WO2006120704A1

WO2006120704A1 - A method for drafting of sliver and apparatus thereof

Info

Publication number: WO2006120704A1
Application number: PCT/IN2006/000148
Authority: WO
Inventors: Prabaharan Balakrishnan
Original assignee: Sree Ayyanar Spinning and Weaving Mills Ltd
Current assignee: Sree Ayyanar Spinning and Weaving Mills Ltd
Priority date: 2005-05-06
Filing date: 2006-04-27
Publication date: 2006-11-16
Anticipated expiration: 2007-11-06
Also published as: CN101171375A; BRPI0610234B1; BRPI0610234A2; CN101171375B; TR200707478T1

Abstract

The present invention relates to a method which performs levelling, by ensuring that there is a uniform distribution of leading-ends per unit length on the delivered output. The invention particularly relates to a method for drafting of a fiber mixture of a textile the method comprising detecting variations in the leading-ends in the feed, computing the corrections needed for minimizing the variations in the output, and applying corrections during operation.

Description

A METHOD FOR DRAFTING OF SLIVER AND APPARATUS THEREOF. Field of the Invention

The present invention relates to a method for drafting fiber mixture of textile which involves detecting variations in the feed thickness, computing the corrections needed for reducing the variations and applying the necessary corrections providing precise control over feed thickness.

Background

Auto-leveling is the name of the technology used to control the mass variations in spinning preparatory-machines. The apparatus for auto-leveling is typically used in the carding and drawing stages to control the hank of the sliver delivered. Several open-loop and closed-loop systems exist, that control the mass variations in the delivered sliver. However, these methods depend on the direct measurement of the thickness of the sliver. Subsequently, based on the thickness of the sliver, the speed of the auto-leveling rollers is adjusted. Conventional drafting apparatus consists of a system of three pairs of rollers. A break draft is usually fixed, while a variable main draft is used to control the leveling process. As the material passes through the drafting system, several corrections may have to be applied to vary the main draft. This makes the relationship between the measured thicknesses and the applied corrections unclear.

US Patent 5,018,248 relates to a drafting system and an auto-levelling apparatus consisting of two rollers referred to as the "tongue" and the "groove" rollers. The two rollers are arranged such that the "tongue" roller and the "groove" roller of the sliver sensor are placed ahead of the first drafting rollers. These two rollers serve to measure the thickness of the sliver between the tongue and the groove. The output measured as a response to the distance between the two rollers is used, subject to a time delay, to vary the speed of rotation of the subsequent drafting rollers. The auto-leveller throughput speed is adjustable and the time delay in the draft ratio variation is automatically adjusted in response to the speed selected. It should be noted that in the stated invention the auto leveling is accomplished through the corrections done at the macro level using the thickness measurements, whereas the underlying micro variations responsible for the non-uniformity in the feed are not considered.

The conventional auto-leveling systems thus attempt to determine the optimal value for the correction to be applied by trial and error on the presumption that the series of thickness measurements can be used to compute the series of needed corrections. These corrections should be applied when the material reaches the optimal correction point.

Object of the Invention

The main object of the present invention is to provide a system so as to obtain a sliver possessing high degree of uniformity by reducing variations at the micro as well as at the macro level.

Another object of the present invention is to ensure that the fibre ends reach the delivery rollers at a constant rate per unit time throughout the drafting process. Yet another object of the present invention is to determine the thickness of the sliver by correlating the thickness with other parameters such as the number of leading fibre-ends in the sliver

Another object of the present invention is to enable the error correction to be applied to the draft rollers based on the fibre-ends distribution in the input and output.

One more object of the invention is to provide an auto levelling drafting apparatus in which the throughput rate is adjusted, so as to provide a sliver having high degree of homogeneity by increasing/improving auto levelling accuracy. Summary of the invention

Accordingly, the present invention provides a new model for the feed slivers which describes the sliver as a system with micro variations in fiber-end distributions, and the sliver thickness as a convolution of this distribution. The present invention utilizes the micro-level structure of the sliver. The present invention retains and surpasses the ability of conventional mechanisms to correct errors occurring due to the variations in the mean thickness. It corrects variations, referred to as stationary variations, which do not involve a shift in parameters like mean or average thickness. The method is based on a sound mathematical model of the sliver, and uses well established digital signal processing techniques to recover the distribution of the micro-variations that fully characterizes the feed at the micro level. The system achieves corrections at the micro level, and hence has been termed the "Micro-leveller". The ability to ensure corrections at the micro- level provides the capability to correct the feed to a far better extent than the existing systems.

