GB2107858A - Fibre defect detection - Google Patents

Abstract

An array of fibrous material (1) for analysis is carried past an examination station with a light source (30) and a photosensor (34). To enhance the visibility of defects, the array is immersed in a liquid (12) having a refractive index substantially equal to that of the fibres of the array. The preferred embodiment incorporates a microprocessor which analyses the output signal from the photosensor (34) so as to detect the presence of defects. <IMAGE>

Description

SPECIFICATION Fibre defect detection The present invention relates to the detection and characterisation of defects in arrays of fibres, and particularly, though not exclusively, to methods and apparatus for detection and classification of foreign fibres in tops, for example coloured fibres or dark (black) hair in wool tops.

According to one aspect of the present invention, there is provided a method of detecting defects in an array of fibres, comprising observing the array whilst immersed in a liquid having a refractive index substantially equal to that of the fibres.

A preferred embodiment of the method comprises moving the array through the liquid, illuminating the array whilst immersed in the liquid, a photo-sensor being positioned to receive light from the array of fibres, and detecting the presence and character of the defect from signals produced by the sensor during passage of the defect.

The defects can be distinguished from the rest of the fibres by optical means. The so-called normal fibres are made of a relatively transparent and homogenous material, but when observing without special precautions a bundle of such fibres in a medium like air, the dispersion and refraction of light at the surface of the fibres make them appear dark to the naked eye. In such an elementary setup, and under the constraint of using a very thin sample, only the most pronounced defects will be discernible.

Placing the fibres into a liquid with a refractive index equal to that of the fibre material suppresses all light reflections and refractions at the fibre-liquid interface, so that a mass of fibres in the liquid will present only an homogenous attenuation of light, related to the material they are made of and not to their geometry.

A defect such as a naturally coloured fibre will of course retain its opacity and strongly contrast with its background, even if contained in a relatively thick sample.

The image of a transverse strip of the top is preferably focused onto a linear array of detectors by means of an objective, and the detected signals then processed.

In another aspect, the invention provides an apparatus for detecting defects in an array of fibres, comprising a container for liquids, means for conveying a fibrous array through liquid in the container, means for illuminating the array when in the liquid, and a photo-sensor disposed to receive light transmitted or reflected by the array, and means for analysing signals produced by the sensor.

By means of an appropriate scanning circuit, a time-varying analogue voltage can be obtained from the array of detectors. The amplitude of this signal at a given time will be proportional to the light reaching the corresponding detector, so producing a "video" signal.

The means for analysing the signals, for instance a minicomputer with associated interface electronics, fulfils among others the task of recognising defects within the array, preferably taking into account the case of several defects appearing at once under the sensor, and rejecting detected events which are not to be considered as defects.

The necessary data upon which this analysis is performed are extracted from the video signal.

The amount of raw information conveyed by such a signal exceeds by far the processing capabilities of a reasonably affordable computing system. Fortunately, part of this information is irrelevant and ought to be discarded as early as possible in the process. Furthermore, if there are times when a lot of data must be processed as quickly as possible (which may be the case when several defects at once are being carried past the detector), there are also some intervals when there is no useful information at all to extract (such as when there is nothing to be seen in the sample which could be interpreted as a defect). The rate at which relevant data are produced by the sensor assembly is therefore extremely variable.In a preferred arrangement, a device is provided to extract from the video signal the information strictly necessary for the subsequent process of recognising, identifying and counting defects. This device outputs the useful information as a string of digital words, which are temporarily stored in an electronic "FIFO" (first in-first out) memory, which is by design able to be filled at a high rate for short periods of time, and which is emptied at leisure for subsequent computer processing, performed at a slower but more uniform rate.

Some embodiments of the invention will now be described, by way of example, with reference to the accompanying drawings, in which: Figure 1 is a side elevation of one form of apparatus according to the invention, for detecting defects in wool tops; Figure 2 is an enlarged view of part of Figure 1; Figure 3 is a block circuit diagram of the sensor of the apparatus of Figure 1; Figure 4, 5 and 6 illustrate the scanning of the top and processing of video signals from the sensor; Figure 7 is a block diagram of parts of the signal processing circuitry of the apparatus; Figure 8 is a graph illustrating the analysis of a typical video signal; Figure 9 is a block diagram of the signal processing circuitry; Figure 10 shows a typical computer print-out; and Figure 11 shows a modified version of the apparatus.

