The usual way of making dual density products is by providing a continuous mineral fibre web which contains binder, separating this web depthwise into upper and lower sub-webs, subjecting the upper sub-web to thickness compression so as to increase density, rejoining the sub- webs to form an uncured batt and then curing the binder to form the cured batt. The upper sub-web thus provides the higher density upper layer intermeshed with the lower density lower layer.

Typical disclosures of conventional dual density processes are given in, for instance, W088/00265 and US-A- 4,917, 750. In each instance the web which is separated into sub-webs is a web as formed initially on a conveyor.

As shown in W088/00265, the web may be formed by cross lapping. As shown in both of those specifications the web is passed under some rollers as it approaches a device for separating the web into upper and lower sub-webs.

If no lengthwise compression is applied to the web before the separation, the fibres in the web will be substantially oriented parallel to the conveyor, because this is the predominant orientation during normal fibre lay-down processes. However in EP-A-1, 111, 113 the web is subjected to longitudinal compression before it is separated with the result that the fibres no longer have an orientation substantially parallel to the conveyor but instead have an orientation which has either a macro vertical component (so as to give significant visible pleats as shown in Figure 2 of EP-A-1, 111, 113) or a micro configuration (in which the vertical reconfiguration of the

fibres has occurred but is not so visible to the naked eye, for instance as described in Figure 12 of EP-A-0,889, 981).

In all these processes the lower web is subjected to little or no treatment between the positions where the upper web is separated from it and then rejoined on to it.

This means that the ultimate performance of the totality of the product is dictated predominantly by the effect of the thickness compression on the upper layer and the structure of the web just before separation of the web into upper and lower sub-webs and by the effect of any post-treatment after the sub-webs are rejoined.

The thickness compression results in some length extension of the upper web. However as mentioned in EP-A- 1,111, 113 it is also possible to subject the upper web to longitudinal compression to compensate for the elongation of the upper web.

Unpublished research by us has demonstrated that the upper layer and the lower layer serve different but inter- related functions in providing the overall properties of the dual density batt, and that the properties of each layer are influenced significantly by the macro and micro fibre arrangements within each layer in the final batt.

Since the initial fibre orientation of the upper sub-web and the lower sub-web is the same this restricts the ability to obtain optimum properties. Thus a fibre arrangement in the initial web which is optimum for the lower layer may not be optimum for the upper layer, and vice versa.

Another disadvantage with this type of system is that if advantage is to be taken of longitudinal compression of the starting web, as in EP-A-1,111, 113, the overall apparatus is very lengthy because of the length associated with longitudinal compression of the thick initial web followed by the dual density separation, thickness compression and rejoining.

A process is described in W094/16162 in which upper and lower sub-webs are derived by separating an initial web

and are subjected to independent treatments before they are rejoined. Thus in Figure 1 one sub-web is subjected to pleating by longitudinal compression, optionally followed by thickness compression or length compression, while the other sub-web is subjected to cross lapping and then length compression and then thickness compression and/or more length compression. This process does allow for independent configuration of the two sub-webs and the attainment of a dual density product, but suffers from the inherent disadvantage that the major processing steps conducted separately on the two sub-webs necessitate extremely complex and lengthy apparatus.

Simpler processes, in which the. lower sub-web has the same fibre configuration as the initial web, are also shown in W094/16162 but these suffer from. the traditional disadvantage that the properties of the primary web may not be optimum for both the upper and lower sub webs.

We have now found that it is possible to make a very.' simple modification of a conventional dual density process so as to obtain an improved combination of product quality and apparatus simplicity. In particular we can obtain product quality at least as good as and often better than is obtainable by the, elongated production line of EP-A- 1,111, 113 but using an apparatus production line which can be significantly shorter. In particular, we find that it is possible to achieve a unique fibre orientation by this process as a result. of which improved properties, per unit weight of product, are obtainable, especially when compared to the product quality obtained by simple processes such as in W088/00265 and US 4,917, 950.

