DE1774927A1 - Arithmetic matrix - Google Patents

Description

Computation matrix (separation from application P 1 ¢ 21 7 62.0) The invention concerns the adaptation of certain lengths or areas to certain overall lengths or areas, what for the division of plate or sheet material with arbitrary Defect areas in predetermined defect-free sizes is typical, and in particular cutting flat glass into salable sizes after inspection in an assembly line glass manufacturing process. Flat glass, such as sheet or sheet material, becomes veneered after grinding and washing checked before it is cut into various market sizes will. The glass is of various types and size and unpredictable flaws Subject to location, thereby creating platters that are cut so that they are the areas of failure contain, can be completely unusable, or thereby the quality classification the salable Glass plates can be affected. To customize or completing various combinations. of different sizes In the usable areas it is of course desirable to use as little glass as possible, that is free from defects in a thickness that affects the quality of the glass during cutting is to be thrown away or as. to be classified as inferior. There is always time and time control to be taken into account, since the work process is a secondary circumstance of manufacturing high Production is in the case of glass of a certain width, either in the form of large, separated Plates or sheets or sometimes as a continuous belt at a constant speed moved during its shaping and finishing. Achieving high efficiency when staking out the gradual defined error-free area of the total produced Glass in usable sizes has been limited by a number of factors, including Consideration on the part of experienced, operating people has proven to be expedient Has. There are many such factors. For example, a large number are certain Sizes available on the market, although more than one sales quality may be available. Control Is necessary for the total output, since the error distribution is not is predictable. Relative preference or demand for different sizes can be also be a fact. The demand for special sizes may vary according to the Aengen change that have already been cut, Apparently the number of Choices: of glass sizes that are assessed before a cutting process must, grown very large, all factors or choices fall into account will Human judgment is at best under such conditions intuitive and at least makes it difficult, but always more expensive, to make decisions faster to be achieved by multiplying human labor output. Still need it Coordinated problems usually have time for their full solution, and economic considerations of manufacture are based on the cost of the time-consuming part of the determination a cutting process with minimal waste. In the interest of reducing the Storage and handling of the produced glass prior to transportation is also a close inspection of the elapsed time included. The work sequence of the error control, the Establishing all relevant data, making decisions regarding economical cutting process, the actual cutting operation and guiding the cut glass sizes to the corresponding packing stations are more desirable Essentially an assembly line process. The invention makes this possible. According to the invention is a method for determining the rectangular division of a glass surface or -plate created on the basis of various predetermined width and length sizes, the sizes in groups with an equal dimension longitudinal to the glass plate aligned and all possible combinations of the predetermined sizes, which have said one dimension in each group, are determined, where the total of the other dimensions of these sizes within each group the Glass plate width does not exceed.

In addition to dividing the glass or others plate-shaped Materials it is noted that the lengths or area sizes in general by Whole numbers are represented, which are divided into combinations of numbers are supposed to represent which members of a group of dividing element numbers.

The invention is apparent from the detailed description that follows illustrated in the drawing.

