The present invention relates to a method for operating a metal detection system that uses at least two operating frequencies and to a metal detection system that implements this method.

An industrial metal detection system is used to detect and reject unwanted metal contamination. When properly installed and operated, it will help reducing metal contamination and improving food safety. Most modern metal detectors utilise a search head comprising a "balanced coil system". Detectors of this design are capable of detecting all metal contaminant types including ferrous, nonferrous and stainless steels in a large variety of products such as fresh and frozen products.

A metal detection system that operates according to the "balanced coil"-principle typically comprises three coils that are wound onto a non metallic frame, each exactly parallel with the other. The transmitter coil located in the center is energised with a high frequency electric current that generates a magnetic field. The two coils on each side of the transmitter coil act as receiver coils. Since the two receiver coils are identical and installed with the same distance from the transmitter coil, an identical voltage is induced in each of them. In order to receive an output signal that is zero when the system is in balance, the first receiver coil is connected in series with the second receiver coil having an inversed sense of winding. Hence the voltages induced in the receiver coils, that are of identical amplitude and inverse polarity are cancelling out one another in the event that the system, in the absence of metal contamination, is in balance. As a particle of metal passes through the coil arrangement, the high frequency field is disturbed first near one receiver coil and then near the other receiver coil. While the particle of metal is conveyed through the receiver coils the voltage induced in each receiver coil is changed (by nano-volts) . This change in balance results in a signal at the output of the receiver coils that can be processed, amplified and subsequently be used to detect the presence of the metal contamination . The signal processing channels split the received signal into two separate components that are 90° apart from one another. The resultant vector has a magnitude and a phase angle, which is typical for the products and the contaminants that are conveyed through the coils. In order to identify a metal contaminant, "product effects" need to be removed or reduced. If the phase of the product is known then the corresponding signal vector can be reduced. Eliminating unwanted signals from the signal spectrum thus leads to higher sensitivity for signals originating from contaminants. Methods applied for eliminating unwanted signals from the signal spectrum therefore exploit the fact that the contaminants, the product and other disturbances have different influences on the magnetic field so that the resulting signals differ in phase. The signals caused by various metals or products, as they pass through the coils of the metal detection system, can be split into two components, namely resistive and reactive components, according to conductivity and magnetic permeability of the measured object. The signal caused by ferrite is primarily reactive, while the signal from stainless steel is primarily resistive. Products, which are conductive typically cause signals with a strong resistive component.

Distinguishing between the phases of the signal components of different origin by means of a phase detector allows obtaining information about the product and the contaminants. A phase detector, e.g. a frequency mixer or analogue multiplier circuit, generates a voltage signal which represents the difference in phase between the signal input, such as the signal from the receiver coils, and a reference signal provided by the transmitter unit to the receiver unit. Hence, by selecting the phase of the reference signal to coincide with the phase of the product signal component, a phase difference and a corresponding product signal is obtained at the output of the phase detector that is zero. In the event that the phase of the signal components that originate from the contaminants differ from the phase of the product signal component, then the signal components of the contaminants can be detected. However in the event that the phase of the signal components of the contaminants is close to the phase of the product signal component, then the detection of contaminants fails, since the signal components of the contaminants are suppressed together with the product signal component.

In known systems the transmitter frequency is therefore selectable in such a way that the phase of the signal components of the metal contaminants will be out of phase with the product signal component.

GB2423366A discloses an apparatus that is arranged to switch between at least two different operating frequencies such that any metal particle in a product will be subject to scanning at different frequencies. The frequency of operation is rapidly changed so that any metal particle passing through on a conveyor belt will be scanned at two or more different frequencies. In the event that for a first operating frequency the signal component caused by a metal particle is close to the phase of the signal component of the product and thus is masked, then it is assumed that for a second frequency, the phase of the signal component caused by the metal particle will differ from the phase of the signal component of the product so that this signal components can be distinguished. By switching between many frequencies, it is expected that one frequency will provide a suitable sensitivity for any particular metal type, size and orientation.

