DE102009044376A1 - Method for determining evaluation index from values of to be evaluated real parameter of memory cell, involves assigning maximum non-closed value ranges of real parameter to highest memory cells - Google Patents

Abstract

The method involves outputting an evaluation index by an output unit, and carrying out assignment of values of a real parameter to multiple memory cells of matrix dimensions of n-dimensional matrix according to a predetermined distribution pattern. The memory cells of each matrix dimension are provided by the distribution pattern. The distribution pattern is designed such that values of closed values range of the real parameter are assigned to one of the memory cells. Maximum non-closed value ranges of the real parameter are assigned to highest memory cells. Independent claims are also included for the following: (1) a device for determining an evaluation index from values of to be evaluated real parameter of a memory cell (2) a computer program product having a set of instructions for executing and determining an evaluation index from values of to be evaluated real parameter of a memory cell (3) a data storage medium comprising a set of instructions for executing and determining an evaluation index from values of to be evaluated real parameter of a memory cell.

Description

The present invention relates to an apparatus and a method for determining an evaluation index. More particularly, the present invention relates to an apparatus and method for evaluating, in particular, the quality of a plurality of real-size values (p _i ).

Apparatuses and methods for evaluating a plurality of real-valued values (p _i ), such as the 6-σ method, as well as similar methods and their application to appropriately configured apparatus are known. With these methods, attempts are made, in part via statistical methods, to obtain a multiplicity of values of at least one real variable, such as, for example, As resistance values, length values, eccentricity values but also empirical variables and economic values, or other types of values to evaluate and without comparing each value in detail to make statements about the sizes and, where appropriate, the distribution of these values.

It has been found to be disadvantageous in such methods and devices for carrying out such methods that so-called "outliers", that is to say values which deviate particularly strongly from a nominal value, strongly influence the evaluation of the entire value chain and thus possibly lead to a distortion of the result of the evaluation being able to lead. Similarly, using known methods, a substantially uniformly distributed deviation of the values by one setpoint can lead to a positive evaluation for the entire set of values, although in some cases large deviations from the setpoint are present.

The object of the present invention was therefore to provide an apparatus and a method for determining an evaluation index of a plurality of values of a real, preferably measurable size, which can easily and yet effectively avoid a misjudgment of the range of values, and in particular Outliers "do not unduly adversely affect the outcome.

This object is achieved by a method and a device according to the independent claims 8 and 1. Preferred embodiments are subject of the dependent claims.

Accordingly, a device for determining a weighting index (Q _i ) of N> 1 values (N ₁ , N ₂ , ... N _N ) of at least one real variable (p ₁ , p ₂ , ... p _X ) to be evaluated has means for recording the individual values (N ₁ , N ₂ , ... N _N ) of the real variable (p _i ) and storage means which contain an n-dimensional storage matrix, the number of matrix dimensions being equal to the number of real variables (p ₁ , p ₂ , ... p _X ). By means for the assignment of the values (N ₁ , N ₂ , ... N _N ) of the real variable (p _i ) to a plurality k> 1 of discrete memory cells (a _{1, i} , a _{2, i} , ... a _{k , i} ) the matrix dimension (i) of the n-dimensional memory matrix, an assignment of the individual values to the discrete memory cells according to a predetermined distribution pattern, wherein the number (k) of the memory cells of each matrix dimension is predetermined by the distribution pattern and the distribution pattern is carried out such that Values of at least one first closed value range of the real variable (p _i ) are assigned to a first memory cell (a _{1, i} ) and values of at most two non-closed value ranges of the real variable (p _i ) at most two further memory cells (a _{k-1, i} , a _{k, i} ).

By means for assigning a discrete evaluation variable (Z _j ) to each memory cell (a _{1, i} , a _{2, i} ,... A _{k, i} ) of the matrix dimension (i) of the n-dimensional memory matrix, an assignment of the evaluation variable (hereinafter Z _j ) to each memory cell, wherein a counter for determining the number (A _j ) of the each memory cell (a _{1, i} , a _{2, i} , ... a _{k, i} ) of the matrix dimension (i) of the n-dimensional memory matrix associated Values and means for forming the memory cell rating index (q _j ) by multiplying the number (A _j ) of the values assigned to each memory cell by the rating (Z _j ) of the respective memory cell, a memory cell rating index (q _j ) for each memory cell (a _{k-1, i} , a _{k, i} ). Finally, by a summation device, the evaluation index (Q _i ) to the respective real size (p _i ) from the memory cell evaluation indices (q _i ) of all memory cells (a _{1, i} , a _{2, i} , ... a _{k, i} ) of the matrix dimension (i) the n-dimensional memory matrix is formed by summation and output by means of an output device.