Brief Description of accompanying drawings Figure 1 is a schematic representation of a micro-leveller system

Figure 2 is a representation of the scanning roller system behaving similar to a mass-spring damped harmonic system of a micro-leveller system Figure 3 is a simulated comparison between the levelling technique of the present invention with conventional auto levelling techniques and with situations where no auto levelling is used.

Figure 4 shows an understanding of instantaneous draft using leading-ends

Figure 5 shows an "All or None" type of correlation between thickness and leading-ends.

Figure 6 shows a graphical representation of fibre impulse response. Detailed Description of the Invention

Uniformity is one of the most desired quality parameters of spun yarns not only because it affects the productivity in subsequent stages in the textile industry, but also because it has a significant impact on the appearance of finished products. The uniformity of the spun yarn can be measured and compared based on short and long-term variations in mass per unit length, variations in strength, variations in blend proportions etc.

The processes prior to spinning are called preparatory stages. It is the purpose of the preparatory stages to prepare the feed material by parallelizing the fibers, and to make the feed uniform. This improves the efficiency and quality of the spinning process. The uniformity of the feed material is achieved through a process referred to as auto-leveling. Auto-leveling is the name of the technology used to control the mass variations in spinning preparatory- machines. The apparatus for auto-leveling is typically used in the carding and drawing stages to control the hank of the sliver delivered.

The processes prior to spinning are called preparatory stages. It is the purpose of the preparatory stages to prepare the feed material by aligning all the fibers in a parallel direction, and to make the feed as uniform as possible to improve the efficiency and quality of the spinning process. In this respect, the auto-leveller drawframe plays a very important role.

The schematic of a micro-leveller system is given at Figure 1. The system includes displacement sensor 20, the scanning rollers 21 , accelerometer 22, dashpot 23, controller 24, feed rollers 25, middle rollers 26, delivery rollers 27, and the delivery funnel with sensors 28 as indicated in the diagram. The elements 20, 21 , 22, and 23 together allow the direct estimation of the change in thickness of the feed even when the speed of the feed is varied over the operating range of the machine. The middle rollers 26 and the delivery rollers 27 have sensors attached to them that communicate their respective displacement to the controller 24 during the drafting process. The controller 24 also gets inputs from displacement sensor 20, accelerometer 22 and sensors 28 at precisely synchronized positions of the middle rollers 26 and the delivery rollers 27. The controller is responsible for storing user input data, calibration and cross calibration of sensors, and the synchronization of measurements, computations and corrections. The controller 24 may also act as a servo-control device, or interact with a dedicated servo controller that implements the motion control of the drafting rollers. The feed rollers 25 and middle rollers 26 usually have a fixed ratio of speeds depending on the break draft selected by the user.

The correction point as indicated previously is typically within unit fiber- length away from the delivery rollers. In the methods and apparatus disclosed, the correction point and the correction that need to be applied are determined by evaluating the number of leading fibre-ends in a feed. The figure 4 clearly shows that the distribution of leading-ends in the output is controlled when the draft is varied. Case 1 represents a high degree of draft being applied while case 2 represents a low degree of the applied draft. In case 1 , by the time the leading-ends at location 4 reach location 5, the nip of the delivery rollers 3 (running at a constant speed), the leading-ends at 5 would have reached location 6. This is achieved by reducing the speed of the leading-ends in zone 8 by slowing the speed of the middle rollers (not shown in the diagram). To reduce the draft, the speed in zone 8 is increased. The instantaneous draft can then be given by the ratio of the distances travelled by the leading-ends on the delivery side to the distance travelled on the feed side.