The invention broadly provides. a continuous method of forming a bonded. mineral fibre batt comprising an upper layer intermeshed with a lower layer having a lower density than the upper layer in which each layer is a bonded non- woven fibre network, wherein the method comprises providing a continuous mineral fibre web which contains binder, separating the web depthwise into upper and lower sub-webs,

subjecting the sub-webs separately to lengthwise compression, and subjecting the upper sub-web to thickness compression before, during or after longitudinal compression and optionally subjecting the lower sub-web to thickness compression generally before the lengthwise compression, whereby the upper layer of the batt has higher density than the lower layer, and then rejoining. the sub- webs to form an uncured batt wherein the upper sub-web provides the upper layer of the batt, and curing the binder. Optionally either or both sub-webs may also be subjected to lengthwise stretching.

The upper sub-web is subjected to a very much greater thickness compression than the lower sub-web to give the required higher density (and indeed it is not absolutely essential for the lower layer to be subjected to any thickness compression), and some or all of the thickness 'compression on the upper sub-web is usually after longitudinal compression. As a result, the effect of the lengthwise compression on the two layers leads to very different fibre orientations in the two sub-layers even though the lengthwise, compressions nominally may be substantially the same.

Usually the two sub-webs are subjected to substantially the same lengthwise compression, and they have substantially the same speed of travel when they are separated and when they rejoin. Minor differences in speed just before, they rejoin can be tolerated provided that any resultant tension in other or both sub-webs when they are rejoined is so low that there is no distortion or delamination of the batt. However they may be subjected to different lengthwise compressions and/or either or both may be subjected to lengthwise stretching. For instance one- sub-web (e. g. , the lower sub-web) may be subjected only to lengthwise compression and the other may be subjected to more lengthwise compression followed by lengthwise stretching.

Preferably the path lengths of the two sub-webs are not significantly different. For instance if the path length of the sub-webs are different, the path length of the longer (usually the upper) sub-web is usually not more than 50%, preferably not more than 30% and most preferably not more than 15% longer than the path length of the lower sub-web, between the separating and rejoining points.

In some embodiments there may be benefit in having a significant vertical component in the fibre orientation at the time of separation of the web into the upper and lower sub-webs, as a result of longitudinal compression of the total web prior to separation. Best results, however, are obtained when the web which is separated has its fibres substantially oriented parallel to the surface of the web.

By this. we mean that the fibres in the web have the traditional essentially horizontal configuration which is typical for mineral fibres collected by an air-laying process, without any deliberate longitudinal compression or other vertical rearrangement of the fibres. Naturally the lay-down is not wholly horizontal, but the predominant orientation is clearly visible to the naked eye as being essentially parallel to the surface of the web.

The web at this stage may be a web formed by direct collection of mineral fibres by air-laying to the desired thickness or it may be a web formed by laying several such primary webs on one another or, more usually, by cross lapping a primary web to form a web of the desired thickness, optionally followed by mild thickness compression.

The web is then separated depthwise into upper and lower sub-webs in conventional manner by a knife or other splitting device which is usually. arranged substantially horizontally at a desired spacing above a conveyor on which the web is carried continuously. The positioning of the separating device is chosen to provide the appropriate relative thicknesses of the upper and lower webs. The thickness of the upper sub-web, at the time of separation,

is usually from 10 to 90% of the thickness of the total web. Usually it is at least 20% and often at least 30% of the total thickness, because the upper web is usually subjected to very high thickness compression and requires adequate thickness after this. Generally the upper sub-web is not more than about 70% or, at the most, about 80% of the total web thickness because usually it is required that the lower layer has sufficient thickness and structural content to impart significant properties to the final product.

Throughout this specification we are using the terms "upper"sub-web and layer and"lower"sub-web and layer in their conventional usage wherein conventionally a dual density batt is considered as having the higher density layer on its topmost surface. However the invention does, of course, include batts which are used the other way up and production processes in which the higher thickness compression is applied to the sub-web which is beneath the other sub-web, although in practice this is less preferred.