The device according to the invention or the method according to the invention thus provides as a result at least one evaluation variable (Q _i ) for a real quantity (p _i ) which contains all the values (N ₁ , N ₂ ,... N _N ) of the real variable (p _i ) is included in the rating and does not exclude "outliers" from the rating.

The evaluation index (Q _i ) of the real variable (p _i ) has the special property that such "outliers" are not only included in the evaluation, but also do not unduly influence the result, since, for example, they are equally strong can experience how this is the case for values that deviate only a certain, predetermined amount from the target. This is done by the particular embodiment of the distribution pattern, which is a structure which, applied to the values (N ₁ , N ₂ , ... N _N ) of the real variable (p _i ), these values in k different but discrete memory cells (a _{1, i} , a _{2, i} , ... a _{k, i} ) and the assignment of the discrete evaluation variable (Z _j ) to each memory cell (a _{1, i} , a _{2, i} , ... a _{k, i} ).

The device according to the invention or the method according to the invention also has the special property that values which are evenly distributed around a desired value, even if their arithmetic mean were close to the desired value, easily lead to a clearly distinguishable evaluation, whether the dispersion the values around the setpoint are large or small and whether they are evenly distributed or inhomogeneous. The evaluation index (Q _i ) thus represents a simple, singular variable which allows a large amount of information about the size and distribution of the individual values (N ₁ , N ₂ ,... N _N ) of the real variable (p _i ).

In a preferred embodiment, the device according to the invention comprises means for combining the weighting indices (Q _i ) of several real quantities (p ₁ , p ₂ , ... p _X ) to form an overall weighting index (Q), so that even in the case of a plurality of real Quantities as a result of the analysis of a singular size, namely the total weighting index (Q) is issued, which, like the valuation indices (Q _i ) allows a direct and clear statement about the sizes and distributions of several real sizes.

Particularly preferably, comparison means are provided which are set up for the evaluation indices (Q _i ) of a plurality of real variables (p ₁ , p ₂ ,... P _X ) and / or the overall evaluation index (Q) with one or more predetermined evaluation thresholds (T _i ). or (T) and output a single or overall comparison result by means of the output device, wherein for Q _i or Q <T _i or T the evaluation result "NEGATIVE" is output, for Q _i or Q = T _i or T the evaluation result "NEUTRAL" is output and for Q _i or Q> T _i or T the evaluation result "POSITIVE" is output.

On the basis of this output, for example, switching operations, such as a sorting process or a marking function can be controlled.

Preferably, means for determining the evaluation threshold (T _i ) for the respective real variable (p _i ) are based on the product of the arithmetic mean of all the evaluation quantities (Z _j ) of the matrix dimension (i) of the n-dimensional storage matrix and the total number of values (N) of the real quantity (p _i ), so that the evaluation threshold (T _i ) is indirectly determined by the values (N ₁ , N ₂ , ... N _N ) of the real quantity (p _i ).

In addition, it is advantageous to provide means for mapping a modified evaluation threshold (T _i ) by means of a predetermined evaluation parameter (E _i ) so that influence can be taken on the evaluation threshold (T _i ) by adjusting the evaluation parameter (E _i ) and thus also on the single comparison result. Thereby, the quality requirements can (p _i) are engaged and, for example, over a certain period, for example, in the otherwise largely automated running process the evaluation of the real size of a manual raising or lowering can be effected.

In a further embodiment, the evaluation parameter (E _i ) can be modified automatically as a function of the deviation of the evaluation index (Q _i ) from the evaluation threshold (T _i ), and thus the tolerance relative to a certain desired value can be set iteratively. It is advantageous that a change of a multiplicity of parameters, which influence the values (N ₁ , N ₂ ,... N _N ) of the real variable (p _i ), can be carried out simultaneously by changing the evaluation parameter (E _i ). It can be determined what effect the parameter change has on a plurality of valuation indices (Q _i ).

For example, the evaluation parameter (E _i ) is raised for Q _i > T _i and the evaluation parameter (E _i ) is reduced for Q _i <T _i . In this way, and depending on the sensitivity of the change in the evaluation parameter (E _i ), optimization of the overall system can take place simply via a large number of value measurements of the real variable (p _i ) and thus, for example, a quality criterion lasting at a very high level being held.

For ease of understanding, the present invention is explained below by way of example of the preparation of cellulose sulfate capsules, but it is expressly understood that with the aid of the present invention, a plurality of processes can be monitored and controlled in terms of quality, especially if the process not a continuous result, but one set of results to compare with more sets of results.