It is important to note that in each time step of control indicated by short vertical lines, the relative displacement between fibers takes place in a controlled manner only for the fiber-ends about to reach the delivery rollers. The fibers already gripped by the delivery rollers, and the fibers that will not be gripped by the delivery rollers in the current time step are unaffected by the variations imposed on the speed of the rollers by the leveling mechanism. Therefore it is of utmost importance that the draft applied depends on the number of leading-ends in the zone about to be gripped by the delivery rollers. The existing auto-leveling processes apply corrections based on the macro-level variations in the feed thickness. The feed thickness is determined by measurements taken at regular sampling intervals. However, thickness is actually a complex sum over all the fiber-ends distributions in the vicinity of the measurement. Such fiber-ends distributions are infinite in number for the same sequence of measured thickness. Therefore, such a method lacks correlation between the measured thickness and the number of leading-ends in any correction zone. In the absence of such a correlation the thickness measurement is ineffective in determining the draft. The "All or None" type correlation of Figure 5 corresponds to non-stationary and stationary cases respectively. In the stationary case, when any measured portion reaches the predetermined location of correction 10, there is no correlation between the measured thickness and the number of leading-ends presented to the delivery rollers at that instant in zone 8 (or any other zones such as 11 or 12 either), because of the extra degrees of freedom in the arrangement of fibers. However, in case of non-stationary variations in the feed, because of the shift in average thickness, the leading-ends in all the zones within 13 are likely to correlate with the measured thickness. Therefore, except during non-stationary variations in the feed, in general, there will not be any correlation in any of the zones with the measured thicknesses. This is the drawback of the existing systems. In other words, for the same set of thickness measurements a different set of corrective actions may be appropriate, except during non- stationary variations. This lack of correlation between thickness and correction needed can be explained with the help of a simplified example. Assume that there is a uniform feed made up of 30 mm fibers. On this is placed a bunch, nearly 30 mm long, of a well-aligned arrangement of fibers that make the thickness double over this 30 mm length. The present systems will apply multiple corrections corresponding to each measurement of higher thickness over this 30 mm length. However, the proper way to correct the feed, as indicated by the method of the current invention is to apply a higher draft only once, when the extra fibers reach the delivery rollers.

Many approaches have been taken to improve the auto-leveling system. The two parameters of the intensity of correction and the point of correction are not sufficient to fully determine the correction needed that is to be applied to the feed. Additional information using techniques such as Fourier analysis also is not useful because these provide information that is not local to the point being corrected. Even wavelet transforms are unsuitable when the variations in the feed are truly random. The present invention differs from the known auto-levellers. The invention, inter alia, accurately tracks the variations in the feed at the micro- level so that the appropriate corrections can be applied at the optimal correction point. The following paragraphs explain the manner in which the present invention is enabled. The method of micro-leveling is based on a mathematical model of the fiber feed that explains the manner in which the variables in the said model affect the variations in the thickness of the feed. It is obvious that each fiber can contribute to the thickness of the feed only to the extent of the length of the fiber. It should be noted that while the length of all individual fibers in the feed is difficult to predict/calculate, the fiber-lengths distribution of the said fibers are likely to be statistically similar at any location in the feed. Therefore in case the lengths of a large number of fibers are actually measured, it is likely that the distribution of fiber lengths in a cross-section at any given point on the feed would not significantly vary from the measured distribution. In other words, even if the individual fibers lengths vary significantly, the mean of fiber- lengths and the standard deviation of fiber lengths in different cross sections of the feed will not significantly vary. Similarly, the average number of folded fibers, the average extent of folding, the average inclination of fibers, and the averages of other configuration-properties of the fibers will not vary significantly over the length of the feed. In the instant case, therefore, the emphasis is given on the variation in leading ends of the fibres to account for the variation in thickness. The number of the leading ends of the fiber would be less as compared to the number of fibers present across a given cross-section. It should be noted that for a given distribution of leading fiber-ends, the thickness of the feed would vary with length in a manner that is characteristic of the average properties of the fiber-distribution in the feed. Conversely, it is possible to compute the leading-ends distribution along the length of the feed from the feed thickness and the average properties of fiber-length distribution.

In a system where the leveling action is regulated by varying the speed of the middle roller while keeping the speed of the delivery roller fixed, it is clear that the fibers gripped by the delivery rollers can no longer be controlled. Conversely, best leveling can be achieved by ensuring that the number of new fibers (leading-ends) presented to delivery rollers per unit time is constant. This will ensure that there are equal numbers of leading-ends per unit length on the delivery side. This in turn will ensure that the thickness on the delivery side is uniform as indicated by equation (1).

Figure 2 represents a scanning roller system of a typical micro-leveller system. The scanning roller system behaves in a manner similar to a mass- spring damped harmonic system. This scanning roller system is a system possessing a single degree of freedom with a random forcing function. This means that the system is acted on by an external, variable, non-periodic force. The system is designed to be "stiffness-dominated" i.e. the spring-mass system would not harmonically oscillate or continue to oscillate, if the scanning roller is displaced by the variations of the incoming feed. The requirement of the spring-mass system as depicted in Figure 2 is that the acceleration measurement should show linear behaviour at frequencies in the range of about 1000 hz. Accelerance is given by:-

Accelerance =

where y=v-4 The spring constant is k, dashpot coefficient c, mass m, and frequency is ω. The forcing function energizes the system, while the dashpot dissipates the energy.