Looking at this method from a different angle it can be stated that for one optimal frequency setting numerous other frequency settings have been applied, disclosing that this method requires considerable efforts. Various frequency settings need to be applied when measuring a single product. This means that for the frequency setting, that provides the best result, only a small measurement period is available. Consequently the result of the measurement will not be optimal. Furthermore, since the measurement is performed for all selected frequency settings the major part of the data, which is processed with considerable efforts, will be disregarded. Hence, this method, which requires considerable efforts in the signal processing stages, is characterised by a relatively low efficiency. The present invention is therefore based on the object of providing an improved method for operating a metal detection system that uses at least two operating frequencies as well as on the object of providing a metal detection system operating according to this method. Particularly, the present invention is based on the object of providing a method that allows detecting contaminants, particularly metal contaminants, with reduced efforts and a high efficiency.

Further, the present invention is based on the object of providing a method that allows detecting small sized metal contaminants with higher sensitivity.

Still further, the present invention is based on the object of providing a method that provides information about the capability of the metal detection system that can advantageously be used for the automatic configuration of the system.

SUMMARY OF THE INVENTION

The above and other objects of the present invention are achieved by an improved method for operating of a metal detection system as defined in claim 1 and a metal detection system operating according to this method as defined in claim 16.

The inventive method serves for advantageously operating a metal detection system that comprises a balanced coil system having a transmitter coil that is connected to a transmitter unit, which provides transmitter signals with a selectable transmitter frequency, and with a first and a second receiver coil that provide output signals to a receiver unit, which compensate one another in the event that the metal detection system is in balance and, in the event that product is present in the balanced coil system, provide an output signal that is forwarded to a signal processing unit, which suppresses at least the components of the product signal and delivers the signal components caused by metal contaminant contained in the product . The inventive method comprises the steps of determining the phase and magnitude of the related signals at least for a first metal contaminant for at least two transmitter frequencies and for at least two particle sizes of the first metal contaminant; determining the phase and magnitude of the related signal for a specific product for the at least two transmitter frequencies; comparing the information established for the at least first metal contaminant and the information established for the product; determining at least one preferable transmitter frequency with which the signal components of smallest sized particles of the at least first contaminant differ most in phase and amplitude from the phase and amplitude of the product signal; and selecting the preferable transmitter frequency for measuring the specific product. The inventive method therefore allows obtaining optimal transmitter frequencies with which the smallest possible particles of one or more metal contaminant types can be detected. Accordingly the inventive metal detection system will optimally be configured for any measurement, involving products of any consistency and any potential metal contaminant type.

Measuring a product at unsuitable transmitter frequencies and analysing the related data is avoided. The inventive method always applies the optimal frequencies so that measurements are performed with reduced efforts and high efficiency. Since measurements are not performed at unsuitable transmitter frequencies, the time available for measuring a product, i.e. for detecting metal contaminants in a product, is dedicated to the application of one optimal transmitter frequencies. As a result, more measurement data of high-quality are available for an individual metal contaminant. This leads consequently to a significant improvement of the sensitivity of the metal detection system for all products and metal contaminant types measured. Optimal transmitter frequencies are therefore determined for all metal contaminant types that may occur in a product and for all available products for all transmitter frequencies that can be selected. In a preferred embodiment at least two curves of a first array at least for a first metal contaminant are established. Each curve is established for a separate transmitter frequency representing the phase and magnitude of the signal for a progressively increasing particle size of the first metal contaminant. Hence, a curve or response locus is established at least for the first metal contaminant for at least two separate transmitter frequencies that are used as fixed parameters and with the particle size as a variable parameter. Each curve established for a specific transmitter frequency is part of a first array that relates to the first metal contaminant. For each metal contaminant a first array with at least two curves is established.

The information established at least for the first metal contaminant and the information established for the product for at least a first and second transmitter frequency are then compared in order to determine the preferred transmitter frequency, for which the signal components of smallest sized contaminant particles differ most in phase and amplitude from the phase and amplitude of the product signal. In the event that information has been gathered for each transmitter frequency for more than one metal contaminant, then the complete information established for all metal contaminant types for a first and second transmitter frequency is compared with the information of the product established for this first and second transmitter frequency. For this purpose, for each transmitter frequency a second array is built with curves of different metal contaminant types recorded with the same transmitter frequency. Then for a specific transmitter frequency a superposition of a second array and the product information is arranged, which allows determining, which parts of the curves lie within or outside of the area or range of the product signals. Parts of the curves that lie outside the range of the product signals indicate particle sizes of the metal contaminant, which will not be masked or suppressed together with the product signals.

Hence, the inventive method and the metal detection system not only allow determining the optimal transmitter frequency of a metal contaminant but also allow determining the minimum particle sizes of the metal contaminant types that can be detected. This valuable information can be used for configuring the metal detection system most efficiently.