Also, it should be understood that although the invention is described wholly in. terms of upper and lower layers and upper and lower sub-webs the invention also extends to processes in which there are one or more other layers and corresponding sub-webs in the final product, wherein these other sub-webs may be subjected to the same or different thickness and/or lengthwise compressions as the upper sub-web and/or the lower sub-web. In particular there may be a higher density layer above the upper layer, for instance as described in WOOO/73600.

Because, in the preferred process, the web which is separated into upper and lower webs has the initial fibre lay-down orientation (substantially parallel to the web surface), optionally with some thickness compression, the apparatus required for carrying out the process does not have to include preliminary lengthwise compression apparatus, for instance as described in EP-A-1 ; 111, 113.

Instead, the entire apparatus can be confined within

approximately the space occupied solely by the thickness compression stages for the upper sub-web shown in that specification or, for instance, in US 4, 917,750 or.

W088/00265.

. The upper sub-web, and optionally also the lower sub- web, is subjected to thickness compression between the separating and rejoining stages. The extent of thickness compression can be indicated by the percentage reduction in thickness. It is possible to perform the process without any thickness compression of the lower sub-web but generally it is subjected to a thickness compression of at <BR> <BR> least 5% (i. e. , so that its thickness after the thickness compression is not more than 95%. of its thickness when initially separated from the upper sub-web) and is usually . at least 10%. Usually the thickness compression of the lower layer is not more than 60%, and preferably not more than 50%.

Preferably the actual thickness compression of the lower sub-web. is equivalent to about 0.5 to 2 times, most preferably about 0.7 to 1. 5 times the thickness of the upper sub-web at the time when it rejoins the lower sub- web. Typically therefore the extent to which the lower sub-web is subjected to thickness compression is such that its thickness is reduced by the thickness of the upper sub- web at the time of rejoining, so that the uncured batt formed by rejoining the sub-webs has the same, or substantially the same, thickness as the thickness of the lower sub-web when it was initially separated from the upper sub-web.

The thickness compression of the upper sub-web is always large, in order that this sub-web provides the required high density upper layer. Generally the overall thickness compression of the upper sub-web when it rejoins the lower sub-web is above 50%, preferably above 70% and most preferably above 85% (so that the final thickness of the upper sub-web is less than 15% of its thickness when initially separated from the lower sub-web. Usually, the

overall thickness compression is less than 97%, and most preferably less than 95% of the initial thickness.

The thickness compression of the lower sub-web (when this is applied) is preferably conducted, and usually completed, before the longitudinal compression of the lower sub-web. Preferably, however, significant thickness compression is applied to the upper sub-web after it is subjected to some or all of the longitudinal compression to which it is to be subjected. Thus normally the upper sub- web is subjected to at least half, usually at least three quarters and preferably substantially. all of its longitudinal compression and is then subjected to significant thickness compression.

The thickness compression which is applied after the longitudinal compression may be the only thickness compression which is applied to the upper sub-web but usually the upper sub-web is also subjected to thickness compression before the longitudinal compression ; Thus typically the upper sub-web is subjected to moderate thickness compression between separation and the longitudinal compression, for instance being reduced in thickness to from 90% to 30% of its original thickness, is then subjected to most or all of its longitudinal compression, and is then subjected to subsequent thickness compression which reduces the thickness of the sub-web to less than 50%, and usually less than 30%, of the thickness of the sub-web after the preceding thickness compression.

It seems that applying significant thickness compression to the upper sub-web after applying significant longitudinal compression is particularly beneficial to the final configuration and properties of the upper layer. The process can usually be optimised by subjecting the upper layer to substantially all the longitudinal compression before subjecting it to the final half or three quarters, or more, of the total thickness compression.