When the spring effect is dominated (ω² » k/m), and the damping can be neglected, accelerance « 1/m. This means that beyond a certain frequency, the measured acceleration will not be affected by frequency. This is an essential requirement because levelling mechanism will vary the speed at which the feed moves across the scanning rollers. The lowest speed at which the error in measured acceleration remains within tolerance will determine the maximum applicable draft in the system. Since the measurements of interest are in the range of about 1 khz, the k/m value has to be selected with that in mind. Also, the measured data should be suitably filtered to eliminate frequencies outside the above range.

It should be noted that in order to calculate and apply the correction appropriately it is important to correlate the displacement and acceleration of the scanning rollers with the leading fibre-end distribution. The displacement of the scanning rollers is caused by the force exerted on the spring-mass system of the scanning rollers. The variation in the force is a direct result . of the variation of the thickness of the sliver in the feed wherein the variations are both due to the presence of the macro as well as the micro level variations. The acceleration measured is in response to the force exerted by the feed material which is correlated to the change in number of fibers in the feed. The acceleration of the scanning roller system is measured in conjunction with the displacement of the scanning rollers.

The main properties of relevance are the bulk modulus, B, of the feed and the specific volume C of the fibre. The bulk modulus is given as

B =Δp/(Δv/v) where Δp is the pressure needed to compress the material at volume, v, by Δv. The specific volume C = Δv/Δn relates the change in volume of the compressed feed to the change in number of fibers, Δn, in the cross section. These two equations, along with the relationships between force and acceleration (f = ma), pressure and force (p = f/area), and volume and area of cross section (v = area.d), are combined to relate the measured acceleration to Δn.

Δn = m.a.d / (B.C) where m is the mass of the spring-mass system, a the acceleration, and d, is the displacement measurement of the spring. It would be noted that by directly obtaining the variations in thickness rather than the thickness, the accuracy of the measured Δn value is greatly improved. Also, the acceleration measurement has a much higher bandwidth than displacement measurement, and this is essential to monitor variations at the micro level.

An equation relating leading-ends distribution to the thickness of the sliver/feed is essential to properly process the data collected. In this respect it is beneficial to view the data collected, the thickness, and leading-ends per unit length as a set of time-series. The contribution to thickness by a single fiber over the time-series representing the sliver depends on length of the fiber, its fineness, its orientation, its folding configuration, etc. However, given that there are large numbers of such fiber-leading-ends in the unit length of interest, the average total contribution to the thickness (of the sliver), of all these fibers together at specific locations from the starting position (of the fibers), will not vary significantly across the entire length of the feed. Furthermore, this relationship between position and contribution to thickness is a property of the feed that depends on the distribution of fiber lengths, fineness, orientation, etc. in the feed. This relationship is analogous to the impulse response of an electronic filter, and the thickness time-series, t(n), can be interpreted as the convolution of this response, h(), with the leading-ends time-series, x(n).

t(n) = ^j="∑_j=o hG) x(n-j) (1) Δn measured above directly relates to t(n), and thus through deconvolution to x(n) as follows:

Δn = t(n) - t(ri-1) =^y=%₀ h(j) x(n-j) - ^y=%₀ h(j) x(n-j-1) (2)

From this one can derive the following: x(n) = {Δn 4>%₀ h(j)

h(j) x(n-j)}/h(0) (3)

From the above equation it can be seen that the current values are determined from previous values. This equation therefore needs to be initialized with the correct values at start up. For this purpose, the thickness of the drafted material at the point where the material just leaves the front delivery rollers before the drafted web is condensed into a sliver is also measured. This can be done using any standard measuring technique such as the capacitance measurement method. This information together with the Δn measurements is used to determine the desired fiber-ends distribution as shown below. In order to explain how to determine the fiber-ends distribution we take the simple case of a parallel arrangement of fibers for illustration. However, the method is applicable to the general case also as shown later. Let the fiber- length distribution in the feed be k(0), k(1) ...k(L), where k(L) is the percentage of fibers of length L in the feed. From a percent cumulative distribution of the above, the impulse response that would be characteristic of the feed is computed and is shown in Figure 6. From the figure it can be gathered that a single fiber's contribution to thickness in a unit length as a function of distance from its own leading-end is given averaged over the entire population of single fibers. It is possible for the contribution to be more than unity, as in the case of inclined fibers or folded fibers. Different impulse responses would correspond to different length distribution and/or configurations of the fibers in the feed. The impulse response is just a weighting factor consolidating into a single function of position, all the physical and conformational properties of the fibres that relate to the thickness of the feed. The thickness at any location in the feed is influenced by fibers whose leading-ends are up to a fuii fiber length away. The total thickness at any location is a summation over the weights in the impulse response scaled by the number of leading-ends in each corresponding zone. Let the actual distribution of leading fiber-ends in the feed, the sequence needed for proper leveling, be given by the time series x(n) = a_o, a₁, a₂, ... a_n (4) and let the impulse response of the above fibers be given by h(n) = h₀, 1I₁₁ R₂, ... h_L (5)