Show it: 1 shows the stages of a logical two-cut process for staking out sheet material according to the invention manure; Fig. 2 shows an example of a leading part of a Glass plate on which a special error history is indicated locally; 2A is a table for the example error lschema showing the faulty square units of Fig. 2 with regard to the local Zage defi- ned; 2B shows a table of the programmed, for example Quantities with degrees and values that appear from the glass panel the pig. 2 are to be cut out; FIG. 2C is a table showing the division of the experiments Plate of Fig. 2 illustrates the choice a certain group of cutting commands according to the programmed sizes of the Pig. 2B lead; 2D, 2E and 2F are graphic representations of the search divisions, give way to the table of the pig, 2C represents; Fig. 3 is a graph of a general Solution of a distribution problem; Fig. 4 is a graphic representation of a selective Solving a distribution problem; Fig. 5 shows the representation of an electrical 2-mission matrix, which is analogous to the distribution definition and the example of Fig. 3; 6 shows an illustration of an electrical command trix analogous to the selective apportionment defi- nition example of FIG. 4; Fig. 7 is a circuit diagram of a size switch and Able- display circuit for the iliatrix of FIG. 6; Fig. SA a 12-unit, length of a plate or table fel, the errors in the first and ninth ness wears; Fig. SB is a circuit diagram of a 12-unit error recovery relay circuit for reporting the errors of Figure 8A; distributor, which is controlled by the fault relays of the FIG. 8B for the purpose of exciting the rilatrix of FIG. 6 is operated; 8F shows a representation of the voltage connections to the matrix used by the voltage distributor of the Fig. 8B and Fig. 9 is an explanation of the electrical symbols. While certain specific embodiments of the invention have been shown and described in detail herein, it is not intended that the invention be limited in this manner to the specific forms or details illustrated. The invention also embraces all alternatives, modifications, and equivalents falling within the scope outlined by the claims, and the description has been structured to provide the reader with as brief and clear information as possible, for this purpose, as the complicated context of FIG allow some of the industrial embodiments described. data in cutting decisions for the ejected glass of a flat or sheet glass factory. Section IV describes the automation and the mode of operation of such a computer in a production line. The symbols of Fig. 42 serve as a preliminary comment. Since a computing system is not easily suitable for display on circuit diagram drawings on small sheets of paper, certain agreements are made to make things easier for the reader. For example, switches for which no actuating solenoid or other actuating device can be found in the drawing are denoted by the symbol according to (a). Normally open and noria -!: -. Normally closed switches, which are operated by a magnetic coil winding shown in the drawing, are marked with symbols according to (b) and (c). Relay windings for operating such switches are shown as circles surrounded by a reference number according to symbol (d). Capacitors are shown with the symbol (s) to a potential to avoid confusion with switch symbols. I. Practical Implementation of the Method According to the Invention A. Area Division The overall method of area division of plate or sheet material according to the invention comprises a process which can be referred to as a two-step logic. For a generalized presentation of the method in the form of the existing method steps, the reader is referred to FIG. 1 , which refers to flat glass, such as plate or sheet material, from which error-free, prescribed sizes are to be cut. To begin with, when cutting flat glass as a slat or sheet material, it should be noted that three basic cutting conditions must be met. First, the board or continuous belt is intended to be divided into rectangles with previously planned or programmed dimensions, as opposed to a method in which the defective area is merely cut out of the board or the largest available defect-free piece of glass sheet is always cut. The planned sizes are typically very much larger than the rectangular remnants with defects that have to be eliminated as scrap. Second, each cut must be entirely across a dimension of the piece of glass in question. This cutting condition is in the nature of the glass, since the cutting process scratches the glass and breaks it off in a jumping manner. The same condition is important for other slat materials to the extent that the costs for secondary expenditure and equipment for cuts of partial lengths or cuts that rotate over corners are excessively high. Third, the staking or partitioning must be selective so that the waste has a smallest surface area in accordance with the choice of sizes into which the defect-free glass can be partitioned. Since the lighter strength influences the classification, it should be noted that a glass area can be free from defects with regard to one quality, but not free from defects with regard to another quality. The sheet of glass described in connection with an industrial embodiment of the facility may, for example, be a 6.35 mm thick smooth produced as a continuous ribbon over a width of 3.05 m at a speed of approximately 5.08 m / min . A transverse cut through the glass ribbon width is the first cut of each pass and is referred to here as the Z-cut, and a group of second cuts, here called S-cuts, slit the cut Z-piece into smaller widths and eliminate the Failure. This is the so-called two-cut logic. The manner in which it is advantageously and uniquely used is described below. With reference to FIG. 1, the information required for dividing the glass consists in the indication of the position of the leading or leading edge of the glass, the position of the imperfections and the program of the sizes to be cut. For determining the local position of the defects, various devices can be used, namely zwecikmäBigerweiee to the loading end tracht total area before the first cut is made in the form of a visual or photoelectric control of the glass plate. This information can be viewed as being stored directly in and on the glass, but can be more easily kept by adding markings or x-marking devices. Preferably, however, the error is stored secondarily, for example on tape or other recording medium, which contains the dimensional size of each error square with respect to the leading edge and the side edge. The entire requirement program contains the absolute sizes and the relative requirements for these, which is characterized by assigned values for each size. In terms of sizes per se, the lengths and widths of the various rectangles to be cut are set so that the corresponding width dimensions are grouped which have a common length dimension. All length dimensions are grouped in this way, with each of these being associated with one or more width dimensions. The relative for which either the entire bandwidth or parts of it are available for splitting as error-free spans or fields depending on the number of the local positions of the defects which influence the quality grade to be considered. Every defect is contained in an error strip as long as the programmed length is considered. Next, the programmed widths for that particular programmed length are fitted into the available widths according to an assigned rank. The waste, which contains mismatched portions of the defect-free widths as well as the defect stripes per se, is reduced to a minimum by choosing the number and type of programmed widths that best divide the available defect-free widths. For each programmed length and corresponding widths experimentally used in this way, the values of the programmed sizes used are stored as the value which is assigned to these experimental lengths or cross-sections. The stages of fitting the programmed widths into available error-free widths and subsequent value stores are repeated for each programmed length before any cutting decision is made. In practice, a number of first trial cuts are made across the width of the glass, and by this logical procedure the cut for which the best fit is obtained, ie the cut for which the value of the fitted programmed sizes is greatest, is selected. After the selected cut determination has been made, commands are issued to cut this programmed length and then to cut the fitted widths of this programmed selected length. A large number of programmed lengths or test cross-sections can thus be considered, and it depends on the specific. Errors and the cutting program, whether the cut selected first detects the smallest or largest or an intermediate of the programmed lengths. Of course, it can be seen that the larger the cross-sections (ie, programmed lengths) that are considered, the greater the number of errors. The values of larger sizes can, however, compensate for the larger absolute drop per cut, in order to obtain the most efficient and most economical way of working in accordance with the fulfillment of the production conditions during long periods of operation. After the local outline of the new front edge is systematically determined in this way, the process steps are repeated. So each programmed length is considered again, but with regard to the new leading edge. The fault positions with respect to the leading edge are of course changed, and the resulting leading edge initiated. The process continues as one decision at a time as to the length of the plate material. If the entire longitudinal dimension is very large, as is the case with a continuous strip of flat material, the process steps are repeated continuously. The programming is usually set up with daily or weekly production in mind, and a farm count is conveniently made of sizes that are cut but not properly stored, as the program can be revised when the volume requirements are met. The programming must of course be adjusted to take into account ongoing economic factors and requirements. It should be noted that the actual time at which cutting occurs can be delayed while the sequence of operations is being determined. Thus, a panel can be marked for subsequent cutting of successive selected lengths (with accompanying adjusted widths), with each selected length representing the leading edge for the next. In some cases the length of the plate material may not be very great in relation to the various programmed lengths, so that a noticeable residual length at the end of the rod could appear as waste. A further method step is then preferably added in which no programmed length is considered which, if selected, would make the remaining length of the disk indivisible. In this form, the two-cut logic requires a separate determination of the length division as a permit to take into account the first cut length. The Z-point in time logic also responds to lengthening as a basic procedural cut in a Z-point in time multiple system. A greater number of choices among the same Z and SW size and value program is obtained by choosing the highest value composition Z in a given: length with respect to a leading edge. Each compound Z is a given number of programmed Z- Values in a particular order with respect to the leading edge, and its value has been subject to change by the order in which its proportional Z-sizes are arranged. This requires separate evaluation of the various permutations of the fractional Z-lengths for a given composite Z. The requirements / requirements / information storage are very high in such a system where substantial numbers of, sizes of Z by S are considered, and the preferred systems specifically described here work on the basic Z-point in time logic. To determine the required, error-free sizes, a basic dimension unit is set up for which all length and width conditions are whole multiples. With a belt width of 3.15 m, as in FIG. 2, a unit of 5.08 cm was chosen, giving the glass an effective width dimension of 62 units (S dimension or dimension). Naturally, if the programmed sizes for cutting are to change by intervals as small as 25.4 mm, the unit size would be selected as 25.4 mm. However, the unit square remains much smaller than the programmed sizes as a first step in reducing the amount of flawless glass that, together with the actual flaw areas, must be discarded as scrap. If the sheet or board is viewed as a l-mosaic or a matrix of 50.8 mm squares, the defects are localized by the Z and So coordinates for the unit square in which they lie. The glass is measured in units back from the leading edge along the length or Z dimension and the width or S dimensions are measured from one of the side edges. As shown, the coordinates relate to the outer boundaries of the squares themselves. The beginning bounding edge of the first Z or S stripe is considered to be zero. Thus, a square Z6, 8 22 is bounded by Z unit dimension lines 21 and 22. With reference to FIG. 2, a number of spots which are speckled on the glass are localized by Z and S coordinates in the table of FIG. 2A. The voids areas are assumed to be arbitrary or accidental because their number and location cannot be predicted in actual production. The defects may have microscopic size and be much smaller than a unit square, or they can cut several mosaic units as z.Bo in the illustrated scribing or scarring-like flaw you. "However, any defect considered in every case as if it is a Likewise, since each square must be examined, either in whole or in part, according to the nature of the defects present to determine defect classification, the smallest unitary dimension can be provided by the limitations of the defect detection devices. Although bulbous edges on the sides of the tape are practically continuous , elongated voids, such foreseeable defects are eliminated simply by edging or trimming, and consequently the locating edges mentioned herein are intended to be reference axes or lines which are considered physical edges after any predetermined edging or trimming operations have been performed irrespective of whether such cutting operations are delayed until such time as defects are measured or size cuts have been made. The table of FIG. 2B illustrates a simplified of sizes to be cut, listing eight sizes which can be reduced to three different length or Z dimensions, each of which has an associated group of two or three widths or dimensions. Have S dimensions. Thus, for any given Z-cut along the length of the sheet of glass, there may be a choice for the S-cuts. In the table of Figure 2B, the larger each S in a group of S dimensions for a particular Z ranks above each smaller S. Further, each size is Z x. S is assigned a value which, in this particular case, corresponds to the S dimension of this size: For the example given, this means that different areas with the same S dimension have the same value. However, this is not necessarily abnormal as will be seen. With this information supplied with regard to the desired sizes and the feriler distribution, the usable areas are marked out by the two-cut logic (as indicated in the table according to Pig. 2C). As a result, each of the Z dimensions (35, 210 and 9) forms the basis of an experimental first cut, taken from the front edge of the glass. Then each defect, measured from the reference side edge of the glass, is marked as a strip equal to the width of a unit, with certain available defect-free widths remaining. When fitting the S dimensions into an available width, the division is finitely determined for the smallest waste in that the highest-ranking S dimensions are always used where a choice is made. If no exact division is available, the usable width is reduced by one and the fitting is tried again. This process is repeated until (; in (, - full division is reached or the error-free width has been used up. FIGS. 2D, 2E and 2F graphically illustrate the S-fits for the three Z-tests listed in FIG. 2C Crossed hatched area represents the flawed area, and the single hatched areas represent the broken glass with the flawed strip, if the Z-cut is chosen, due to the impossibility of producing an initial fit. Sufficiently large pieces of waste can be saved if necessary, If one stakes out special or non-programmed quantities. By adding up the assigned values of the quantities used for each Z attempt, the choice of Z is determined by the highest storage. In this case, Z is chosen to be 20, since the value is 58, so that only defect strips of two units and two additional strips of one unit of unmatchable cutout oussglas are available. In the case of a link to the fourth , the link is systematically interrupted in that an initial selection is made to include the first (or the last) of the link values considered, and the order of the Z trials is chosen accordingly. For this no-link example, the Z attempts were made in the order of decreasing Z length, and if a link of the largest value occurred it would be in favor of the first summed value, ie, the largest Z, has been interrupted.