The operator can input the metal contaminant type that shall be detected. Based on this information provided by the operator, the computer program implemented in the metal detection system will often find a transmitter frequency that will be suitable for the detection of two or more metal contaminant types. During the measurement process the metal detection system may therefore be configured and operated with one of at least two transmitter frequencies that preferably meets all requirements set by the operator.

In the event that a single transmitter frequency does not satisfy the requirements of the operator, then the computer program will select two or more transmitter frequencies that are optimal for individual metal contaminant types and that are applied during the measurement of the product. The selected frequencies are then applied according to a suitable method. The selected transmitter frequencies can be applied alternately or simultaneously, e.g. as a mixture of the selected transmitter frequencies, which are filtered accordingly in the receiver stage. Hence, only optimal transmitter frequencies are applied that allow measurement of metal contaminants for the maximum available time so that contaminants can be detected with the highest possible sensitivity.

In a preferred embodiment the operating program is designed in such a way that the operator can input the minimum particle sizes for the metal contaminant types that shall be detected. This allows the operating program to select a transmitter frequency that is suitable for two or more contaminant types, for which the specified particle size can be detected. The required information of the metal contaminant types and the product can be gathered in various ways. Information can be pre-stored and is downloaded. Alternatively, a calibration process can be performed for the product and the metal contaminant types, in which e.g. a product and metal contaminants of at least one particle size are measured.

The product information can be obtained when scanning a product, for which typically various signal components occur that have an individual phase and magnitude. Connecting the vectors of all signal components leads to an envelope that is the boundary of the area of the product signals or the product signature that will be suppressed by a signal discriminator, typically a signal processor that is programmed to suppress the components of the product signal. The area, in which signals of the product and contaminants are suppressed, is closely adapted to the product signature but typically slightly larger, so that a safety margin is provided. The product signature changes from transmitter frequency to transmitter frequency. For a general product, an algorithm, based on empirical data allows to establish the product information preferably based on only one measurement performed at a single transmitter frequency.

During a setup period in the factory or before the start of a measurement process data of metal contaminant types are gathered by measuring metal particles with at least one particle size. Preferably, only one or a few points of the curve are measured, while the remaining part of the curve is obtained by applying empirical data that are typical for that metal contaminant. In preferred embodiments, mathematical models or formulas are used to establish the curves or to interpolate sections between two measured points. In this way, the calibration of the metal detection system requires only a few measurements that provide at least the starting points of the curves or first and/or second arrays.

The gathered information is preferably stored in a memory of the control unit or a computer system that is attached to or integrated into the metal detection system. The information stored can then selectively be downloaded and used for the future calibration and configuration of the metal detection system.

BRIEF DESCRIPTION OF THE DRAWINGS

Some of the objects and advantages of the present invention have been stated, others will appear when the following description is considered together with the accompanying drawings, in which: Fig. 1 shows a block diagram of an inventive metal detection system; and

Fig. 2 shows the transmitter stage of the inventive metal detection system; Fig. 3a shows a curve or response locus established for a first metal contaminant MCi for a transmitter frequency f _TX i used as a fixed parameter and with the particle size as a variable parameter;

Fig. 3b shows a first array of curves established for the first metal contaminant MCi for three different transmitter frequencies f _T xi, ίτχ3/

Fig. 3c shows a second array of curves established for three different metal contaminant types MCi, MC ₂ , MC ₃ for one transmitter frequency f _T xi; Fig. 3d shows an area A _PS of the product signals for a scanned product with signal vectors of different phases and amplitudes that define the envelope of the area A _PS of the product signals and discriminator lines D that delimit the area A _PS of the product signals, that will be suppressed;

Fig. 3e shows a superposition of the second array of curves shown in figure 3c and the area A _PS of the product signals shown in figure 3d; and