In the apparatus shown in Figures 1 and 2 for detection of dark fibres in wool tops, a drafting unit 20 is used to prepare the material for the analysis.

The top 1 is supplied by the drafting system 20 comprising a double apron 22 and front rollers 21. The use of the apron reduces the adjustments required to accommodate different types of materials to be analysed. The top 1 then passes from the drafting unit 20 into a detector area 10 where it preferably moves at constant speed, with only a weak additional drafting. At this point, the material may typically have a linear density of 1.5 g/m (1.5 Ktex) with a width of around 13 mm.

In the detector area 10, the top passes into a container or dish 11 containing a liquid 12 having a refractive index substantially equal to that of the fibres of the top. In the case of wool, liquids such as O-dichlorobenzene, methyl-salicylate or benzyl alcohol can be used. Among these, benzyl alcohol has been found particularly satisfactory. Of course, care should be taken that the materials in contact with the liquid remain unaffected. For example, the dish 11 may be made of stainless steel, or of glass plates glued with silicone.

At a detection station 10 the top passes between a lower plate 13 of white reflective material, immersed in the dish 11 with its upper surface about 4 mm below the surface of the liquid, and a glass plate 14, about 10 mm thick, which is placed horizontally over the drafted top and then moved downwards thus forcing the top to be immersed in the liquid. In its final position the glass plate 14 lies over the white plate 13 with 2 mm spacing between them and the top passing through this gap. The lower surface of the glass plate is immersed while its upper surface remains dry. The edges of the glass plate are rounded in order to make easier the sliding of the sample under the plate.

After leaving the container 11, the top 1 is moved by further rollers 24 of the drafting unit, and is wrung dry between two such rolls after emerging from the liquid. The wringing rolls 24 should be made of, or coated with, a suitable material resistant to the liquid being used, and positioned so that the liquid wrung from the top is collected in the dish. A damping plate 1 9 helps to keep the surface of the liquid still, despite drops falling from the wringing rollers.

After passing between these rolls, the top is carried into a vessel 26 by an additional roll 25. The level of liquid with respect to the glass plate 14 is not critical as long as its bottom surface remains wetted. However, an addition of liquid can be made from time to time by means of a suitable tank 17 to compensate for loss of liquid carried away by the top or evaporated.

The apparatus also includes an optical detection system consisting of a light source and an observation device. The lighting device comprises an intense source of white light 30 and a condenser 31 in order to produce a uniformly illuminated zone of several square centimetres on the immersed sample. A slide projector lamp (150 W) with its reflector and condenser is well suited for this purpose.

The angle between the incident light beam and the plane of the sample is preferably about 45 degrees. In addition, a cylindrical lens 32 made for example of a lucite rod can be placed on the glass plate 14 in order to produce on the sample a more strongly illuminated strip corresponding to the region under observation.

The observation device comprises a focusing lens 33, for example a camera objective with iris, and an array of photodetectors 34 on which the image of a transverse strip of the sample in the illuminating zone is focused. In a preferred embodiment of the invention, the detector consists of an array of 1024 adjacent photodiodes whose sizes are 25x25 microns. With the optics adjusted to give a magnification factor of 2, the observed strip on the top is about 12.8 mm 12.5 microns. The tilt angle of the observation device's optical axis should be different from that of the illuminating light beam in order to avoid specular reflection problems. The spacing between the glass plate 14 and the white plate 13, which determines the thickness of that sample, should be kept reasonably small (typically, as mentioned, about 2 mm) to avoid depth of field problems.

In the present description of the invention, the observation is made by reflection on the white plate 1 3. In the absence of coloured fibres, the fibrous structure of the top immersed in the liquid of identical refractive index is not discernible. The light diffused by the white plate and passing through the top results in a uniform illumination reaching the detector. Coloured fibres, foreign fibres or other types of defect contrast strongly with the background. As such a defect crosses the illuminated zone, its image is focused onto the detectors, producing local areas of weaker illumination, i.e. a detectable event. Due to the preparation method of the top (i.e. drafting), the fibres are aligned in the direction of motion. Hence, as the top moves in the observation region with a known speed, the time during which an event is detected can be related to the length of the corresponding dark fibre. Moreover, the width of the detected fibre is given by the number of photodiodes undergoing a change in light intensity. For example, in the above apparatus, a 50 microns diameter fibre will give rise to an event extending over four photodiodes.