As is known, the production of cellulose sulfate capsules is often carried out in so-called batches, that is, discontinuously by dropping cellulose sulfate into a PADMAC fixing solution, wherein a variety of parameters such as cellulose flow, acceleration voltage, vibrating frequency of the dropping nozzle, etc. are geometric and strongly influence the functional properties of the capsules.

It has been found that different batches of cellulose sulfate capsules in relation to their geometric characteristics, such as capsule diameter and shell thickness already deviate significantly even with the slightest change in the production parameters of the desired capsule quality, however, affect different properties of the capsules and therefore a batch with only a few larger capsules must not necessarily be discarded, if others Capsule parameters are consistently good or above standard.

The present invention offers opportunities to compare individual batches from the production of cellulose sulfate capsules with each other in terms of a variety of properties, such. B. to make their size and shell thickness.

For this purpose, measured values of the sizes of the cellulose sulfate capsules of a batch and their shell thickness or even wall thickness are ascertained and, if appropriate, further properties, such as, for example, the sphericity of the capsules, are measured. The determination of the sphericity makes it possible, for example, to make statements about the quality of the dripping process or fixation.

The measured values (N ₁ , N ₂ ,... N _N ) are read into the device according to the invention, optionally automated, and read in accordance with the distribution pattern of the memory cells (a _1,1 , a _2,1 , ... a _5,2 ) assigned. The memory array of the device according to this embodiment has, for example, a number k = 5 memory cells and X = 2 matrix dimensions. By way of example, the memory cells (a _1,1 , a _2,1 , ... a _5,1 ) are evaluated as evaluation quantities Z _j = {1; ...; 5} and the memory cells (a _1,1 , a _2,1 , ... a _5,1 ) are evaluated as evaluation variables Z _j = {0, 1; ...; 5}.

The distribution pattern selected for the present embodiment assigns magnitude values corresponding to the target value to the memory cells (a _3,1 ) and (a _3,2 ), respectively, values that differ up to ± 10% from the target value of the memory cells (a _{2, 1} ), (a _4.1 ) or ( _a2.2 ), ( _a4.2 ) and values which differ from the nominal value by more than ± 10% for the memory cells ( _a1.1 ), (a5 _{, 1} ) or (a _1,2 ), (a _5,2 ).

Hereinafter, for each memory cell (a _1,1 , a _2,1 , ... a _5,2 ), a memory cell evaluation index (q _j ) is formed by multiplying the number (A _j ) of the values assigned to each memory cell by the evaluation quantity (Z _j ) of the respective memory cell: _qj = _Zj * _Aj = _Zj * COUNT ( _{aj, i} )

It is essential here that values which deviate from the nominal value by ± 11% influence the memory cell evaluation _index (q _j ) as well as values which deviate from the nominal value by 90%. Thus, "outliers" are given less weighting and in this way minimized the influence of measurement errors or actual isolated defective cells, as long as such defective cells are present only to a limited extent.

The memory cell evaluation indices (q _j ) determined in this way are subsequently summed up and thus result in the evaluation index (Q _i ): Q _i = SUM (q _j, )

The evaluation index (Q ₁ ) indicates in the present example a quality assessment for the distribution and scattering of the sizes of the cellulose sulfate capsules of a batch, wherein the influence of measured value errors in an automated determination of the capsule sizes can be minimized by the particular embodiment of the present invention. The same applies to the shell thicknesses of the capsules, which receive a rating in the present embodiment by the evaluation index (Q ₂ ).

In addition, by summing the memory cell rating indices (q _j ) of different matrix dimensions (i) an overall rating index (Q) can be determined. In the present case, this means, for example, that the combination of the evaluation indices (Q ₁ ) and (Q ₂ ), for example by addition, can bring about a mutual compensation, at least to a certain extent, that is to say that e.g. For example, a batch with a particularly high homogeneity in capsule sizes is still POSITIVE, when compared to an overall rating threshold (T) even if the shell thickness requirements are not met.

Equally, by means of this form of mutual influencing, only a slight undershooting of both variables, that is to say falling short of the nominal dimension AND falling below the shell thickness of the capsules, can result in a sensitive change in the overall evaluation index (Q) and thus a NEGATIVE evaluation, although the evaluation indices (Q ₁ ) and (Q ₂ ) would be positive on their own.

This makes it possible, so far only with considerable effort to achieve opportunities in the quality control of a wide variety of products and processes, in a simple way.