When selecting the rank in the manufacture of S-.Einpassungen, the value when selecting a Z to the tentative fitting and the order of Z-tests for breaking links can virtually be anticipated every Progranusierungsanforderung without personal judgment requiring after the program has been set up. It is added that in the example shown, a small scrap area per Z unit length logic is determined by making the value proportional to S. 1-If the Lert were proportional to Z x S in dimension units, then in this example Z would have been chosen to be finite 35 in spite of the larger scrap amount. If in the S-fitting, #; the edge is not in the order of the S-dimension, links of the division in the S-dimension are interrupted in favor of the smaller sizes and, of course, are cut. In the course of time, ie in the course of the production of a day, a week or a year, the reject rate is by definition minimal, since the programmed values have been cut in the programmed order, among other things, to meet the requirement calculated in this way. In the event that the tape is cut into cut sheets of, for example, 4.57 m or 90 units prior to the inspection, it is preferable to avoid remaining excess lengths. This is achieved by always choosing a Z, which is part of an exact fit of the various Z dimensions in the overall length. If the highest value Z is not available, the next higher value is chosen, etc. In this case j, any of the values 9, 20 or 35 can be selected as the first section f; e without impairing the longitudinal fit. Choosing Z with 20 in the first cut leaves 70 units into which two ZLs of 35 can fit. However, a Z of 20 (or 9) would be eliminated from second cut considerations.