In an alternative embodiment of the invention (not shown), the detection can be done in transmission mode. In that case, the observation device would look directly toward the light source, the immersed sample being placed between them.

It is worth noting that the arrangement described is able to separate defects which are simultaneously under the sensor.

The manner in which the signals from the sensor 34 are processed will now be described.

The sensor 34 is a monolithic self-scanning linear photodiode array and comprises (Figure 3) a row of silicon photodiodes DO, D1 ... D1023 each with an associated storage capacitor CO, C1,...

C1023 and a multiplex switch (transistors TO, T1 . . . T1023) controlled buy a shift register SR. The shift register SR is driven by a clock; each scan is initiated by a start pulse. The start pulse loads a bit which is clocked through the register, successively opening and closing the switches (TO etc.) and thus connecting each photodiode (DO etc.) in turn, to the video line. As each photodiode is accessed, its capacitor is charged to the potential of the video line and is left open-circuited until next scan. During the interval between scans, the capacitor is discharged by an amount equal to the instantaneous photocurrent in the diode, integrated over the line scan. Each time a diode is sampled, this integrated charge loss must be replaced through the video line.The resulting video signal is a train of charge pulses, each proportional in magnitude to the quantity of light reaching the corresponding photodiode.

The sensor chip may typically be a RETICON type RL 1 024G, encapsulated in a 22 pin dual-in-line package, which contains 1024 photodiodes on 25 Mm centres and 26 iJm width, protected by a transparent window.

As the top 1 is moved in the "Y" direction (Figure 4) under the glass plate successive strips of the top are scanned by the sensor, as illustrated in Figure 5. The direction of the scan (across the sample) is referred to as the "X" direction. Figure 5 also shows the form of the resulting video signal and the corresponding X coordinates (diode number 0,1,2... 1023) and Y coordinates (line number n, n+ 1, etc.). The video signal from the sensor 34 is fed to a front end unit 100 which amplifies the signal and also provides the start and clock pulses for the shift register SR. The pulses are so related to the drafting speed that one scan occurs every 1/8 mm of material.

It is assumed that among the fibres of the sample a defect (dark fibre 35) crosses the illuminated strip 36 which is imaged onto the photodiode array 34 by the optics 33. During one scan the voltage V1 (t) at the output of the front end unit will take successive values proportional to the amount of light striking each photodiode in turn, from 0 to 1023. The video signal V1 (t) is sketched in Figure 4, from which it will be noted that the waveform is curved as the average light intensity striking the sensor is higher at the centre of the array than near the edges partly due to the non-homogenous illumination of the top: the distance from the light source and hence the illumination is not the same for every point (L , < Lb on Figure 6). A simiiar effect occurs for the light travelling back towards the sensor.The presence of the defect 35 causes a local decrease of the light intensity which is seen- as a dip in the recorded voltage waveform V1 (t).

A corrector circuit 101, by sampling and averaging the video signals from successive scans, produces a reference voltage which is then subtracted from the instantaneous video signal Vl(t) to give a "normalised" signal V2(t) with a flat baseline. As seen in Figure 7, the corrector has a circuit 102 in which the curvature of the reference signal is extracted from the video signal itseif which is sampled and held in a set of capacitors during the scanning. A special circuit constructs a fitting signal joining each other memorised level. Subtraction of the video signal from the reference signal is performed in differential amplifier 103 to the 'flat' normalised video signal where it is easier to use fixed voltage as threshold levels of the analogue to digital conversion.The reference signal can also be used to perform automatic regulation of the intensity of the light source.

The normalised video signal V2(t) then passes to an analogue to digital converter or threshold detector 104 where the instantaneous value of the signal is compared with five reference values or "thresholds" (TH 1 to TH5 on Figure 8). Levels TH 1 and TH5 are adjustable by potentiometers 106, 107, levels TH2, 3 and 4 being equally distributed between levels 1 and 5. When the signal stays below the first (and iowest) threshold, no information at all is generated by the circuits. A detected "event" begins when the signal exceeds the first threshold (for X=13 on the example) and ends when the signal returns below it (for X=24). There may be several such events during one scan. The detector 104 produces, for each X coordinate, a value from 0 to 5 according to the signal level.