Although suitable thickness compression in the prior art is often achieved merely by the use of pairs of

rollers, the unusual stresses created on the upper sub-web by the preferred process of the invention are such that the thickness compression after the longitudinal compression is preferably achieved by passage of the upper sub-web between converging endless surfaces. These may be converging conveyors, or a conveyor and a plate which converge.

The lengthwise compression in each of the upper and lower webs should be at least 1.2 : 1 and preferably at least <BR> <BR> 1.5 : 1 (i. e. , the speed of the web leaving the lengthwise compression is not more than. two thirds of the speed of the web entering the lengthwise compression stage). It is generally not more than 5: 1 and often not more than 3: 1.

Each longitudinal compression can be achieved in conventional manner by passing the relevant sub-web from one set of conveying surfaces. (which may be belts or rollers) to a second set which are travelling slower. For instance the upper sub-web may be passed from a series of rollers or belts travelling at one speed to the converging passage between two conveyors which are travelling at a slower speed (so as to cause lengthwise compression followed-by thickness compression). The lengthwise compression of the lower sub-web can be achieved by passage from rollers or converging belts that provide thickness compression to a set of rollers or belts which are moving slower and which are parallel to. one another so that they do not provide thickness compression.

Although there is uncured binder in the upper and lower sub-webs, and this may be sufficient to achieve adequate integrity of the final batt, it is generally preferred to apply additional binder at the interface between the upper and lower sub-webs where they are rejoined, so as to promote the integrity of the final batt.

The uncured batt is formed by pressing the upper and lower webs together with sufficient pressure to achieve intermeshing and integrity but preferably insufficient to cause thickness compression, because additional thickness compression at this stage is unnecessary, and indeed is

generally undesirable since it may impair the pronounced vertical fibre orientation which is preferably achieved in the lower layer.

The batt is then passed through a curing oven in order to cure the total binder in conventional manner.

The invention not only includes the process but also includes the novel apparatus comprising the means for separating the web into sub-webs, subjecting each sub-web independently to treatments selected from lengthwise compression and thickness compression and rejoining the sub-webs, and wherein preferably the apparatus is supplied with web direct from a fibre lay-down process or direct from a cross-lapping process.

The invention also includes the mineral fibre batts made by the process and batts having the structural characteristics of these. The preferred batts have an upper layer having a density of 100 to 300kg/m3, often around 120 to 250kg/m3. They have a lower layer which has a'density which is usually not more than 80% but'usually more than 30% of the density of the upper layer, often around 40 to 70% of the density of the upper layer. It is usually 50 to 150kg/m2. Usually the upper and lower layers in the final product have a thickness of 30 to 300mm. The lower layer is usually 25 to 275mm thick and is usually at least 75mm thick. Generally it is at least 50%, and often 75 to 95%, of the combined thickness of the upper and lower layers.

The mineral fibres may be any suitable mineral fibres such as glass, rock, stone or slag. The invention is of particular value when applied to mineral fibres obtained by centrifugal fiberisation, and in particular by fiberisation of a rock, stone or slag melt by a cascade centrifugal spinner.

We find that it is possible, by the invention, to provide a lower layer which has a unique structure relative to the structure of lower layers provided in other dual density processes and that this provides excellent support

for the upper layer with the result that the overall properties of the combination of the high density upper layer and the unique lower layer provide a product having exceptional properties. This is discussed in more detail below.

The Figures 1 and 2 of the accompanying drawings are each a diagrammatic side view of apparatus according to the invention in use in the process of the invention.

In Figure 1, a web 1 is supplied direct from a cross lapping system which, in turn, is supplied direct from the collector of a collecting chamber from a conventional cascade spinner for rock fibres. Accordingly the overall and predominant orientation of the fibres in the web 1 is substantially parallel to the upper and lower surfaces of the web. The web 1 may have been vertically compressed and has a thickness TW and the rollers 2 and 3, and all the associated components with those, are set at a spacing corresponding to TW. The web 1 enters the apparatus at speed VW.'' A separating knife 4 separates the web depthwise into an upper web 5 having a thickness TU1 and a lower web 6 having a thickness TL1. As shown, TU1 and TL1 are approximately the same, but they can be different. Lower web 6 passes between converging belts 7 and 8 driven by roller train 9 in the direction of travel of the web, as a result of which the belts 7 and 8 cause thickness compression of the lower web 6. As the lower web emerges from. the converging belts at the position 10 it has a thickness TL2 which, as shown, is about three quarters of TL1.