The thickness t(n) is then related by a convolution of the leading fibre- end distribution and the impulse response of the fibres as t = h * x (6)

Applying z-transform to the above equation we obtain, T(Z) = H(Z) X(Z) (7)

Similarly the z-transformed x(n) and h(n) are as follows:

H(z) = h_o+ h₁z+ h₂z²+ ... h_Lz^L (8)

X(z) = a_o+ _aiz+ a₂z²+ ... a/ (9)

The Z-transform converts a discrete time domain series, which is a sequence of real numbers, into a complex frequency domain representation.

Let us denote by x_d(n) the leading-ends sequence on the delivery side.

Let us assume further that a constant integer-valued draft, D, is applied. Then, assuming that the fiber-ends are distributed uniformly within a single segment, ^x _d(^z) = [ (^a6^{+ a}o^{z+ a}o^{z +} — ® ^times) ⁺ (^ai ^{z + a}i ^{z + a}i^{z +} - D ^times) ^{+ • • •} + (a_n z^nD+ a_n z^nD+1+ a_nz^nD+2+ ... D times) + ...}/ D (10)

The thickness of the delivered sliver is then given by

T_d(z) = H(z) X_d(z) (11)

Let us now assume that the x(n) values are not known accurately, so that the values known are given by the time series

Let a_o= α₀ + ε₀; a = Q₁ + ε.,; a₂= α₂ + ε₂ ... etc. , (13)

The ε₀, E₁, etc. represent the error in the estimates of the leading-ends, which because of the nature of the feedback filter of equation (3) will lead to errors in the future estimates also. These errors in turn will lead to errors in the subsequent estimates and so on. Conversely, any correction to reduce the error in these will lead to improvements in the estimates of future values of fiber-end distributions.

The difference between the actual time series of delivered sliver thickness T_d(z), and the expected (erroneous) thickness, t_d(z), is given below: T_d(z) - t_d(z) = H(z) [ε₀ (1 + z+ z + ... D times) + E₁ (z + z + z + ... D times) +

LD LD+1 LD+2

... + ε_L (z + z - + z + ... D times) + ...] / D (14)

Assuming that the measured value of thickness on the delivery side is used to get T_d(z), and the value of ϊ_d(z) is computed from the convolution of drafted x(n) values, the above set of equations (obtained by matching the coefficients of the powers of z in equation (14)) can be solved by regression to get the values of the error terms, ε₀, E₁,... ε_L . These may then be used in equations (13) to better estimate the correct values of the initial values of the fiber-ends time series. With these new initial values, equation (3) is used once again to re-compute the yet to be corrected portions of the feed. The system can continuously correct itself in this manner. It may also make corrections to the estimate of the impulse response H(z) periodically using the same technique.

The above set of equations (14) using standard techniques can be described and solved as a matrix. Though equation (14) has been shown for a constant integer draft, the method can also handle variable drafts with non- integral values in the matrices. This can be done by calculating the weighted average for all the elements in the matrix, every time the draft changes in the middle of a measured segment.

Once the distribution of leading-ends is estimated, it can be ensured that the leading-ends reach the delivery rollers at a given constant rate, using techniques described later.

The method and apparatus of the above invention were simulated in a computer, assuming: a) a randomly selected set of fiber lengths for a maximum fiber length of 30 mm and ranging from 19 to 30 mm; b) the distance of the scanning roller to the correction point is about 1 m; c) speed on the feed side is about 1 m sec^"1; d) the recomputed initial values should be available by the time the first 30 mm of fiber have been drafted; e) Readings of thickness and change in thickness are made every 1 mm; f) The feed consists of about 150,000 fibers in the cross section; g) There are on average 5,000 leading ends per mm of feed, with a stationary random variation of plus or minus 15%; h) There is an error in measurement of plus or minus 3%; i) The mechanical system is capable of producing the desired changes in roller speeds; j) The feed is considered to be made of a parallel arrangement of fibers for the computation of the impulse response.

The above assumptions are good approximations to the conditions that would exist when the machine would actually be running. The results shown in Figure 3 demonstrate the superior correction achieved by the invention compared to the similarly simulated results for 1) no auto correction and 2) auto correction using conventional auto-leveling techniques. Typical values of 1 m cv% obtained are of the order of 0.02 as against the usual 0.5% in today's machines.