Usually, by including a large number of Z's in the program, a large number of length choices remain open. The number of tentative Z-scores that are subject to the S-test subdivisions and evaluation is naturally increased. However, the effort of the additional decisions is made easier by their still systematic solution. If the choice is likely to be limited due to fewer or only large programmed Z numbers, a small remainder in the total length can be tolerated. A preferred method comprises dividing the total number of lengths alternately on a trial basis, reduced by, for example, one, two or three. An adjustment leads to only a small drop in any case. C. - Basic definition of apportionment. The starting point of the two-dimensional or areal division is a basic method of one-dimensional or length division. Such a division can generally be expressed in terms of groups of integers, the sum of which is exactly equal to a given integer. Dividing a large sum into selected smaller integers, like prime factorizing a large number, is often easier to do than explain. When calculating prime factors, calculation tables can be used for the sake of simplicity, and the results are valid regardless of whether they are explicitly expressed by 10 procedures are determined. Since a definition of call sharing as distinct from factorization is not easy to set up, the following expression is proposed as a definition, at least for the purposes of this description and the claims. A = n1 a1 + n3 a3 + .... + nn an. In this expression for A, the integer is divided (i.e., a compound), and ai can generally refer to the integers of a group a1, a2, a3, etc. that fit into A with no remainder. The number of each such integer ai contained in a partition is n1, n2, n3 etc. or n .. The value of ni can be zero in the case when its associated ai is not present in the partition. The use of the base number indices 11, 22, 33 need not involve sequential or quantitative ordering, nor should any numerical identity be assumed between the ni and ai to which the same index number is applied. The only division is obvious if the only a1 is the integer 1. It is equally obvious if the only a1 has a value equal to A. There would be no practical problem, or even the need to refer to the process as "splitting", if it only included these trivial splits. However, this is not the general case. With regard to the discovered usefulness of the defined concept of partitioning, two types of solution are of particular interest. One type can be called the general solution, in which all possible divisions of A are recorded and evaluated by the integer ai. The other type - can be referred to as the selective solution, in which conditions for the uniqueness of the solution are imposed by assigning rank or priority order to the various given integer ai. General splitting is used in the Z-splitting method discussed earlier. The information required indicates whether a given ai is or is not present in any of the possible partitions of A (ie, whether its ni is different from zero). Information as to the number of such partitions in which this ai is present or the number of times (ni) that ai exists in any partition may not be required. Such a determination is of no use if the ai group contains the number one, and integral higher ai are necessarily included. Fig. 3 graphically illustrates the possible Insertions of any or all of the integer ai - values 3, 5, 8 and 12 in a gangzahliges A with the value 20. This means A is an ordered group of verlau- fenden parallel in gleichrAäßigem spaced horizontal lines 0, 1, 2, 3, .....: ... 20 shown. The divisions are represented by vertical link chains that have corresponding length spacing lines, the number of which differs by 3, 5, 8 or 12 as shown in the illustration. Thus, for an A of 20 (with reference to FIG. 4) links of 3, 5, 8 and 12 in different chains are each shown ending at line 20, indicating that for an A of 20, each of these are integral ai can be present in a breakdown. The selection of any values from 3, 5, 8, or 12 as the first Z-cut out of a Z-length of 20 is thus acceptable. FIG. 3 is further designed to show the availability of various integer ai for each of all subsequent percentages. In this way t: If A is equal to 17, if an ai of 3 is the first choice, and it is observed that all integer ai still remain available as a second choice, since terms of 3, 5, ß and 12 are from the line 17 depend. If the second chosen a1 is 12, a link that extends from line 17- to line 5, then only ai of 5 is available for the third choice, since only one of 5 is attached to line 5 . The complete division is thus 3, 12, 5 and the other possible divisions can easily be taken from the table. This representation contains permutations of the groups of integer ai and coefficients nix which form a division (e.g. 3-12-5, 3-5-12, 5-3-22, 5-12-3, 12-3-5 and 12- 5-3).

