JP2002183250A - Analyzer in manufacturing process, method of analysis in manufacturing process and computer-readable storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To analyze the correlation between process control data and quality data, which can not be obtained by a scatter diagram or a correlation coefficient. SOLUTION: In a iron and steel process, the range of values that can be taken by a manufacturing line velocity V in a certain manufacturing period is divided into L pieces of ranges, the probability distribution Pi of the number N of surface faults is calculated in each range (i), and the number NiPc of fault occurrences, where a prescribed cumulative probability in each range (i) becomes Pc is calculated. The relation between the manufacturing line velocity V and the number NiPc of surface faults to be the prescribed cumulative probability is expressed by an approximate expression or a table, an appropriate manufacturing line velocity to obtain a product with few surface faults is obtained and the maximum number of surface faults of a product, which occurs in the probability Pc is predicted in the case of being a certain manufacturing line velocity by using the approximate expression or the table.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an analyzer, a method, and a computer-readable storage medium in a manufacturing process for analyzing a correlation between quality data in a steel process or the like and process operation data.

[0002]

2. Description of the Related Art As a general method for finding a relationship between two data, such as a relationship between process operation data and quality data, observation and evaluation of a scatter diagram or evaluation by a correlation coefficient is performed. According to these methods, when there is a linear or curved relationship between two data, in a scatter diagram, the dispersion of points concentrates around the straight line / curve, and the phase relationship is increased. In terms of numbers, the relationship between the data is clarified by the absolute value of the linear correlation coefficient and the absolute value of the curve correlation coefficient (that is, a value close to 1).

As a method for predicting the quality in a steel process or the like, for example, there is a method disclosed in Japanese Patent Application Laid-Open No. 6-304723. There, quality control diagnosis is performed by inputting process operation data and quality data to a neural network and learning the neural network.

[0004] As a quality control / prediction method using a probability distribution, a property that a randomly generated surface quality defect is approximated to a Poisson distribution is used. A method of estimating the product yield from the number is known.

[0005]

However, depending on the quality data, there is a case where each process operation data does not have a linear or curvilinear relationship. In this case, even if the scatter diagram and the correlation coefficient are evaluated, there is a problem that the two are determined to have low correlation and the relationship between the two cannot be grasped.

In the method disclosed in Japanese Patent Application Laid-Open No. 6-304723, as process operation data, physical property values such as the carbon amount of a slab, continuous casting operation values such as a sheet width, and the temperature of each cooling zone are described. In addition, presence / absence of a surface defect is input as quality data. However, in an actual steel process, the causes of surface defects are numerous, and it is often difficult to artificially set or measure them. In this case, the quality data appearing as a result of the operation also includes uncertainty, and the quality data is given as two values, that is, whether or not there is a surface defect, and the relationship with the input process operation data is learned. However, there is a problem that it is not always possible to obtain a sufficiently accurate learning result.

In a very simple manufacturing process, the number of surface defects can be approximated by a Poisson distribution. However, in a real manufacturing process that is complicated over many steps, for example, a steel process or a semiconductor process, etc. However, the number of occurrences of surface defects in the final product does not always indicate a Poisson distribution, and there is a problem in that quality-related information cannot be represented only by the average number of occurrences of surface defects.

The present invention has been made in view of the above points, and has as its object to analyze a correlation between process control data and quality data which cannot be captured by a scatter diagram or a correlation coefficient. And

[0009]

SUMMARY OF THE INVENTION An analyzing apparatus in a manufacturing process according to the present invention is an analyzing apparatus in a manufacturing process for analyzing a correlation between process operation data and quality data. Dividing means for dividing the range into a plurality of ranges; probability distribution calculating means for obtaining a probability distribution of quality data for each of the ranges; and quality data value for calculating a quality data value having a predetermined cumulative probability in each of the ranges. It is characterized in that it has a calculating means.

Another feature of the analyzing apparatus in the manufacturing process of the present invention is that the probability distribution calculating means includes:
The point is that the probability distribution of the quality data for each range is approximated using a probability density function representing an exponential distribution.

Another characteristic of the analyzer in the manufacturing process according to the present invention is that the process operation data and the quality data value which is a predetermined cumulative probability in each range calculated by the quality data value calculating means. And an approximate expression calculating means for expressing the relationship with the approximate expression.

Another feature of the analyzing apparatus in the manufacturing process of the present invention is that a new process operation data is generated at a predetermined probability using the approximate expression obtained by the approximate expression calculating means. In this case, there is provided a prediction unit for obtaining the maximum value of the predicted quality data.