The invention can also be used for other optimizations in the auto- leveling process. For example, one may try to ensure uniform distribution of trailing-ends, rather than leading-ends, considering the fact that these trailing- ends will be the leading-ends in the next stage in the process. Such changes will only involve some minor modifications to the basic method described here but would still be within the scope of the present invention.

The invention can also be used in situations where the fibers are not all aligned parallel to each other. This factor can be incorporated into the algorithm by first using two different impulse response functions, one for the feed and a different one for the delivery (because of the tendency for parallelization after drafting). These impulse responses can be periodically estimated using the multiple regression in the same manner as above by rearranging the matrix equation generated from equation (14).

The present invention may be implemented with many combinations of hardware and software. If implemented as a computer-controlled apparatus, the present invention is implemented using means for performing all of the steps and functions described above. The present invention can also be included in an article of manufacture (e.g., one or more devices) having, for instance, sensors and controllers. The controller has embodied therein, for instance, machine readable program code, means for providing and facilitating the mechanisms of the present invention. The article of manufacture can be included as part of a machine, computer system or sold separately. The system can be used to process a wide range of natural and synthetic materials including, but not limited to cotton, wool, polyester, viscose, acrylic, etc.

Along with the procedure described above for the calculation of the leading fibre-end distribution, a number of practical aspects need to be addressed to make the realisation of the above outlined process feasible. These practical aspects include synchronized measurement, calculation, and control. The process described in the above section had access to the initial values, data obtained at different instants, at different locations, by different types of devices, and yet it was assumed that the data could be perfectly matched, both for analysis and correction, to obtain the desired result. It was also assumed that the corrections applied could be effected instantaneously by the servo control system. The techniques used to make the above described system work essentially as described, will be disclosed. Synchronization

The process for the auto-levelling described above requires that the data from the two sensors are synchronized with the applied speed changes in the middle roller. This requires accurate estimates of the time delay from the instant when the material is sensed by the scanning rollers, to the instant when it is about to be gripped by the delivery rollers. This information is required for every segment, and the actual delay needed for each segment may vary considerably because of the speed variations introduced by the levelling mechanism. Likewise, the time taken for the material to go past the delivery- side sensors could also vary. These delays are further dependent on the roller settings used, and this factor has to be taken into consideration.

For the typical drafting system in which a fixed break-draft is applied one encoder/resolver on the middle bottom-roller, and another on the front bottom-roller is installed. While the distance from the scanning rollers to the delivery rollers is fixed, the distance between the middle and front rollers could vary depending on the settings used. Likewise, the distance from the feed roller to the middle roller could also vary. But, once the machine is prepared for processing, these settings are not disturbed for the duration of the run. Thus synchronization is achieved using the following procedure after every setting change. Synchronization Procedure

The process for synchronisation begins by first setting the machine for a low, fixed draft of, say, 6. Cut one sliver just before the scanning rollers. Disable the stop motions, and run the machine until the cut sliver completely clears the delivery-side sensors. The system controller is designed to sense the large deviation in thickness (because of the cut sliver), and record the instantaneous middle roller encoder readings as the material enters, first the scanning rollers, and subsequently the delivery-side sensors. The above procedure is repeated for a high, fixed draft, say, 10. The data collected is used to compute the parameters needed for synchronization as shown below.

Let XL and X_H be the revolutions of the middle roller needed for the material to reach the delivery-side sensor,, from the scanning rollers, for low and high drafts, respectively. Let Y be the revolutions of the middle roller needed for the material to reach the front roller nip, from the scanning rollers. Let Z be the revolutions of the front roller needed for the material to reach the delivery-side sensor, from the front roller nip. Let the draft between the middle rollers and the delivery rollers be DL and D_H, for the two cases. Then, (assuming same roller diameters), X_L = Y + Z / D_L (15)

XH = Y + Z / D_H (16)

From the above two equations, the values of Y and Z can be estimated.

Y represents the time delay, in revolutions of the middle roller, between the measurement at the scanning rollers and the application of the corresponding correction. Similarly, Z represents the delay between the front rollers and the front sensors in revolutions of the front roller.