It must be noted that FIG. 3 does not illustrate all the divisions for the integer A's less than 20, but only those divisions which are obtained by subtracting one or more of the integer ai. Thus, no links end at line 16 despite the fact that it is evenly divided evenly by two numerical values 8, since 16 is not a subdivision of 20.

Although the general solution illustrated in Figure 3 shows acceptability or future freedom of choice with respect to any chosen ai, it contains multiple choices which would make the S-partition described above indefinite. In the above-mentioned ranking, in which, for example, a larger number always precedes, the subdivisions with smaller numbers of the highest integer ai are eliminated. Considering the most senior ai as the ai member connected at its upper end to line 20 of the parallel family in FIG. 3, the highest ai is chosen from any of those available in a general split of 20. This is of course 12, leaving a next A of 8 for which the general divisions are 8, 5-3, or 3-5. According to the rules, 8 rules out the others, so that a solution is automatically obtained for n1 (a1), which is one, and for n2 (a2), which is 8.

FIG. 4 further illustrates selective division for all integer A up to and including 12 and for integer ai of 4, 3 and 2 in this rank. The constants of the apportionment equation are changed here to indicate the generality of the approximation to a solution and to provide a sufficiently simple example for which a mechanical correspondence can be clearly and precisely illustrated in the following section. In this example it can be noted that each A is divisible and that there are no links or undecidable choices. For other examples, no division may be available and the chosen solution is the next highest, divisible A. In terms of method A is decreased by one, until it finds an adjustment for which there is no rest. The same result can be obtained in this particular example by substituting an ai equal to 1. However, a method which obtains an ai equal to one hides the fact that such ai may not in itself be a useful or desirable part of a partition. Thus, when the other ai values are high, a partition solution can form a large number (ni) of ones.

Since the general solution includes the unique solution, it is recognized that the unique or selective solution also indicates the existence of the integer ai, albeit in one and not in all of the possible divisions. span which are spaced numerically in an integer ai assigned to the cell. When a voltage that is positive with respect to the null bus is applied to any of the other bars, current will flow down through any unbroken cell strings between the end layers. To prevent upward flow of current, each cell contains a unidirectional conductive or rectifying device, conveniently a silicon diode 45. The diodes are polarized so that they have a negligible forward voltage drop from an upper rail to a lower rail. (The contact pairs shown in each cell are described later and are closed to operate the matrix). It is noted that the numbering of the busbars is of structural and operational importance with regard to the mutual diode connections, although the order of the diode connections from left to right is a matter of choice.

Apparently, each ai busbar span is connected to a cell of the corresponding span. The geometry chosen to represent the circuit according to FIG. 5 links the cells with the same span in chains and groups the chains in blocks in order to form the complete meshing. In this way, the 3-span block for ai = 3 has three vertical cell chains, the first of which extends from the rail 20, the second from the rail 19 and the third from the rail 18. In the 12-span block are Connection cell of each block responds. As shown in Fig. 5, the cell of each block, the lower end of which is connected to the 0 bus bar, contains a sense relay winding. As indicated, the relay windings (each shown here and in the following figures as a circle bearing a reference number) for the respective blocks 3, 5, 8 and 12 are designated as 48, 49, 50 and 51. Other types of indicators can be used, but the relay windings have the advantage that they easily provide a number of switch contacts for any display or control function that may be required. No diode 45 is required in the last cell, which contains the read-out relay, since no return current paths are possible as a result of the matrix end connection. All cells thus remain members acting in one direction.

In order to facilitate the solution of dividing equations, the respective blocks can be switched depending on which integers are required in the group in or outside the matrix. For example, referring to the block of 3 span cells 41, there is a matrix size relay 50 which has a number of normally open contact pairs. It includes 1? Contact pairs 50a, namely a pair of diodes in each cell and another pair 50b in the Ablesezelle. Relay winding 50 is suitably energized by connection through selector switch 51 from a suitable control voltage source, shown as battery 52. When switch 51 is closed, relay 50 is energized and all 3-span cells 41 are switched into the matrix. Numbers 3, 5, 8 and 12 included in various ways in some of the possible divisions of 20 so that the reading relay of each block is energized. Any of the integers are allowed, ie their subtraction from 20 leaves a remainder that can be divided. If, for example, an approved cut of 12 units is selected from a plate with a length of 20 units, the remainder of 8 can still be divided by a combination of 8, 5 and 3. This is easily determined by setting the selector switch 71 to the bus bar 8, noting that the lamps L46, L47 and L48 are on.