Finally, it is possible by the integration of the evaluation parameters (E ₁₎ and (E ₂₎ to influence the eventual evaluation in which, for example, the evaluation parameter (E ₁₎ to the evaluation index (Q ₁₎ is added or subtracted from this and As a result, the weighting of the capsule size is temporarily modified with respect to the shell thickness. A customer of the capsule manufacturer, for example, on the website of the capsule manufacturer, the quality requirements individually or in the whole for subsequent batches of capsules temporarily or continuously increase or decrease by modified Become the evaluation parameters (E ₁ ) or (E ₂ ) or a Enter total evaluation parameter (E).

An automated modification of the evaluation parameter (E ₁ ) or (E ₂ ) or (E) can also ensure an optimization of the cost-benefit calculation by producing a consistent quality of capsules at the lowest possible price, if assumed is that the quality of the capsules can be increased at a higher cost. If, due to a change in the manufacturing conditions, the quality of the capsules is continuously higher than needed and thus Q _i > T _i , an automatic change in the evaluation parameters can lead to lower costs but still maintain the necessary quality of production, as long as Q _i ≥ T _{i is} given. In this case, for example, the evaluation parameter would be linked to the cost of production and would iteratively strive for a minimum.

Claims (16)

Method for determining an evaluation index (Q _i ) of N> 1 values (N ₁ , N ₂ , ... N _N ) of at least one real variable (p ₁ , p ₂ , ... p _X ) to be evaluated, comprising the steps : a. Detecting the values (N ₁ , N ₂ , ... N _N ) of the real quantity (p _i ); b. Assignment of the values (N ₁ , N ₂ , ... N _N ) of the real variable (p _i ) to a plurality of k> 1 discrete memory cells (a _{1, i} , a _{2, i} , ... a _{k, i} ) the matrix dimension (i) of an n-dimensional matrix, the number of matrix dimensions corresponding to the number (X) of the real variables to be evaluated (p ₁ , p ₂ , ... p _X ); c. Assigning a discrete evaluation variable (Z _j ) to each memory cell (a _{1, i} , a _{2, i} , ... a _{k, i} ) of the matrix dimension (i) of the n-dimensional matrix; d. Determining the number (A _j ) of the values assigned to each memory cell (a _{1, i} , a _{2, i} ,... A _{k, i} ) of the matrix dimension (i) of the n-dimensional matrix and determining the memory cell evaluation _index (q _j ) Multiplying the number (A _j ) of the values assigned to each memory cell by the evaluation quantity (Z _j ) of the respective memory cell; e. Summing the memory cell rating indices (q _i ) of all the memory cells (a _{1, i} , a _{2, i} , ... a _{k, i} ) of the matrix dimension (i) of the n-dimensional matrix for determining the weighting index (Q _i ) of the respective real size (p _i ); f. Output of the rating index (Q _i ) by means of an output unit; characterized in that the assignment of the values (N ₁ , N ₂ , ... N _N ) of the real variable (p _i ) to the memory cells (a _{1, i} , a _{2, i} , ... a _{k, i} ) of the matrix dimension (i) the n-dimensional matrix is carried out according to a predetermined distribution pattern, wherein the number (k) of the memory cells of each matrix dimension is predetermined by the distribution pattern and the distribution pattern is executed such that values of at least one first closed value range of the real size (p _i ) a first memory cell (a _{1, i} ) are assigned and values of at most two non-closed value ranges of the real variable (p _i ) are assigned to at most two further memory cells (a _{k-1, i} , a _{k, i} ). Method according to claim 1, characterized in that the evaluation indices (Q _i ) of several real quantities (p ₁ , p ₂ , ... p _X ) are combined to form an overall evaluation index (Q). Method according to one of the preceding claims, characterized in that the evaluation indices (Q _i ) and / or the overall evaluation index (Q) are compared with a predetermined evaluation threshold (T _i ) or (T) and an individual or overall evaluation result is output, wherein for Q _i or Q <T _i and T, the evaluation result is "NEGATIVE", for Q _i or Q = T _i or T the evaluation result is "NEUTRAL" and for Q _i or Q> T _i or T the evaluation result Is "POSITIVE". A method according to claim 3, characterized in that the evaluation threshold (T _i ) is based on the product of the arithmetic mean of all the evaluation quantities (Z _j ) of the matrix dimension (i) of the n-dimensional matrix and the total number of values (N) of the real size (p _i ) for the respective real size (p _i ) is determined. Method according to Claim 3 or 4, characterized in that at least one predetermined evaluation parameter (E _i ) for mapping a modified evaluation threshold (T _i ) is