Fig. 4 shows an illustration of the computer program 50 that is used to implement the inventive method.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Figure 1 shows a block diagram of an inventive metal detection system, which comprises a transmitter unit 1, a balanced coil system 2 with a transmitter coil 21, a first and a second receiver coil 22, 23, a receiver unit 3, a signal processing unit 4, and a control unit 5 that comprises standard interfaces, input devices and output devices, particularly a monitor. Figure 1 further shows a conveyor 6, on which products P are transferred through the transmitter coil 21 and the receiver coils 22, 23. The transmitter unit 1, for which a preferred embodiment is shown in detail in figure 2, provides a transmitter signal si to the transmitter coil 21 of the balanced coil system 2. Further, the transmitter unit 1 provides a reference signal slO having the transmitter frequency f _TX to the receiver unit 3. The transmitter signal si induces signals s22, s23 in the identical receiver coils 22, 23 that are of the same magnitude but inverse polarity as long as the system is in balance, i.e. as long as the conveyed products P are not influencing the magnetic field themselves and are not contaminated with metals. In the event that a product PC is contaminated with an electro- conductive object, then the signals s22, s23 in the identical receiver coils 22, 23 will change while that product P _c passes through the balanced coil system 2. As a result the transmitter frequency f _TX induced in the receiver coils 22, 23 gets modulated with a baseband signal, whose amplitude and frequency are dependent on the property, dimension and travelling speed of the electro-conductive object.

The output signals s22, and s23 of the receiver coils 22, 23 are applied to center-tapped primary windings of a balanced transformer 31 that mirror the receiver coils 22, 23. Further, the balanced transformer 31 comprises two identical center- tapped secondary windings whose opposite tails are connected to an amplifier 32. The outputs of the amplifier 32 are connected to a filter unit 33 which provides the amplified and filtered signals to a demodulation unit 34, which provides at its outputs the in-phase and quadrature components of the demodulated monitoring signal s30 and in-phase and quadrature components of the baseband signal, which originates from the conveyed products P. The in-phase and quadrature signals provided at the outputs of the demodulation unit 34 are forwarded to a further filter unit 35, which allows the desired signals to pass through to a gain unit 36 that allows setting the amplitudes of the processed signals to a desired value. Subsequently the filtered and calibrated signals are converted in an analogue to digital converter 37 from analogue form to digital form. The output signals of the analogue to digital converter 37 are forwarded to a signal processing unit, such as a known digital signal processor 4, which compares the demodulated and processed monitoring signals with reference values. The data resulting in the evaluation process are then forwarded to a data processing unit such as the central processing unit of the metal detection system, an internal or external control unit such as a computer terminal 5. In order to control the measurement process the signal processor 4 or the control unit 5 is capable of controlling the functions of the various modules provided in the transmitter unit 1 and in the receiver unit 3. For this purpose, the signal processor 4 is forwarding a first control signal c32 to the amplifier unit 32, a second control signal c33 to the first filter unit 33, a third control signal c35 to the second filter unit 35, a fourth control signal c36 to the gain unit 36 and a fifth control signal c37 to the analogue to digital converter 37. With these control signals c32, c33, c35, c36 and c37 the amplification and filter characteristics in the individual receiver units 32, 33, 35, 36 and 37 can be selected or adjusted. A sixth control signal cl2 is forwarded to the transmitter unit 1 as described below.

The receiver stage 3 described above is of course a preferred embodiment. The inventive method however can be implemented in metal detection systems that use differently structured receivers 3.

Figure 2 shows a block diagram of the transmitter unit 1 of the metal detection system shown in figure 1.

The transmitter unit 1 comprises a reference unit 11 that provides a reference signal sO with a reference frequency f _REF to a signal source 12, such as a frequency synthesiser 12 that is controlled by the sixth control signal cl2 received from the signal processor 4 or the control unit 5. The signal processor 4 can therefore select a suitable transmitter frequency f _TX that is forwarded with signal slO to a power amplifier 13, which is providing the amplified transmitter signal si to the transmitter coil 21 of the balanced coil system 2. Signal slO is also forwarded to a module 38 in the receiver stage 3 that provides in-phase and quadrature components of the reference signal slO to the demodulation unit 34. The metal detection system described above allows to measure products and contaminants with the application of various transmitter frequencies f _TX that are selected according to the inventive method. The inventive method is implemented by means of a computer program (see figure 4) that is stored preferably in the signal processor 4 or the control unit 5. Modules of the computer program can also be implemented in distributed processors .