Not all A can be divided up using this number program. For example is then, if the voltage is connected to the busbars 7 or 4, no Rung available through any combination of 3, 5, 8, and 12. of course is even an A smaller than 3 cannot be divided, since 3 is the smallest programmed whole Number is. However, there are connections from cells to all busbar spans in terms of the density of the scheme in illustrating the matrix and to anticipate incomplete division in the event that further ai numbers added or replaced by switching operations in other cell blocks will.

When the matrix is formed so that it is only a single whole Dividing number, some cells can be omitted. For an A of 20 and the one shown ai program, for example, the busbars 19, 18, 16, 13, 7, 4, 2 or 1 connected cells may not be required. these are however, special and not general cases, and the note thereon, that all spans are interconnected does not intend the connection of the unusable span widths to be requested, which are only available in the case of boundary conditions or specific examples occur.

As long as the matrix does not contain the combinations contained in the split identify or specify the frequency with which a particular whole Number is applied in the matrix current paths, proof or reading is sufficient of the current through a terminal cell of each cell block. This follows from the fact that all permutations find current paths between the end connection bars. Consequently has a different one of the permutations of each combination part on its rung through each end relay each block of numbers at each point in a current path lies. Reading relays can be provided in additional cells if necessary to provide additional information. Such modi fications are clearer in the following subsection in connection with the description of a selective Distribution matrix illustrated.

B. A Simple Selective Partitioning Matrix A selective partitioning matrix that provides a unique solution is shown in FIG. It is a constructive analog of the solution to the apportionment equation illustrated in FIG. Compared with a permissions matrix of the type shown in FIG. 5, the stated whole readings are only limited to those in the highest-ranking distribution combination. Such a matrix, which also indicates the frequency with which each displayed integer is present, can be referred to as a command matrix, since it contains a specific or unambiguous group of instructions that best adapt to the input conditions according to their built-in logic or decision-making instructions supplies.

The applicability of the command matrix for S-division can easily be realized will. In addition, the matrix of Figure 6 also illustrates other useful features for the construction of the distribution matrix with regard to the expansion of its tasks and their versatility. Because the S division always adjusts to errors there where these might occur are the A-spans of the matrix shown limited by the error without affecting the A-span with the zero busbar is related. While the number of busbars in the example shown is small, this number can be very large, and several A (i.e. whole Numbers) can be split at the same time.

In the matrix of FIG. 6, a horizontal array represents busbars 60, numbered 0, 1, 2, 3 .... 12, a maximum A-span of 12 represents a correspondence to the selective distribution example of FIG.

As before, the perpendicular directional characteristic of the matrix is produced by cells or members 61, 62 and 63 which are conductive in only one direction and which are connected between busbar spans of 2, 3 and 4, respectively. Each such cell or member includes a diode 64 which is polarized for current flow from a higher numbered or more positive busbar to a lower numbered busbar. All available spans of 2, 3, and 4 are connected by the links that again- Reading circuit. In operation, a voltage applied to the positive terminal of busbar 12 and positive with respect to the null busbar causes current to flow through a path which corresponds to the only division given for such a value of A in FIG. 4, namely a chain of three 4-span cells 63 (from busbars 12 to 8, 8 to 4, and 4 to 0) that form the only current path. Of course, none of the other busbars is energized and the portions of busbars 12, 8 and 4 to the left of the 4-span cell block 63 are broken by the opening of the contact pairs 67a associated with the energized relays of the power connection cells.

When a voltage is applied between the positive terminal of bus bar 11 and the negative terminal of bus bar 2, the current seeks a path over a span of 9 bus bars. With the number program and order of precedence given here, the exact decision must be in the order of 4, 3, 2. It becomes the only current path from busbar 11 via a 4-span cell 63 to busbar 7, from busbar 7 via a 3-span cell 62 to busbar 4 and from busbar 4 through a 2-span cell 61 tracked to the busbar 2. Each of the busbars 11, 7 and 4 is interrupted by the opening of the cell relay contact pair on the left side of the upper connection of each energized cell. In this way, all other current paths are excluded.

Matrix size relays can expediently be used for the matrix can be used to solve various partitioning problems; program. As shown in Fig. 7, the matrix size relays 68, 69 and 70 are each provided with a number of contacts 68a, 69a and 70a located in the corresponding 2-, 3- and 4-span cells 61, 62 and 70a, respectively. 63 (Fig. 6) are located. The relays are energized via respective size switches 71, 72 and 73, each switch and relay being connected in parallel to a suitable voltage source 74. All contact pairs are of course considered closed when analyzing the matrix design for the ai program of 4, 3, 2. It is noted that busbar contact pairs 65a are significant when other cell blocks are or can be switched into the matrix on the left-hand side of the 2-span cells 61.

