the nature of the package being tested. For example, the more flexible the package, the less pressure need be applied. The member 26 should be shaped to apply the pressure to the package without damaging the container and should therefore match the basic shape of the surface of the package 10.

Fig. 2 illustrates the different directions in which pressure may be applied to containers of different shapes and sizes. Thus, in Figs. 2a and 2f, pressure is applied to the top of a tray pack 30 or box 31 respectively. In Figs. 2b and 2e, pressure is applied to opposite sides of a rectangular package 32 (e.g. of the type often used for milk or fruit juice) or cylindrical can 33 respectively. In Figs. 2c and 2d, the pressure is applied to one side or two opposing sides of a bag 34, the actual direction being dependent on the orientation of the bag 34 as it is presented to the testing station.

Clearly in those cases where the pressure is applied horizontally to the package, it is necessary to support the package on the other side; this support could take the form of a second cylinder 20 with associated piston 22, piston rod 24 and member 26 or may be provided by the conveyor 12 as shown in Fig. 1.

As a further alternative, instead of using the assembly of the piston 22 in the cylinder 20, a second conveyor belt positioned above the package may be used. Rollers may act on the second belt to compress the package between the two belts. This alternative system would allow a higher throughput of packages, allowing perhaps 100 to 150 packages to be tested per minute.

It should be noted that ultrasonic emissions are generally only caused by gas escaping through a relatively

small leak. Thus, if a large hole is present in a package 10, the detector 16 and processor 18 may not detect that the package is in fact faulty; in a fully automatic system, such a package may be passed as satisfactory. In order to check for large leaks in a package 10, a pressure- or force-sensitive resistor 29 may be fixed to a plate 27 fixed beneath the lower belt 12 (as shown in Figure 1) or to the lower surface of the pressure member 26 where a piston 22 as shown in Figure 1 is used. As the pressure member 26 comes into contact with a package 10

(which will generally be after a known amount of downwards travel of the piston 22) , the pressure- or force-sensitive resistor 29 will detect the resistance to downwards movement of the piston 22 caused by a package 10 which has no leaks or only small leaks. If the detector 29 indicates that no or very little resistance to downwards movement of the piston 22 occurs as it contacts the package 10, it can be assumed that the package 10 is likely to have a large leak, through which the flow rate of escaping gas is large, and that such a package should be rejected.

Figures 3 and 4 show the output of the processor 18 against frequency for a non-leaking pack and for a leaking pack respectively. In each case, there is a large "spike" A at a relatively high frequency which is due to background noise from the equipment used (such as the piston 22) and other instruments, for example. Two smaller peaks B can be seen in the trace for a non-leaking pack, again due to background noise. Otherwise, the output of the processor 18 for a non-leaking pack is substantially zero. For a leaking pack, the trace shown in Figure 4 shows the burst of signals across a range of frequencies at the lower part of the frequency scale shown at C which corresponds to gas leaking through a relatively small hole in the packaging. There may well be peaks corresponding to the peaks B in the trace shown in Figure 3, but these are hidden by the dominant output caused by the leaking gas.

Figures 5 and 6 show the output of the processor 18 against time for a non-leaking pack and for a leaking pack respectively. In each case, there is an initial signal D of significant amplitude caused by the movement of the piston 22 when compressing the package. For the non- leaking pack, the output thereafter is substantially flat as shown in Figure 5.

For the leaking pack, as shown in Figure 6, the output increases with time as gas escapes from the package as can be seen at E in Figure 6. Eventually (beyond the trace shown in Figure 6) , the output will decay again as the rate of escape of gas drops.

It may be advantageous to connect the output of the transducer 16 or the processor/data logger 18 to a neural network. In addition, it may be advantageous to connect the output of the pressure- or force-sensitive resistor 29 to such a neural network in the example where such a resistor is used. A neural network used in this manner would be able to distinguish between the characteristic signatures of a leaking pack and a non-leaking pack more quickly and accurately, particularly in a batch process where substantially identical packages are being tested, allowing higher throughput and/or more reliable testing.

In some cases, such as cans of drink, the pressure within the can increases naturally after the can is sealed so that it is not always necessary to apply any external influence to the can to increase the pressure.