All the precision and sophistication of the above methods are not very useful if the leading-ends that were at the same transverse line as the material went past the scanning rollers do not reach the drafting system at the same instant. This can happen because the slivers on the feed-side tend to fan-out after the compressive forces at the scanning rollers are released. As a result, the distance travelled by the slivers in the middle is less than the distance travelled by the ones at the sides, to reach the feed rollers. The few millimetres of difference could introduce significant errors in the Microleveller computations. Hence, a simple arched bridge type off device needs to be introduced on the feed side between the scanning rollers and the feed rollers to increase the path length of the middle slivers. Motion Control

The recent developments in the field of Programmable Motion Control make it possible to regulate speed changes with almost a 1 khz bandwidth. These systems work with encoders/resolvers to achieve the desired speed profiles. These have built-in support for achieving smooth changes in speed. The Microleveller algorithm will calculate the leading-ends in each correction zone, estimate the encoder positions and possibly, the speed change moves for each zone, and request the motion controller to execute the moves. Calibration

The process for correcting the stationary variations as outlined above works using the number of fibres in the cross section, in its calculation. However, the two sensors in the system do not directly measure the number of fibres. The acceleration measurements and the capacitance measurements have to be converted to corresponding values for the number of fibers, and then to the number of leading-ends per unit length. The tendency for variations in the capacitance measurements with varying humidity will have to be accounted for. The two sensors will have to be cross-calibrated to ensure that there is no scaling error between the sensors. The delivery-side sensor can be easily calibrated by using a cut and weigh method for a known length of sliver that has been independently measured using the capacitance sensor. The average capacitance value can be related to the average mass per unit length, which can then be converted using the denier/micronaire value to the number of fibers in the cross-section. This information, together with the fiber-length information can be used to compute the average leading-ends in the feed material. The cross-calibration between the sensors is achieved by the method described below.

The Microleveller has two independent ways of determining the thickness of the output material. One method uses the delivery-side sensor. The other uses the feed-side sensor data to compute the same using convolution of the drafted material. The two methods produce two synchronized time-series representing the same property. Therefore, the statistical properties such as the average and the standard deviation should be the same for the two time-series. The scaling needed to match the series is determined by the ratio of the standard deviations.

It is convenient to view the two series mentioned above as a Cave-plot. One of the series is displayed as stalactites while the other is displayed on the same time axis as stalagmites. This allows the users to evaluate the efficacy of the Microleveller algorithm in the tracking of the leading-ends. The user may use the plot (computed and refreshed at periodic intervals), to monitor the progress of the levelling function. It may also be convenient to provide a simple phase correction mechanism in the programmable motion control system that might be controlled manually while continuously monitoring the Cave-plot. Data Collection

As stated earlier, the data collection, the computations, and the control with variable delays have to be precisely synchronized. This requires that proper book-keeping is done by the Microleveller algorithm.

Assuming that the data from the scanning roller and the delivery-side thickness data are collected every mm, and the change in encoder readings corresponding to these are E_m and E_d, for the middle and delivery roller encoders, respectively, the following tables of data need to be collected. Table below is populated for every E_m change in the middle encoder.

The above data is based on feed-side measurements and computations using that data. The first column is the middle encoder reading when the material sensed is at the scanning rollers. The leading-ends data and the required draft are computed for this material, and the middle encoder position when the correction should be initiated is recorded by adding Y to the first column. The last column is the computed value for the change in the delivery encoder position as the middle encoder traverses to the next 'correction start' position. This effectively determines the instantaneous draft to be applied. Data from this table is used to drive the programmable motion controller. It may be preferable from a system design perspective to vary the speed of the delivery-side systems also (including the coiler), in addition to the feed-side rollers so that the desired relationship between the two sets of rollers can be maintained with lower capacity motors than would otherwise be needed.

Table below is populated for every E_d change in the delivery encoder, starting from Z revolutions of the front roller after the middle encoder reaches the correction start point of the previous table.

The delivery-side data recorded in the above table is used to compute the error in the estimates of the leading-ends by multiple-regression as explained earlier. It may be sufficient to carry out this feedback computation intermittently. Also, during the initialization of the Microlevelling, this feedback data may be used to correct the estimated impulse response of the feed material. It would be noted that even though the simulations were carried out under certain assumptions, these are not limitations of the Microleveller system, but just the limitation of the simulation method used so as to exemplify the invention. For example, the Microleveller system can work even in non- parallel arrangement of fibers in the feed, and even when there are leading/trailing hooks in the feed. These effects are easily accounted for by changing the shape of the impulse response function. The working of the micro-leveller will be similar to a typical auto-leveller, except for the initialization procedure after a setting-change/change in feed material and is described in an exemplifying manner through the following steps:

1) input of data in relation to the fibre properties, break draft and machine settings, feed hank etc.