As shown, the cell relay is connected in conjunction with the diode in each cell or member of FIG. 6, and the matrix size contact pair is a resistor that is optionally added to help maintain the relay current within applicable operating limits. The resistances in each cell of a certain range have the same resistance value. As shown, resistors 75, 76 * and 77 are provided for cells 61, 62 and 63, respectively. The usability of the resistors and the choice of its value depends on the power supply used and treated in a soft direct section. A simple readout circuit is conveniently made by counting the number of energized cells in each block in a suitable manner by means of normally open contact pairs from each cell relay in a display or control network (Fig. 7). To detect the flow of current through any of the 2-span cells 61, a galvanometer-type display device 78 'is connected in parallel to the voltage source 74 via parallel circuits consisting of the normally open contact pair 65b of each cell relay in series with a reading or value resistor 79. With such a display circuit in parallel with voltage source 74 or other stabilized voltage supply device, if no two span cells are energized, the current flow through the galvanometer is zero and increases with the number of contact pairs 65b closed.

Corresponding readout circuits are for the It is also noted that features of the matrix of FIG. 6 can be applied in a permissions matrix. For example, the normally closed pairs of contacts 65a, 66a and 67a in the busbars and the associated individual cell relays can be omitted so that the division is not selective. The galvanometer reading for the appropriate blocks then indicates the number of cells of each span that are energized. Included in this information is of course the indication of whether any cells or no cells at all are contained in a division. If only the more general information is desired, the reading network resistors 79, 81 and 83 can also be omitted. Such a modification, which requires the use of cells in each cell relay is advantageous if the relationship to be divided spans no beneficial loading to the same or resistant Enäanschluß- found (0) -Sammelschiene.

Regardless of whether the device according to FIG. 6 as well as the device according to Fig. 5 simplified to serve as a Permission matrix, it ensures Anachlußverbindung of at least one cell in JE dem.Block. It can also be seen that by opening the busbar switches , the matrix of FIG. 6 selectively excludes all but one on each energized busbar.

Briefly summarized, it can be said that immediately after the busbar voltage is applied, several current paths are established through the matrix of FIG. 6 , which correspond to all partition permutations as in a permissions matrix of the type shown in FIG. In this sense, a cell in each block connected to either the upper or lower busbar of the energized span also carries current if integers of that cell span are included in a general apportionment solution. After the energized relays open their normally closed contacts, the redundant paths are eventually eliminated in favor of the unique selective apportionment solution. It is an advantage of the circuit design that relay pick-up or actuation times are not critical, and the correct solution is also obtained if the cell relays in the lower tier cell spans should have shorter response times than relays in the higher tier. G. - Matrix voltage distributor Although voltages can be applied to the selected busbar span of a matrix, e.g. of the type shown in Fig. 6, by means of manually operated switches, the automatic correlation of applied matrix voltages with the local position of error-free spans of a flat material to be divided offers advantages for high-performance operation.

To economically meet the requirements of a particular facility it lies to some extent in the choice of the designer, either the number of Matrix busbars for the purpose of simultaneous division of several spans increase or tension at time intervals to appropriately suitable ranges of a to create a smaller matrix. Regardless of whether in practice of the invention the various new circuit arrangements in part or be chosen altogether, it cannot be regarded as the slightest advantage that such There are options here at all.

The voltage splitters of Figures 8C and 8E are such new constructions. In each arrangement, double pole changeover switches are used to apply the required voltage between respective busbars. As a simple example, Figure 8A shows a strip of sheet material 12 units long with voids in the first and ninth units. This means that lengths of seven or three units are available for splitting. In the fault switching system shown in Fig. 8B, twelve fault relays 85-1 to 85-12 correspond respectively to the units 1 to 12 of the board of Fig. 8A. Relay controls shown as simple switches again. Although the example is relatively simple to better illustrate the full connections to the small 12-span matrix of Figure 6, it will be noted that voltage switching is advantageously available for much larger matrices which have a much more complex fault scheme.

1. Series-Type Voltage Splitter The series voltage splitter of Figure 8C applies a voltage to each clear or error-free span that is proportional to the length of the span. Twelve residual current voltage sources 88-1 to 88-12 form the Looking at the details of the circuit of FIG. 8C, the fault relay contact pairs in FIGS. 8C aligned horizontally with the representation of the corresponding fault relay winding in FIG. 8B to illustrate the following switching operations. The contact pairs a, b of each relay are a first single pole changeover switch, since they have a common pole or center connection, and the contact pairs c, d or each relay form a second single pole changeover switch. With reference. to any unit or a corresponding switch or voltage source in the form of the number of units from the lower side of Fig. 8A, 8B or 8C as m is the negative terminal of every m-th voltage source in Fig. 8C with the common pole of the contact pairs e, d of the m-1 error relay and their positive Kelmme connected to the common pole of the contact pairs a, b of the m + 1 error relay. The end connections of the contact pair c of the relay m @ and the contact pair b of the relay m + 1 are connected for the purpose of series connection of the voltage sources. The connections of the m-th busbar of the matrix run from the negative end of the bar to the end connection of the contact pair d of the relay m and from the positive end to the end connection of the contact pair a of the relay m + 1.