Figure 3a shows a curve or response locus established for a first metal contaminant MCi for a transmitter frequency ΐτχι , which is used as a fixed parameter, and with the particle size of the first metal contaminant MCi as a variable parameter. This curve can be obtained in various ways. With one method the operator sequentially transfers metal particles of steadily increasing sizes, e.g. 1mm, 2mm, 3mm, through the balanced coil system 20 and records the related signal vectors svl, sv2, sv3. By connecting the endpoints of the signal vectors svl, sv2, sv3 a program module constructs the related curve. In the event that only a small number of signal vectors svl, sv2, sv3, have been recorded, then the line sections between two points are obtained by interpolation. In the event that the typical progression of such a curve or the characteristics have been recorded for the metal contaminants MC, then it is sufficient to measure only one or two signal vectors svl, sv2, and to construct the curve based on empirical values. In the event that the metal detection system has not changed its status and a calibration is not required, then the pre-stored curves for one or more potential metal contaminant types MCI, MC2, MC3, can be downloaded from memory.

Figure 3b shows a first array of curves established for the first metal contaminant MCi for three different transmitter frequencies f _TX i , ΐτχ2, fτχ3 · The curves can again be obtained by measuring the signal vectors for different sizes of the metal contaminant MCi for each of the frequencies ΐτχι , ίτχ2, fτχ3 · Alternatively, the curves can be obtained by taking one measurement only, e.g. for a particle size of 2 mm at the transmitter frequency f _TX 2 and by applying empirical data. Figure 3c shows a second array of curves established for three different metal contaminant types MCi , MC2 , MC ₃ for one transmitter frequency f τχι . The curves were established as described above for figures 3a and 3b. Figure 3d shows an area A _PS of the product signals that were taken while scanning a product P with the transmitter frequency fxi . The product signals are represented by signal vectors of different phases and amplitudes that define the envelope of the area A _PS of the product signals. Further shown are discriminator lines D that delimit the area A _PS of the product signals, which will be suppressed or blanked by the computer program 50 (see figure 4) provided in the signal processing unit 4 and/or the control unit 5. Typically, modules of the computer program 50 that relate to the control of the acquisition of calibration data are implemented in the control unit 5, while modules of the computer program 50 that relates to the processing of signals, particularly the suppression of unwanted signals, are implemented in the signal processing unit 4.

The area A _PS of the product signals is suppressed for example by means of adjusting the product phase until the discriminator lines D enclose the measured product signal. Signals of the metal contaminants MC i , MC2 , MC ₃ , that will extend beyond the discriminator lines D will then be detected, while product signals will be suppressed. However, it is understood, that there are other options of suppressing unwanted signals. E.g., the received signals may be mapped into a two- or three- dimensional representation, in which areas or volumes are defined, that will be suppressed. Signals that lie within this area or volume will be disregarded. Figure 3e shows a superposition of the second array of curves shown in figure 3c and the area A _PS of the product signals shown in figure 3d. In this illustration it can be seen that only small sections of the curves of the first and the second metal contaminant MC I , MC2 lie inside the area A _PS of the product signals. Hence, for this transmitter frequency f _TX i the metal detection system is capable of detecting particles of the first and the second metal contaminant MC I , MC2 that are significantly smaller than 1 mm. However, it is shown that for the third metal contaminant MC3 a large part of the curve, including the point that relates to a particle size of 1 mm, lies within the area A _PS of the product signals.

In figure 3 e the intersection points IP _M ci / IP _M C ₂ / IP _M C3 of the discriminator lines and the curves of the metal contaminant types MCi , MC ₂ , MC ₃ , that form a second array, are shown. These intersection points IP _M ci / IP _M C ₂ / IP _M C3 indicate particle sizes of the metal contaminant types MCi , MC ₂ , MC ₃ that can no longer be measured. However, the comparison of the intersection points IP _M C _I IP _M C ₂ / IP _M C3 obtained at least with a first and a second transmitter frequency f _TX i , f _TX2 allows to determine, with which transmitter frequency f _T xi or f _TX2 smaller particle sizes of the metal contaminant types MCi , MC ₂ , MC ₃ , of interest can be measured. Based on the gathered data, as shown in figure 3 e , the computer program 50 can therefore decide, which transmitter frequencies ΐτχι , £τχ2, - shall be applied. In the given example the computer program 50 may decide that for the metal contaminant types MCi and MC ₂ the first transmitter frequency f _TX i is suitable, while for the third metal contaminant MC ₃ another transmitter frequency ΐ may provide better results.