The web then passes through upper and lower roller trains 11 and then through upper and lower roller trains 12. Within each of the. roller trains, all the rollers rotate at the same speed to carry the web forward.

Longitudinal compression is achieved by roller train 12 rotating slower than the rollers 9 and thus the belts 7 and 8. If roller train 11 rotates at the same speed as roller

train 12 then longitudinal compression will be applied at position 10. If roller train 11 rotates at the same speed as roller train 9 then longitudinal compression will be applied at position 13. Often roller train 12 rotates slower than roller train 11 which rotates slower than roller train 9, in which event longitudinal compression is applied both at positions 10 and 13. The objective is that the speed of travel as the lower web 6 passes through guide' rolls 14 should be the speed VB of the final batt as it enters the curing oven 15, with the ratio VW: VB generally being at least 1.5 : 1.

The upper web 5 is carried between a conveyor belt 16 and its supporting rollers 17 and a converging belt 18 and guide rolls 19. As a result of this the thickness of the upper web 5 is reduced from TU1 to TU2. TU2 may be, for instance, one third of TU1.

Conveyors 16 and 18 and rollers 19 all travel at the same speed and the upper web 5 travels from them to between converging belts 20 and 21 driven, respectively, by roller trains 22 and 23. Rollers trains 22 and 23 all rotate at the same speed, and at a speed less than the roller trains 17 and 19. As a result, longitudinal compression is applied at position 24. The extent of this longitudinal compression is such that the speed of the upper web when it emerges from between the converging belts 20 and 21 is sufficiently close to VB that there will be no unacceptable distortions of the upper or lower layers when they are rejoined at 26 to form batt 29. Thus any stretching or compressing of either or both sub-webs, due to tension in either or both when they are rejoined, should be so low that there is no distortion or delamination of the batt 29.

The converging. belts 20 and 21 apply substantial thickness compression to the upper web whereby the upper sub-web 5, when it emerges from between the converging belts, has an ultimate thickness (after any relaxation which occurs) of TU3, where TU3 is usually well below half of TU2 and typically below one fifth of TU1.

The upper sub-web may then slide over a supporting plate 25 as it travels down to the position 26 at which it rejoins the lower web. Binder is'sprayed between the webs as they are rejoined, from applicator 27.

As is apparent, it is preferred that substantially all the lengthwise and thickness compression steps conducted on the sub-webs are conducted between planar surfaces.

The rollers 28 apply enough pressure to press the upper and lower webs together to form an intermeshed batt 29 but insufficient pressure to cause any significant thickness compression of it. The uncured batt 29 then passes into the curing oven 15 and is then cured and subjected to conventional post treatments, such as cutting into slabs of the desired size.

In Figure 2, the rolls 19 are arranged as separate sets 19a and 19b each of which is covered by a band and is driven. The roller train 17 has been divided into two sets, one set covered by conveyor 16a and the other set covered by conveyor 16b. Bands 16,18 and 19a operate together at the same speed, and bands 16b and 19b operate together at the same speed, which can be lower. Lengthwise compression therefore can occur at both 40 and 24.

In one typical process of the invention, the ratio of the speeds of the belts 7: rolls ll: rolls 12: rolls 14 and 28 is 3: 3: 0.9 : 1 giving length compression at 13 and stretching between 12 and 14. In a second typical process, the ratio is 3: 2: 0.9 : 1, giving length compression at 10 and 13 and stretching between 12 and 14. This results in the lower layer being more relaxed, with less risk of the product distorting out of a planar configuration.