Another feature of the analyzing apparatus in the manufacturing process of the present invention is that the process operation data and the quality data value which is a predetermined cumulative probability in each range calculated by the quality data value calculating means. And a table storage means for expressing and storing these as a table.

Another feature of the analyzing apparatus in the manufacturing process of the present invention is that a new process operation data is generated at a predetermined probability using a table stored in the table storage means. The point is that a prediction means for obtaining the maximum value of the predicted quality data is provided.

Another feature of the analyzer in the manufacturing process of the present invention is that it is applied to a steel process, and the quality data is the number of defects per unit area of a product surface.

An analysis method in a manufacturing process according to the present invention is an analysis method in a manufacturing process for analyzing a correlation between process operation data and quality data,
A process of dividing a range of possible values of the process operation data into a plurality of ranges, a process of obtaining a probability distribution of quality data for each of the ranges, and calculating a quality data value that is a predetermined cumulative probability in each of the ranges. This is characterized in that the following processing is performed.

The computer-readable storage medium of the present invention is characterized in that a program for causing a computer to function as each of the above means is stored.

Another computer-readable storage medium according to the present invention is characterized in that a program for causing a computer to execute the above-described processing is stored.

In the present invention as described above, for example, in a steel process, the range of values that can be taken by process operation data such as the production line speed during a certain production period is divided into a plurality of ranges, and the surface defects for each range are divided. The probability distribution of the quality data such as the number of surface defects is obtained, and the number of surface defects having a predetermined cumulative probability in each range is calculated. Then, the relationship between the production line speed and the number of surface defects having the predetermined cumulative probability is represented by an approximate expression or a table.

[0020]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an embodiment of an analyzing apparatus, a method, and a computer-readable storage medium according to the present invention. In the present embodiment, in the steel process, the production line speed, which is one of the process operation data, and the number of defects per unit area of the product surface, which is quality data (hereinafter, referred to as “the number of surface defects”). Analyze the relationship of
An example will be described in which the number of surface defects predicted to be generated with respect to new process operation data is obtained using the relationship. Here, the surface defect is intended to be generated due to bubbles, inclusions, powder, and the like contained in the steel slab.

FIG. 1 is a diagram showing a configuration of an analyzer in the manufacturing process of the present embodiment. In the figure, 1
Reference numeral 01 denotes a data input unit, from which process operation data including operation setting values and measured values during a certain manufacturing period and surface defect data are input from a data storage unit (not shown). In the present embodiment, the production line speed is input as process operation data, and the number of surface defects is input as quality data.

Reference numeral 102 denotes a probability distribution calculation unit which divides a range of possible values of process operation data (manufacturing line speed) into a plurality of ranges and calculates a probability distribution of quality data (the number of surface defects) for each range. Then, a quality data value (the number of surface defects) that becomes a predetermined cumulative probability in each range is calculated.
At this time, quality data for each range (number of surface defects)
Is approximated using a probability density function representing an exponential distribution.

Reference numeral 103 denotes a correlation analysis unit which calculates an approximate expression indicating a relationship between the process operation data (manufacturing line speed) and the quality data value (the number of surface defects) that provides the above-mentioned constant cumulative probability. A correlation display unit 104 displays a result calculated by the correlation analysis unit 103.

Numeral 105 denotes a defect occurrence predicting unit which uses the approximation formula obtained by the correlation analyzing unit 103 to generate quality data predicted to occur at a predetermined probability with respect to new process operation data (manufacturing line speed). (The maximum number of surface defects) is predicted. A prediction result display unit 106 displays a result predicted by the defect occurrence prediction unit 105.

Reference numeral 107 denotes a histogram calculation unit that calculates a histogram from the process operation data (manufacturing line speed) and the quality data (the number of surface defects) input to the data input unit 101. As described above, the probability distribution calculation unit 102 performs approximation using a probability density function representing an exponential distribution.
7 can be determined based on the histogram calculated.

Next, the processing operation of the analyzer in the manufacturing process of this embodiment will be described with reference to the flowchart shown in FIG. When the manufacturing line speed V and the number N of surface defects during the manufacturing period are input to the data input unit 101, the histogram calculator 107 calculates a histogram (step S201).