2) feeding material at high and low draft for collecting synchronization data. The measurements made by the delivery-side sensors can be used for automatically calibrating the sensors using the fibre properties entered by the user in step 1. 3) data can also be collected from the accelerometer and the displacement sensors. This data is processed for cross calibration between the accelerometer and the delivery-side capacitance sensor. This is the only time in the run the accelerometer is cross calibrated against the capacitance sensor. Subsequently, the accelerometer is used to cross calibrate the capacitance sensor.

This is because of the tendency of the capacitance sensors to drift during the course of a run.

4) the user input data is used to compute the initial estimate of the fibre impulse response. The initial values for the number of leading-ends are guessed to be the average value. The machine is now ready to compute the actual leading-ends.

5) scanning roller data and the delivery-side data is collected, and the error values for the initial value for the leading-ends are computed by multiple regression. This is used to correct the initial values of leading-ends.

6) Equation (3) is now used to compute the subsequent values of leading-ends. Using these new values, the coefficients of the impulse response are determined by multiple regression. Using the new impulse response, the error in leading-ends is computed

7) This process is repeated till the system stabilizes on a proper impulse response and characteristic of the feed.

8) The drafting system now computes the leading-ends in the feed from the scanning-roller displacement and acceleration data. The system should be set up to get feedback on the correction from the delivery- side sensors. Whenever the variations exceed certain limits, the leading-ends and the impulse response estimations should be repeated. The delivery-side sensor should also be monitored periodically to prevent drifting.

9) The leading-ends computation is used to send the control inputs to the motion control system to achieve the desired speed changes to the middle rollers. The object of the speed change is to ensure that the leading-ends reach the delivery rollers at a constant rate. It should be noted that the steps outlined above which are essential for the practical working of the invention are only indicative. The calibration and the synchronisation can be achieved through other means and such methods if carried out with the Micro-leveller process disclosed would still be in the scope of the invention.

Claims

I/We claim:

1. A method for regulating the drafting of a fiber mixture of a textile using properties of the fibres making up a feed, said method comprising : a. detecting thickness, and variation in the thickness of the feed as a function of position or time series or the combination of both, b. computing the average properties representative^»of the entire feed, as an impulse response and the arrangement βfc properties local to the point of measurement, so as to derive a relationship between an input and an output; c. computing the time series of corrections needed to minimize the thickness variations in the output using the above; and d. applying corrections.

2. The method as claimed in claim 1 , wherein the variations are stationary and/or non-stationary.

3. The method as claimed in claim 1 , wherein the variations include variations at the micro-level of the feed that do not affect the statistical properties of the feed.

4. The method as claimed in claim 1 , wherein the detection of the variations is correlated to stationary and non-stationary variations in the feed thickness.

5. A method as claimed in claim 1 , wherein the variations in the feed at the micro-level are computed, so as to apply the appropriate corrections at the drafting rollers.

6. A method as claimed in claim 1 , wherein leading fiber end distribution results in the variations in the feed.

7. A drafting system for fibres with devices for measuring, computing, controlling and correcting the stationary and/or the non-stationary variations in the feed thickness of the incoming fibre, comprising: a. damped spring mass system consisting of i. at least one pair of scanning rollers; ii. a dash pot; iii. an accelerometer; and iv. displacement sensor; b. feed rollers; c. middle rollers with position sensors; d. delivery rollers with position sensors; e. delivery-side sensor; f. motion controller; and g. controller to compute and co-ordinate all measurements and corrections.

8. The system as claimed in claim 7, wherein leading fibre-end distribution causing the variation of the thickness of the feed, is computed from the data as the force exerted on the scanning rollers by the feed.

9. The system as claimed in claim 7, wherein force exerted on the scanning roller is measured in terms of acceleration.

10. The system as claimed in claim 7, wherein the scanning roller system is designed to have near constant accelerance.

11. The system as claimed in claim 7, wherein data computed from the scanning rollers and the delivery-side sensors are used to compute time-series of the corrections needed.

12. The system as claimed in claim 7, wherein variations in the fiber-length distribution, the degree of orientation of fibers, presence of hooked fibers, variations in fiber fineness is/are accounted for in the computation(s).

13. The system as claimed in claim 12, wherein the thickness of the output can be evaluated for any given time-series of data from the scanning rollers and the time-series of corrective actions at the drafting rollers.

14. The system as claimed in claim 13, wherein the time-series is compared to the delivery-side measurements to make the whole system self- correcting.

15.The system as claimed in claim 14, wherein the measurements from the sensors are cross-calibrated.

16. The system as claimed in claim 14, wherein the measurements and corrections are synchronized.