Accordingly it can be desirable to subject the lower web to a plurality of length compressions.

As examples of the invention, products A, B and C were made using apparatus described above wherein the operating conditions were as follows: Value Product A Product B Product C TW 110mm 380mm 360mm

TL1 65mm 330mm 225mm TL2 60mm 185mm 185mm TU1 45mm 50mm 135mm TU2 6mm 6mm 10mm TU3 12mm 15mm 30mm TB 60mm 200mm 215mm VW 32m/min 6.9m/min 6. 3m/min VB 16m/min 2.3m/min 2. lm/min We have established that the orientation of the fibres in the lower layer is unique and that the attainment of this unique orientation results in the lower layer giving better support to the top layer of a dual density product and that the product has improved penetration resistance and performance than is achieved when the lower layer does not have this orientation, when all other conditions are the same. Thus, as a result of obtaining the unique orientation it is possible to obtain equivalent results with a lower fibre amount and/or better results with the same fibre amount, when the upper layer is unchanged.

Similarly, it is possible to obtain better results when using the same upper layer or equivalent results with an inferior upper layer.

The novel fibre orientation obtainable by the methods of the invention is also obtainable by other methods, and thus is another aspect of the invention.

In particularj in this product aspect of the invention we provide a dual density layer wherein the lower, lower density, layer is definable by its Kappa and Tau values in one or more cross sections, wherein these values are obtained by scanning examination of parts of each respective cross section through the thickness of the layer and Fast Fourier Transformation of the data.

In particular, in this product aspect of the invention we provide a dual density layer wherein the lower, lower density, layer is definable by its Kappa and Tau values in one or more cross sections, wherein these values are obtained by measuring parts of each respective cross

section through the thickness of the layer in a flatbed scanner like Hewlett Packard ScanJet 6100C. The product to be examined is placed on the scanner so that it fits on top of the scanner with the shortest distance perpendicular to the scanning direction, see drawing.

For the set-up of the scanner use was made of the scanner-software Desk Scan II with the following settings : Sharp B, and W. Photo, Resolution 120xl20dpi, and automatic adjustment of brightness and contrast. The scanned image (110mm x 270mm) was divided into a number of local windows in a pattern comprising 8 rows each with 33 windows of equal size (32x32pixels) in which the dominant fibre orientation was estimated using Fast Fourier Transformation.

As is known, a two-dimensional pattern, for instance of parallel stripes, can be expressed, by Fast Fourier Transformation, as a small number of dots and a complex two-dimensional pattern, such as a cross section of a mineral fibre network can be expressed by Fast Fourier Transformation as a large number of dots. These dots will be arranged in a pattern, which may be circular but more usually is elliptical.

The Tau value for the cross section is defined as the geometric mean of the ratio of the length of the ellipse to the width for each of the 33 local windows and thus a high value indicates a local well organised pattern (high consistency locally) and a lower value near 1 indicates that the pattern locally cannot be defined. The Kappa value,.. is an indication of the statistical distribution of the different angles at which the ellipse is arranged locally for different parts of the overall structure, which is being examined. A high Kappa value indicates a narrow statistical distribution of angles whilst a low Kappa value indicates a broad distribution.

A description of the principles of the Tau and Kappa values for cross sections through the thickness of mineral fibre networks is described in"S. Dyrbol, Heat Transfer in

Rockwool Modelling and Method of Measurement"Dept. of building and Energy, Technical University of Denmark and Rockwool International A/S. Ph. D-thesis, 1998. Reference should be made to that article for a description of how to examine a cross section and how to conduct a Fast Fourier Transformation on the result of the examination and calculate the Tau and Kappa values for the cross section.