FIG. 3 is a scatter diagram showing the relationship between the production line speed V and the number N of surface defects during a certain production period. From the scatter plot, it is difficult to find the correlation between the two. Also, the absolute value of the primary correlation coefficient at this time, that is, how close the two are to a linear relationship is 0-1.
Is a low value of 0.1, indicating that there is almost no correlation between the two.

Here, a histogram of the frequency of the number N of surface defects (how many plots exist) in FIG. 3 is obtained as shown in FIG. As shown by the dotted line in the figure, it is determined that this histogram can be approximated by an exponential distribution.

Therefore, in this case, the probability distribution of the occurrence of surface defects when the production line speed V is determined is approximated by an exponential distribution. That is, assuming that the number N of surface defects and the occurrence probability distribution P are k, parameters can be represented by the following equation (1) using k as a parameter.

[0030]

(Equation 1)

Next, as shown in the following Expression 2, the probability distribution calculation unit 102 divides the production line speed V from the minimum value V _min to the maximum value V _max into a plurality (L) of sections divided by ΔV. It is divided into ranges (step S202).

[0032]

(Equation 2)

[0033] Then, determine the probability distribution P _i of surface defects for each range (step S203). Figure 5 illustrates a production line speed V, and the surface defect number N, the relation between occurrence probability distribution P _i for each range. Each solid line shown in the figure illustrates the probability distribution P _i for each range (in the figure displays the three ranges).

Next, for the occurrence probability distribution P _i in each range i, using a known maximum likelihood method, an equation of a probability density function corresponding to the above equation (1) Equation (3) shown)
Performs fitting to determine a parameter k _i for each range i (step S204). Each dotted line shown in FIG. 5 illustrates a state of performing a fitting on the occurrence probability distribution P _i for each range.

[0035]

(Equation 3)

When the probability of occurrence of surface defects is represented by an exponential distribution as in the present embodiment, the number of occurrences of defects _NiPc corresponding to a certain cumulative probability Pc is _calculated by the following equation (4).
Is represented by

[0037]

(Equation 4)

Here, in each range i, the 80% certainty is obtained.
Number of surface defects occurring at a rate N _i0.8Think about
Step S205). This means that in the above equation (4), Pc =
0.8, which is equivalent to the number of surface defects,
Equation (5).

[0039]

(Equation 5)

Using the above equation (5), N in each range i
_Plots of _iPc are a plurality of black points shown in FIG. This plot point is approximated by a polynomial of an appropriate order in the correlation analysis unit 103 (step S20).
6) The relationship between the production line speed and the number of surface defects can be obtained. In the present embodiment, it can be expressed by a linear expression as shown in the following Expression (6), and the lower the production line speed V, the smaller the maximum number of surface defects that occur with a probability of 80%. You can see that it can be done.

[0041]

(Equation 6)

At this time, the correlation coefficient representing the magnitude of the primary correlation is 0.95, and the strong correlation between the production line speed and the number of surface defects indicates that the apparatus of the present embodiment is used. It became clear.

Therefore, by using the above equation (6) in the defect occurrence prediction unit 105, the number of surface defects generated for a new arbitrary production line speed can be reduced by 80%.
(Step S207).

In this embodiment, the above equation (1)
Although the exponential function distribution is applied as described in (3), appropriate probability distributions such as Poisson distribution, binomial distribution, beta distribution, and gamma distribution are applied according to the tendency indicated by the histogram, and the respective probability density functions are set. And approximate them.

[0045] In addition, the appropriate probability distribution approximated using is not intended to be necessarily performed, the occurrence probability distribution P _i of surface defects in the range i, which are divided according to the above equation (2), directly, the accumulated The defect occurrence number _NiPc having a probability of Pc may be obtained and plotted as shown in FIG. However, by performing the approximation, noise included in the input data can be removed, and a clearer correlation between the process operation data and the quality data can be obtained.

In this embodiment, the points plotted as shown in FIG. 6 are approximated by a linear expression, but they may be approximated by a polynomial or any other expression that is higher than a quadratic expression. further,
Without performing approximation to the expression, the number of occurrences _NiPc of the defect which is the cumulative probability Pc in each range i of the above expression (2) is stored in a memory as a relation table, and is used to predict the number of occurrences of surface defects. You may.

Further, in this embodiment, the production line speed is used as the process operation data. However, other process operation data, such as physical properties such as the carbon amount of the slab, product dimensions such as plate width, product temperature, etc. Can also be used.