Following the conversion, detection logic 108 stores information in a FIFO (first in-first out) memory 1 15--described further below-each time a threshold level is crossed. This information, called an X information, consists of 16-bit digital word giving the level reached and the number of the sensor element (photodiode) for which this amplitude change occured. This is equivalent to the transverse coordinate of the event with respect to the sample (e.g., in Figure 7 level 4 for X=i 4). This information is formatted as follows:- 10 bits for the photodiode position where the threshold level was crossed (from O to 1023).

3 bits for the threshold level number which was crossed.

1 bit to identify an X information.

(the remaining bits are not used).

This way, information is stored into the FIFO memory only when a defect is effectively seen under the sensor. The X coordinate data is obtained from an X counter 110, forming part of control logic 109, which counts from 0 to 1023 and represents the position of the photodiodes in the video signal. It is reset by the START pulse and incremented by one at each CLOCK pulse.

Similarly, a Y counter 111 is incremented by one at each START pulse and counts the number of scans. (Figure 5 and 7). The control logic 109 also includes units for end of line information (112) and control 11 3 for transmission of information over a data bus 114 to the FIFO memory.

During the scan of one line, an event starts when the normalised video signal crosses the threshold level 1 and ends when the normalised video signal returns under the threshold level 1 (Figure 8). More than one event, in the sense described above, can be simultaneously present on the same scanned line. It is a task of the program of the host computer to eventually determine whether these events are parts of the same natural defect, or effectively reflect several natural defects.

For the duration of an event, a digital integrator 108 computes an approximation of the area lying under the event envelope and above the first level. The number representing this area can be evaluated as the number of clock pulses where the signal is above the level 1, added to the number of clock pulses where the signal is above the level 2, and so on until level 5. Thus, for the waveform shown in Figure 8, the total is given by level: 1 11 2 8 3 7 4 7 5 4 37 This information is transmitted to the FIFO memory at the end of the detected event with a special flag, as a so-called SIGMA word.The X and z information for the waveform of Figure is thus Threshold Address X info 2 1 3tbeginning of the defect 4 14 5 15 4 19 1 21 0 24tend flag 37cdarkness of the defect At the end of each scan during which at least one event was recorded, a "Y" information word is separated, consisting of a digital word giving the length of sample examined thus far (more precisely the number of scans performed since the beginning of the analysis). This is equivalent to the longitudinal coordinate of the event with respect to the sample.The Y information is also a 1 6-bit word, with the following information: 12 bits for the line scanning number (Y counter) 3 bits for flags as END of analysis OVERFLOW of the Y counter GAP indication 1 bit to identify "Y" information.

A Y information word may also be sent alone (without any X information during the corresponding line) merely for the purpose of transmission of the flag information.

The line number is transmitted only when an event appeared in the line.

End flag is transmitted upon reception of the appropriate microprocessor command to stop the measurement, so specifying it is the last information to be transmitted. The end command may be an operator action (keyboard entry) or an automatic action performed by the microprocessor (FIFO FULL error).

Overflow flag is transmitted when the Y counter reached its maximum value, to inform the rest of the system that subsequent Y information will be reset to zero. This fact has to be taken into account for the various software counters in the microprocessor and the host computer.

Gap flag is transmitted when no event has been encountered during N scan lines, which is an information the recognition program of the host computer needs to analysis the data.

X, Y and SIGMA data for all detected events are all the information required for the subsequent computation to recognise and classify defects. Its generation rate is very much lower than the speed at which one would be obliged to process the raw video signal.