Other relevant publications are Russ.,"Computer-Assisted Microscopy. The Measurement and Analysis of Images". Plenum Press, New York, 1990; Larsen and Hansen,"Orientation Analysis of Insulation Materials, A feasibility Studie for Rockwool International A/S". Department of Mathematical Modelling, Technical University of Denmark, 1997. IMM-TR- 2001-03 ; and Ersboll and Conradsen"Analysis of directional data for Rockwool A/S", Department of Mathematical Modelling, Technical University of Denmark, 1998. IMM-TR- 2001-04.

In every instance it is necessary to determine the Tau and Kappa values by taking the mean value of at least 5 separate determinations each consisting of 3 cross sections.

In novel mineral fibre batts of the invention there is an. upper layer having a density of 100 to 300kg/m3 intermeshed with a lower layer having a lower density than the upper layer wherein each layer is formed of a bonded non-woven mineral fibre network the fibre orientation of which is definable by the Tau and Kappa values derived from Fourier Transformation of scanned images of thickness cross sections of the layers wherein Tx and Kx are the Tau and Kappa values determined on the thickness cross section of the layers in the lengthwise production direction X of the batt, and Ty and Ky are the Tau and Kappa values determined . on the thickness cross section of the layers in direction Y which is perpendicular to the production direction X.

We have found that in the lower layers of conventional dual density products K, is always less than Ky and that K, is below 2, for instance 0.7 to 1.4. In the invention we

find that improved performance from the lower layer is achieved when Kx is greater than Ky, with the ratio Kx : ICY preferably being at least 1.3 : 1 and often at least 2: 1, for instance up to 5: 1.

Conventional products have Kx below about 1. 5, but in the invention Kx is preferably at least 2.5 and most preferably at least 3.

We find that in conventional-products T,, is always less than Ty but in the invention Tx is preferably above Ty.

In particular, it-is preferred that the ratio TX : Ty is at least 1.2 : 1 and usually at least 1. 5: 1 and is often as much as 3: 1 or more.

We find that Tx of the lower layer in conventional products normally has a value of 2. 6 or less but in the invention Tx is preferably above 3, and most preferably above 3.5. For instance it may be up to 7 or more.

As an example, a conventional commercial product from a competitor was determined to have a lower layer in which Kx = 0.7, Ky = 2.9, Tx = 2.4 and Ty = 3.8. In contrast, a. product made by the method described above had Kx 3.8, Ky 1.2, Tx 4.2 and Ty 2.6.

The upper layer of the commercial product had Kappa and Tau values in each direction substantially the same as the Kappa and Tau values of the upper layer of the product made by the present process, but the point load resistance of the products made by the present process was very much greater than the point load resistance of the commercial product. Although there was some difference in density and surface weight per unit area, the difference in point load resistance could not be explained by this and so, instead, can be attributed almost entirely to the benefits of. the novel fibre orientation in the lower layer.

We believe that the unique Tau and Kappa values obtainable in the invention are due predominantly to the significant lengthwise. compression applied to the lower layer independently of the upper layer, in combination with the relatively horizontal orientation of the fibres prior

to splitting. Accordingly, if Kappa and Tau values or ratios not within the preferred ranges are obtained in any particular process,. it is possible to achieve the desired results by varying the extent of longitudinal compression of the lower sub-web, the extent of the combination of this with thickness compression of the lower sub-web, and the extent to which the fibres in the web before splitting are substantially horizontally arranged and the extent to which they are arranged predominantly in the Y direction, i. e., transverse to the lengthwise production direction X.

The lengthwise production direction X can usually be determined by observing the pattern impressed on the upper and lower surfaces of the batt by the curing oven, when cured in conventional manner.

Best results are achieved when the lower layer has the fibre orientation described above and the upper layer has the fibre orientation described in PCT application..... reference. PRL04361WO claiming priority from European application 01310773.5 filed even date herewith.

The invention may be utilised for production of roof boards, facade boards or similar boards produced from bonded mineral fibres when a certain point load resistance is required. They may be used generally for thermal insulation, fire proofing, fire protection, sound proofing, sound protection, and as horticultural growth medium.