In this embodiment, the correlation between one process operation data and the number of surface defects is analyzed. However, the correlation between a plurality of process operation data and the number of surface defects is analyzed. It can also be analyzed. For example, when analyzing the correlation between two process operation data and the number of surface defects, the data is divided into L × L ranges corresponding to the range divided into L in the above equation (2), and each range is divided. By determining N _ijPc at ij, the correlation between the two process run data and the number of surface defects can be expressed.

In this embodiment, an example in which the present invention is applied to a steel process has been described.
For example, the present invention can be applied to analysis in a semiconductor process.

(Other Embodiments) The analyzer in the manufacturing process of the present invention may be composed of a plurality of devices or one device.

The above-described embodiment is constituted by a CPU or MPU of a computer, a RAM, a ROM, and the like, and is realized by operating a program recorded in the RAM or the ROM. Therefore, means for supplying a computer with software program codes for realizing the functions of the above-described embodiments, for example, a storage medium storing such program codes is included in the scope of the present invention.

[0052]

As described above, according to the present invention, in a manufacturing process such as a steel process, the correlation between process operation data and quality data can be analyzed using a probability distribution. It is possible to clarify the correlation between the two, which was not captured by the scatter diagram or the correlation coefficient. Therefore, using the results of the analysis, it is possible to obtain appropriate process operation data for obtaining high-quality products, or to predict the quality of products obtained when certain process operation data is used.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of an analyzer according to an embodiment of the present invention.

FIG. 2 is a flowchart for explaining a processing operation in the analyzer.

FIG. 3 is a scatter diagram illustrating a relationship between a manufacturing line speed V and the number N of surface defects during a certain manufacturing period.

FIG. 4 is a diagram showing a histogram of the frequency of the number N of surface defects in FIG. 3;

[5] and the production line speed V, and the surface defect number N, is a diagram showing the relationship between the occurrence probability distribution P _i for each range.

FIG. 6 is a diagram plotting _NiPc in each range i.

[Explanation of symbols]

Reference Signs List 101 Data input unit 102 Probability distribution calculation unit (division unit, probability distribution calculation unit, quality data value calculation unit in the present invention) 103 Correlation analysis unit (approximate expression calculation unit in the present invention) 104 Correlation display unit 105 Defect occurrence prediction Unit (prediction means in the present invention) 106 prediction result display unit 107 histogram calculation unit

Claims (10)

[Claims] 1. An analyzing apparatus in a manufacturing process for analyzing a correlation between process operation data and quality data, wherein the dividing unit divides a range of possible values of the process operation data into a plurality of ranges; A probability distribution calculating means for obtaining a probability distribution of quality data for each range; and a quality data value calculating means for calculating a quality data value that is a predetermined cumulative probability in each of the ranges. Analysis device. 2. The method according to claim 1, wherein the probability distribution calculating means approximates the probability distribution of the quality data for each of the ranges using a probability density function representing an exponential distribution.
An analyzer in the manufacturing process according to 1. 3. An approximate expression calculating means for expressing a relationship between the process operation data and a quality data value which is a predetermined cumulative probability in each range calculated by the quality data value calculating means by an approximate expression. An analysis device in a manufacturing process according to claim 1 or 2, wherein: 4. A prediction means for calculating a maximum value of quality data predicted to occur at a predetermined probability with respect to new process operation data using an approximation formula obtained by said approximation formula calculation means. The analyzer according to claim 3, wherein 5. A table storage means for storing a table of process operation data and quality data values which are predetermined cumulative probabilities in each range calculated by the quality data value calculation means. An analysis device in the manufacturing process according to claim 1 or 2. 6. A prediction means for obtaining a maximum value of quality data predicted to occur at a predetermined probability for new process operation data using a table stored in the table storage means. An analyzer in the manufacturing process according to claim 5. 7. The analysis in the manufacturing process according to claim 1, wherein the quality data is applied to a steel process, and the quality data is the number of defects per unit area of a product surface. apparatus. 8. An analysis method in a manufacturing process for analyzing a correlation between process operation data and quality data, wherein the range of possible values of the process operation data is divided into a plurality of ranges; An analysis method in a manufacturing process, comprising: performing a process of obtaining a probability distribution of quality data for each of the plurality of processes; and a process of calculating a quality data value that is a predetermined cumulative probability in each of the ranges. 9. A computer-readable storage medium storing a program for causing a computer to function as each of the means according to claim 1. 10. A computer-readable storage medium storing a program for causing a computer to execute each processing according to claim 8.