Due to the difference between the instantaneous speed at which the sensors provide information, and the speed at which the host computer can process them, all the 1 6 bits information words are loaded in a RAM (Random access Memory) used as a microprocessor controlled FIFO (First In-First Out). The "FlFO" memory 11 5 is a temporary storage device for the digital data: the information earliest written into is ("first in") will be the first to be retrieved ("first out"). The rates at which data are entered into the FIFO on one hand, and read out on the other hand, may be entirely unrelated as long as the long term average input rate does not exceed the average output rate to avoid overflow.Event detecting logic located to block 107 performs the task of writing in the FIFO memory the "X", "Y" and "SIGMA" data at the appropriate time. The FIFO is unloaded by a control device on request of a remote computing system which uses the information provided to recognise a defect from successive events, classify it according to size and opacity, and accumulate statistics on defects. The FIFO memory comprises a 2K 16 bit per word RAM array and the microprocessor is used as the FIFO controller. Data are loaded in the FIFO at the sensor rate using the DMA (Direct Memory Access) principle, using a write counter 11 6 as an address pointer. The write counter is incremented on every word loaded in the FIFO.

Data are unloaded by the microprocessor with normal addressing, using a software read counter, which is incremented on every word read from the FIFO, via the microprocessor data and address buses 111, 112 and an address multiplexer 117. A suitable circuit is able to detect that a sensor attempts to load the FIFO at a location not yet unloaded by the microprocessors, reflecting a so-called FIFO-FULL condition. This situation is unrecoverable, and the microprocessor is notified to stop the measurement. The FIFO data are unloaded by the microprocessor upon request of the host computer, to which they are transmitted as two 8-bits bytes using one RS232-type serial line per sensor (baud rate: 9600bd).

The control device is a general purpose microcomputer, containing a microprocessor, keyboard and display, the necessary links with the FIFO on one side, and with a computer on the other side, and the necessary program to perform the following tasks: Be the logic controller of the whole measuring system; Inform the operator of the status of the system; Allow the operator to start, interrupt, end or abort the measuring activity; Generate all timing references for the sensors, counters, etc; Read data from the FIFO memory and transmit it to the computer; Take action in case of abnormal conditions such as FIFO overflow.

The computing system is a minicomputer with associated device for printing results, and programmed to perform the following tasks: According to proximity rules, link together the events produced during the scan of one defect.

Reduce each recorded defect to a set of characteristic parameters.

According to classification rules applied to these parameters, decide for each detected "object" on the sample, whether to reject it as spurious information, or to keep it in memory as a recognised defect.

On command from the control device compete the above activities and print statistics data about all recognised defects, including for example size, shape and opacity information.

The overall layout of the interface logic is shown in Figure 9. As previously noted, FIFO arrangement is connected via the multiplexer 11 7 and data buffers 11 8 to the buses 111, 112 of a microprocessor system which also includes a microprocessor 11 9 with associated program store 1 20 and working store 121.

The microprocessor:-- controls the transfer of data between the FIFO and the host computer, displays messages to inform the operator of the evolution of the measurement, generates the required signals to control the interface system (e.g. clock and start pulse, for the front-end unit).

The display and keyboard available on the microprocessor are used by the operator before the beginning of the measurement to specify the material to be analysed, the date, and some other housekeeping information. These informations are also transmitted to the host computer and appear in a "bulletin" printed at the end of the analysis.

These informations are: LOT definition SAMPLE number CUSTOMER name OPERATOR name TYPE of material DRAFT of the drafting machine TEX value of the analysed top.

During the analysis, the display is used to show messages like: WAIT OPERATOR The operator is prompted to begin the analysis when he is ready.

WAIT MOTOR The operator is prompted to turn on the drafting machine.

READY The operator is informed that the interface system is ready to accept a command.

ACTIVE This message is displayed along with a counting value in meters reflecting the length of analysed top, and appears when a measurement is effectively running.

END The operator is informed that the measurement is completed, and he is permitted to request a new one.

ENDING The operator is informed the measurement is to be completed, but he has to wait for the FIFO to be completely unloaded.

ERROR The operator is informed an abnormal situation has occurred, such as a FIFO FULL condition, or a transmission error between the host computer and the microprocessor.

The host computer 122 communicates with the microprocessor via an interface 1 23 and serial lines 124 and is used to calculate a set of parameters for each defect, which are chosen to fully characterise the defect.

The software program inside the host computer need not hold in its memory the total amount of data that represents a defect as they are sent by the interface system. Instead, the program retains a set of values, characterising at any time the "part of defect" acquired up to this point. At every scanning line containing an event, this set of values is updated using the information sent by the microprocessor.

When the defect is recognised to be ended (upon reception of a GAP information), the final set of parameters is computed that fully characterise the defect in size, shape, irregularity, etc., taking into account the data accumulated about this defect from each sensor.