In the event that the area A _PS of the product signals has accurately been mapped, then the intersection points of the boundary of the area A _PS of the product signals and the curves of the metal contaminant types MCi , MC ₂ , MC3, that form a second array, may be determined. When scanning a product P the computer program 50 may alternately or simultaneously apply the selected transmitter frequencies f _TX i and fτχ-χ . In the event that the transmitter frequencies f _TX i and fTx-x are alternately applied then the sequence of application is preferably selected or selectable according to one or more of the process parameters. The number of alternations will typically depend on the size of the products P and the metal contaminant types MC and may freely be selected. Typically, the number of the alternations is selected in the range between 1 and 50.

In the event that the first of two metal contaminants MCi , MC ₂ would provide a strong signal and the second would provide a small signal, then the duty cycles with which the transmitter frequencies f _TX i and f _T x- _x are applied can advantageously be adapted. The time of the application of the transmitter frequency f _TX optimised for the second metal contaminant MC ₂ would typically be by a factor in the range of 2 to 10 higher than the time of the application of the transmitter frequency f _TX that has been selected for the first metal contaminant MCi . The inventive method therefore allows starting the measurement processes, i.e. the process of scanning the conveyed products P with the preferred or optimised transmitter frequencies f _TX i and ίτχ-χ· Hence, the time that is available for measurement, when the product P is passing the balanced coil system 2, is fully used by applying the most preferable transmitter frequencies f _TX . All data collected therefore contributes to the final measurement results. Hence, due to the application of optimised transmitter frequencies f _TX for a longer period of time, the resulting sensitivity for the detection of the metal contaminants MC increases significantly. Further, while at least for the initial setup of the metal detection system, additional efforts are required, the overall efforts of operating the metal detection system are significantly reduced. A process for analysing, evaluating and selecting data is no longer required.

Further, the inventive method constitutes an important improvement for the simultaneous application of more than one transmitter frequency f _TX . Based on the above described calibration process only a few transmitter frequencies f _TX will be required for performing an optimised measurement. Consequently the efforts for separating the reduced number of transmitter frequencies f _TX , e.g. by means of filter techniques implemented in the signal processor 4 or by means of switched filter banks 33 (see figure 1), will be relatively low in comparison to a known systems.

Figure 4 shows an illustration of a preferred embodiment of the computer program 50 that is used to implement the inventive method. The computer program 50 comprises four essential modules 51, 52, 53 and 54.

The first program module 51 serves for establishing the data of metal contaminants MC that possibly appear in the products P. More particularly, the first program module 51 serves for establishing the discussed first and/or second arrays of curves shown in figure 3b und 3c. For this purpose the first program module 51 accesses the data file 553, in which the available transmitter frequencies f _TX i, ΐτχ2, ίτχ3, ^ΤΧΛ, - , are listed that serve as variable parameters for the first arrays and as fixed parameters for the second arrays. The number of transmitter frequencies f _TX is typically in the range between six and twenty, but can freely be selected. The first program module 51 may access further data files 5511, 5512, .... In the data file 5511 empirical data of the metal contaminants MC are listed. The data file 5512 preferably contains pre-recorded curves and/or first and/or second arrays of the metal contaminants MC . In the event that at least one calibration process for a metal contaminant MCi is performed, with one particle size and one transmitter frequency f Tx i , then the related data are forwarded to the first program module 51 via the data bus 5513. The established data, such as a plurality of first and second arrays, are forwarded in data files 510 to the third program module 53.

The second program module 52 serves for establishing data of at least one product P for the selectable transmitter frequencies ^Txir ί τχ2 / ίτχ3 / f τχ4 Λ - · The second program module 52 may access the data files 5521, 5522, .... The data file 5521 contains empirical data of the products P. The data file 5522 preferably contains pre-recorded data of products P of interest. In the event that at least one calibration process with a product P is performed, then the related data are forwarded to the second program module 52 via the data bus 5523. The established data such as a plurality of areas A _PS , each established for a transmitter frequency f x i ," ΐ τχ2 / ΐ τχ3 / ΐτχ4 / ···, are forwarded in data files 520 to the third program module 53.

By selectively using the data files 5511, 5512, 5521, 5522,

... and/or concurrent data of calibration processes the required data for the comparison processes, i.e. the evaluation of the results of the application of the transmitter frequency f Tx i ," f X2 ίτχ3 ίτχ4 can be established in various ways.