The end connections at the ends of the series connection of the twelve voltage sources are simple. The series voltage connections are omitted, so that the contact pairs a and b of the error relay 85-1. and contact pairs c and d of unit fault relay 85-12 are not used (and are therefore not shown in FIG. 8C ). The positive terminal of the voltage source of the twelfth unit is suitably continuously connected to the positive terminal of the busbar 12, and the negative terminal of the voltage source 88-1 of the first unit is connected in the same way to the neutral busbar without the selective one To impair switching function.

The operation of the voltage distributor in the absence of any defects which cause the fault relays to be energized is simple. The twelve voltage sources are simply connected in series between the positive terminal of the twelfth busbar and the neutral busbar, as would naturally be desired in the absence of flaws that limit the span. This is readily apparent from the fact that as the matrix contact pairs A and D are open and all rows -kontaktpaare b and c are closed, wherein the direct connections from the positive terminal of the twelfth voltage source to the positive terminal of the busbar twelve and the negative terminal of first voltage source to the neutral busbar remain.

The operation under fault conditions can be seen in the following example of Fig. 8A, in which the first and ninth units of the plate material are faulty. As a result, the circuits of FIGS. 8B and 8C show closed fault switches 86-1 and 86-9, the contact pairs of the relays 85-1 and 85-9 being switched over from their normal state. As a result, voltages are applied or distributed between two different spans of the matrix, as further illustrated in the simplified circuit of FIG. 8D. Thus, seven voltage sources 88-2 to 88-8 are connected between the positive terminal of the busbar 8 and the negative terminal of the busbar 1. Three voltage sources 88-10, 8 $ -11 and 88-12 are connected in series between the positive terminal of bus bar twelve and the negative terminal of bus bar nine. The first and ninth voltage sources remain disconnected.

The adaptability of the voltage distributor in manufacture direct correlation between the flawless spans of the plate and the corresponding matrix busbar for each possible defect scheme is remarkable. Successive working methods for the different spans are not required, and divisions occur simultaneously in the matrix of FIG. 6. Although the reading can be recorded separately from each cell of the matrix of FIG. 6 remains the resulting reading arrangement in the circuit of FIG. 7 is of particular use in creating an additive reading for all simultaneously derived divisions as a sum for comparison with the sum for dividing another section the plate that has a different defect pattern.

It will be noted that large collector spans have greater voltages, and all of the resistors 75, 76 and 77 previously identified as being optional in the matrix of Figure 6 play a useful role. Thus, if all the relays have a common nominal value, the ohmic resistance of each cell is conveniently proportional to the cell span in order to maintain equal voltages and currents in all cells of a particular current path. If beispiels-, ten volt sources and relay 1000 ohm Widklungen are available with a nominal operating current of 10 mA, the resistor 75 of each 2-spans cell 61, the order of 1000 ohms to the current to the nominal value of limit at 20 volts. Similarly, to maintain the nominal value conditions for the 3 and 4 span cells 62 and 63, the associated 76 and 77 are 2000 and 3000 ohms, respectively. 2. Parallel-Type Voltage Distributors The problems associated with switching high voltages are eliminated by the use of a parallel-type voltage source in which the same given voltage is applied to each fault-free span of the matrix. Such a circuit is illustrated in FIG. 8E in which fault relay contacts a ', b', c ', and d' are again aligned horizontally with their respective fault relays 85-1 through 85-12 of FIG the same switching contacts for parallel operation. Voltage sources 88-1 'through 88-12' represent the twelve voltage sources for the corresponding twelve units of the plate of Fig. 8A and the busbar spans of the matrix of Fig. 6. As in Fig. 8C, the positive and negative busbars are also shown. The contact pairs of a ', b', c 'and d' of the relay 91-102 8E in relation to the series circuit in Fig. 8C are in a different manner in the parallel circuit of Fig. Connected to the common potential intermediate busbars effectively remove, which delimit the sides of an error unit or a strip of units. That is how it usually is It can be seen that with each type of voltage distribution, voltages are applied via a switching system which provides for automatic application to different spans of the same matrix. A relationship with a constant or fixed bus bar becomes unnecessary and, in fact, voltage connecting bus bars are re- selected for such a splitting process.

Claims (2)

P atent claims 1. Computing matrix for determining the possible fit of elements of a group of given integers into a larger given number, characterized by a group of numerically ordered conductors (40, Fig. 5, etc.), members (41 , 42, etc.), which span pairs of conductors with numerical differences, corresponding to each of the said integers of the first group, devices (52, 59, etc.) for applying a voltage between conductors, the numerical difference of which corresponds to the said larger integer , and means (46, 47, etc.) for determining resulting current paths through at least one of the named members to determine whether the corresponding integer is an element of a possible fit. 2. Matrix according to claim 1, characterized in that the only one-way conductive members (61-63, Fig. 6, etc.) are arranged on said conductors (60) in a certain order which corresponds to the relative rank, and that there are devices (65-67, Fig. 6 etc.) which open current paths between conductors through members of lower slope when parallel current paths exist between the same conductors (60) through members of higher rank, whereby the indication of all possible fits up to on one is eliminated.