These parameters are compared to some reference parameters, to determine the class in which the defect must be counted.

When the END flag is detected, the host computer will print the results of the classification process in the form of a "bulletin" (printer 125).

Classification requires to compute parameters as: size shape grey levels projected length discontinuities Figure 9 is a representation of the defect as it is possible to "rebuild" it from the X and Y information received by the host computer. The printed character is the threshold level number crossed. The space of a character on the paper represents a square of 1/8 mm on the top.

In a typical apparatus, the sensors will be scanned at a rate and sensitivity such that each photodiode will "see" contiguous squares of 1/80 of a mm from a 12.8 mm wide top. The speed of the sample moving under the sensor may be more than 1.5 metre/minute.

At the operator's choice, the sample may be moved under the sensor without analysis, in order to give the drafting machine enough time to reach a steady and homogenous production, before starting the measurements.

Means are provided to automatically transmit the measurement results (statistics summary for the unit of length respectively weight of the sample) to another computer for archival and later use.

The sensor arrangement described may also be combined with the dual reflection/transmission defect sensor described in our co-pending U.K. patent application no. 8110057, as illustrated in Figure 11, where such a dual sensor arrangement 60 is placed on the same drafting machine as the liquidimmersion sensor unit 1 0. The signals from the three sensors may be processed by a common interface unit (100 to 11 5), control unit 1119 to 121) and computer with associated printer (122, 125).

In the combined arrangement it can be arranged that the "liquid immersion" sensor detects the presence of defective fibres such as dark hair in a wool top, whilst the dual sensor serves to detect other kinds of defects (including for instance foreign matter like burrs and seeds in wool tops); the dual sensor having a lower resolution (e.g. 1/8 mm as opposed to 1/80 mm) than the liquid sensor.

In appropriate cases, the dual sensor arrangement may be designed to view the sample when immersed in the liquid.

Claims (14)

Claims

1. A method of detecting defects in an array of fibres, comprising observing the array whilst immersed in a liquid having a reflective index substantially equal to that of the fibres.

2. A method according to claim 1, comprising moving the array through the liquid, illuminating the array whilst immersed in the liquid, a photo-sensor being positioned to receive light from the array of fibres, and detecting the presence or character of the defects from signals produced by the sensor during passage of a defect.

3. A method according to claim 1 or 2, wherein the array is continuously moved through the liquid.

4. A method according to claim 3, in which an image of a strip of the array, transverse to the direction of movement, is focused on the photo-sensor.

5. A method according to claim 4, in which the photo-sensor consists of a linear array of adjacent detectors.

6. A method according to any one of the preceding claims, including the step of classifying the detected defects according to their dimensions and darkness.

7. A method according to claim 1, substantially as herein described.

8. An apparatus for detecting defects in an array of fibres, comprising a container for liquid, means for conveying a fibrous array through liquid in the container, means for illuminating the array when in the liquid, and a photo-sensor disposed to receive light transmitted or reflected by the array, and means for analysing signals produced by the sensor.

9. An apparatus according to claim 8, in which the conveying means is arranged to move the array continuously through the liquid.

10. An apparatus according to claim 8 or 9, in which the illuminating means includes a cylindrical lens arranged to iiluminate a strip of the array transversed the direction of movement.

11. An apparatus according to claim 8, 9 or 10, including focusing means arranged to form an image of a strip of the array, transversed the direction of movement, on the photo-sensor.

12. An apparatus according to claim 11, in which the photo-sensor consists of an array of adjacent detectors arranged to produce a serial video signal.

13. An apparatus according to any one of claims 8 to 12 including roller means for wringing liquid from the array after exit from the liquid.

14. An apparatus according to any one of claims 8 to 13, in which two sensors are spaced apart alongside the path of the fibrous array, a source of light is disposed on the opposite side of the said path to one of the sensors and second and third sources of light are disposed respectively on the same and opposite sides of the array as the other sensor.

1 5. An apparatus for detecting defects in an array of fibres substantially as herein described with reference to Figures 1 to 4 of the accompanying drawings.

1 6. An apparatus for detecting defects in an array of fibres substantially as herein described with reference to Figure 5 of the accompanying drawings.