taken into account when determining the evaluation index (Q _i ). A method according to claim 5, characterized in that the evaluation parameters (E _i) is modified in dependence of the deviation of the evaluation index (Q _i) of the evaluation threshold (T _i). Method according to claim 6, characterized in that the evaluation parameter (E _i ) is raised for Q _i > T _i and / or the evaluation parameter (E _i ) is reduced for Q _i <T _i . Device for determining an evaluation index (Q _i ) of N> 1 values (N ₁ , N ₂ , ... N _N ) of at least one real variable (p ₁ , p ₂ , ... p _X ) to be evaluated, comprising: a. Means for recording the values (N ₁ , N ₂ , ... N _N ) of the real quantity (p _i ); b. Memory means comprising an n-dimensional memory matrix, the number of matrix dimensions corresponding to the number (X) of the real variables to be evaluated (p ₁ , p ₂ , ... p _X ); c. Means for assigning the values (N ₁ , N ₂ , ... N _N ) of the real variable (p _i ) to a plurality of k> 1 discrete memory cells (a _{1, i} , a _{2, i} , ... a _{k, i} ) the matrix dimension (i) of the n-dimensional memory matrix; d. Means for assigning a discrete evaluation quantity (Z _j ) to each memory cell (a _{1, i} , a _{2, i} , ... a _{k, i} ) of the matrix dimension (i) of the n-dimensional memory matrix; e. Counting device for determining the number (A _j ) of the values assigned to each memory cell (a _{1, i} , a _{2, i} , ... a _{k, i} ) of the matrix dimension (i) of the n-dimensional memory matrix and means for forming the memory cell evaluation index ( q _j ) by multiplying the number (A _j ) of the values assigned to each memory cell by the evaluation quantity (Z _j ) of the respective memory cell; f. Summation device for determining the weighting index (Q _i ) for the respective real size (p _i ) from the memory cell weighting indices (q _i ) of all the memory cells (a _{1, i} , a _{2, i} , ... a _{k, i} ) of the matrix dimension (i ) of the n-dimensional memory matrix; G. Output means for outputting the weighting index (Q _i ); characterized in that the means for assigning the values (N ₁ , N ₂ , ... N _N ) of the real quantity (p _i ) to the memory cells (a _{1, i} , a _{2, i} , ... a _{k, i} ) the matrix dimension (i) of the n-dimensional memory matrix to assign according to a predetermined distribution pattern, wherein the number (k) of the memory cells of each matrix dimension is predetermined by the distribution pattern and the distribution pattern is carried out such that values of at least a first closed value range of real size (p _i ) are assigned to a first memory cell (a _{1, i} ) and values of at most two non-closed value ranges of the real variable (p _i ) are assigned to at most two further memory cells (a _{k-1, i} , a _{k, i} ) become. Device according to Claim 8, characterized in that means are provided for combining the evaluation indices (Q _i ) of a plurality of real variables (p ₁ , p ₂ , ... p _X ) to form an overall evaluation index (Q). Device according to one of claims 8 or 9, characterized in that comparison means are provided which are adapted to compare weighting indices (Q _i ) and / or the total weighting index (Q) with a predetermined weighting threshold (T _i ) or (T) and a single or total evaluation result is output by means of the output device, the evaluation result "NEGATIVE" being output for Q _i or Q <T _i or T, the evaluation result "NEUTRAL" output for Q _i or Q = T _i or T, respectively and for Q _i or Q> T _i or T the evaluation result "POSITIVE" is output. Apparatus according to claim 10, characterized in that means for determining the evaluation threshold (T _i ) for the respective real quantity (p _i ) on the basis of the product of the arithmetic mean of all the evaluation quantities (Z _j ) of the matrix dimension (i) of the n-dimensional Memory matrix and the total number of values (N) of the real size (p _i ). Device according to one of claims 10 or 11, characterized in that means for mapping a modified evaluation threshold (T _i ) by means of a predetermined evaluation parameter (E _i ) are provided. Device according to claim 12, characterized in that the evaluation parameters (E _i) in dependence of the deviation of the evaluation index (Q _i) of the evaluation threshold (T _i) is modified. Device according to claim 13, characterized in that the evaluation parameter (E _i ) is raised for Q _i > T _i and the evaluation parameter (E _i ) is reduced for Q _i <T _i . A computer program product adapted to implement, when executed on a computer, the method of any one of claims 1 to 7 on the computer. A data carrier containing a computer program which contains program instructions which, when executed on a computer, implement the method according to one of claims 1 to 7 on the computer.