In the third program module 53 the data established for the product P and the contaminants MC are compared for each transmitter frequency f _TX i ; f X2 ίτχ3 ί τχ4 - · As symbolically shown in figure 4, superpositions of data of the product P and two contaminants MCi, MC ₂ are established for a first and a second transmitter frequency f x i f τχ2 · Then it is determined, with which transmitter frequency ΐ ιί ίτχ2 the smallest particle sizes of the contaminants MCi , MC2 can be detected. This process is preferably performed for all selectable transmitter frequencies f _TX i , f _T x2 , ίτχ3 , ίτχ4 , Finally, the results of the third program module 53 are forwarded to the fourth program module 54 that controls and coordinates the individual processes, preferably including the calibration and measurement processes. Particularly, the preferred transmitter frequencies f xi ," ΐτχ2 / ΐτχ3 / ΐτχ4 / ... and the minimum particle sizes of the metal contaminant types MC i , MC2 , ... that can be detected therewith are reported to the fourth program module 54. Depending on configuration parameters selected by the operator, e.g. on a data terminal 5, the fourth program module 54 is setting up the measurement process and forwards the related control signals to the individual electronic modules of the metal detection system. In particular, the fourth program module 54 initiates sending the control signal cl2 to the transmitter unit 1. Furthermore, the status of the measurement process may continuously be observed by electronic sensors, preferably optical sensors that provide signals to the fourth program module 54 allowing for timed- performed measurement sequences.

The fourth program module 54 preferably comprises data storage units, such as data files 541, 542, in which at least the results of the calibration processes are stored. In the data file 541 the results of the calibration processes, particularly the minimum particles sizes of the metal contaminant types MC i , MC2 , ... and the related transmitter frequencies f _TX i , ΐτχ2, ίτχ3 , ίτχ4 _/ ··· · _r are stored. In the data file 542 configuration data for various measurement processes can be stored for repeated use . The fourth program module 54 preferably also communicates with the signal processor 4, the control unit 5 and other devices that are contained for example in a computer terminal.

During the operation of the metal detection system the fourth program module 54 preferably collects further data derived from the product P and the metal contaminants MC in order to maintain optimal conditions with a calibration process running in parallel to the measurement process.

For the operation of the metal detection system it would be desirable to reduce the number of transmitter frequencies f^xi, f _TX 2, so that not for every individual product a specific frequency needs to be selected.

According to the invention different products are assigned to clusters that are assigned each to an optimised transmitter frequency f _TX . Clustering the products therefore allows obtaining improved measurement results with high efficiency.

The process of clustering can be performed in various ways. Preferably the product information is still obtained for all available transmitter frequencies f _TX . As stated above, when scanning a product, typically various signal components occur each having an individual phase and magnitude. Connecting the vectors of all signal components leads to an envelope that is the boundary of the area A _PS of the product signals or the product signature that will be suppressed by the signal discriminator.

Figure 3d shows an area A _PS of the product signals that were taken while scanning a product P with the transmitter frequency fxi . The product signals are represented by signal vectors of different phases and amplitudes that define the envelope of the area A _PS of the product signals. Further shown are discriminator lines D that delimit the area A _PS of the product signals, which will be suppressed.

According to the invention products with similar or equivalent product signatures A _PS are grouped and assigned to an optimised transmitter frequency f _TX . For this group or cluster discriminator lines D are then set, which ensure that each product signature A _PS is suppressed, when the corresponding product is passing through the detector.

Alternatively, products with discriminator lines D that lie within a selected range are grouped for an optimised transmitter frequency f _TX .

As a further alternative, discriminator lines D maybe defined, based on which stored product signatures A _PS or stored discriminator lines D are retrieved, that lie between said discriminator lines D. The operator may therefore select acceptable particle sizes of the metal contaminant types MCi , MC ₂ , MC ₃ and corresponding discriminator lines D. For these discriminator lines D the implemented computer program 50 will list all products that can be grouped for individual transmitter frequencies f _T xi, ΐτχ2 _/ ■■■·

In a further preferred embodiment, the operator may select a first product P that needs to be measured. The computer program 50 can then analyse if further products P exist, preferably within a selected tolerance range that can be clustered together with the first product.

The reduced set of transmitter frequencies fTxi, ίτχ2 , ■■· , consists preferably of transmitter frequencies fTxi, ίτχ2 , ■■· , which are selected in such a way that cluster sizes are obtained containing a maximum number of products. With this clustering process, which can be executed with the computer program 50 based on product data stored in the database 5, the efficiency of the metal detection system can significantly be increased.