JP2995518B2 - Method and apparatus for automatically constructing learning type abnormality diagnosis algorithm - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for constructing a diagnostic algorithm, and is suitable for constructing a diagnostic algorithm for diagnosing abnormalities of various devices used in automobiles, home appliances, power generators and the like. The present invention relates to a method and apparatus for automatically constructing a learning-type abnormality diagnosis algorithm.

[0002]

2. Description of the Related Art Conventionally, an acoustic emission (AE) method has been known as a method for diagnosing an abnormality of a device. As a diagnostic device using the AE method, for example, Japanese Patent Application Laid-Open Nos. 62-282251 and 62-19
755. In the former device, diagnosis is performed using a diagnostic algorithm specific to a diagnosis target, and in the latter device,
A diagnostic algorithm is constructed using specific AE characteristic parameters.

[0003]

In the above-mentioned prior art, no consideration is given to the versatility of performing high-precision diagnosis on various types of diagnostic objects, and even if these methods are employed, the equipment changes. It is necessary to develop a diagnostic algorithm every time, and there is a problem that the development efficiency is low. That is, in the case of the former device, a signal that has passed through a band of a specific wavelength is extracted from the AE signal, the signal is compared with a reference value, and the breakdown of the bearing is predicted from the comparison result. However, since the reference value differs depending on the diagnosis target, it cannot be used for devices other than the diagnosis target. Further, in the case of the latter device, the nature of the abnormality is mainly classified, and it is difficult to determine the abnormality determination level of the diagnosis target with high accuracy, and the diagnosis is performed using the AE characteristic value that is optimal for the diagnosis target. It is impossible. Therefore, the method applied to both devices has to use a diagnostic device developed for each diagnostic object when performing highly accurate diagnosis. An object of the present invention is to provide a learning-type algorithm automatic construction method and an apparatus thereof capable of performing highly accurate diagnosis on various diagnosis targets.

[0004]

The object of the present invention is to measure values relating to a plurality of diagnostic parameters for a group of diagnostic objects classified into those having normal properties and those having abnormal properties, and measuring each diagnostic parameter. Statistical processing is performed on the values, and a plurality of diagnostic parameters that are expected to be valid parameters are extracted from the diagnostic parameter group based on the processing result, and diagnosis is performed based on the measured values of the extracted valid diagnostic parameters. Determine a determination level for determining the quality of the target property, based on the determination level normal, abnormal algorithm determination method or an apparatus for determining an abnormal product, the determined normal,
Calculate the correct answer rate for abnormal products and, if the correct answer rate does not reach the target correct answer rate, further update the combination of diagnostic parameters and the judgment level until the correct answer rate is obtained, and the correct answer rate reaches the target value. This is achieved by being determined from the number of parameters and the determination level. In addition, the diagnosis of the diagnosis target group whose properties are normal or abnormal is unknown by the constructed algorithm, classified into normal or abnormal products, and reevaluated whether the diagnosis result is valid. This is achieved by correcting only the diagnosis results, storing data for each diagnosis in the knowledge base together with the correct answer samples, and reconstructing the algorithm based on the updated database.

[0005]

In the case where the correct answer rate of normal / abnormal judgment based on the effective parameters extracted by the algorithm construction means based on the statistical processing method and the fuzzy method based on the confidence evaluation and the judgment level does not reach a predetermined target correct answer rate, it is effective. By sequentially updating the combination of the parameters and the determination level and reconstructing the algorithm, it is possible to construct an abnormality diagnosis algorithm with higher determination accuracy. In addition, the obtained data is re-edited based on the knowledge base, and a new diagnosis algorithm is reconstructed based on the selection of an effective parameter having the highest accuracy rate and the optimum judgment level. That is,
As the database of the system itself increases, the system itself learns, and finally, a diagnosis algorithm with a very high normal / abnormal judgment correct answer rate can be constructed.

[0006]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a flowchart showing a processing method of a diagnostic algorithm construction method according to a first embodiment of the present invention. FIG. 2 shows the overall structure of the device according to the present invention. In FIG. 2, the diagnosis target is a rotating body such as a steam turbine, a generator, a water turbine, a rolling mill, and a motor.
0 is classified into those having normal properties and those having abnormal properties. Then, the values of various diagnostic parameters of each diagnostic object 10 are detected by the detector 12. The output signal of the detector 12 is sent to a computer 18 via an amplifier 14 and an A / D converter 16. The computer 18 includes a waveform analysis unit 20, a frequency analysis unit 22, and an arithmetic processing unit 24. The arithmetic processing unit 24 is connected to the display unit 26 and the input / output unit 28. Further, the computer 18 includes an average value calculation unit, a first variance calculation unit, a second variance calculation unit, a variance ratio calculation unit,
It is composed of parameter extracting means, deviation amount calculating means, calculating means, individual certainty calculating means, final total certainty calculating means and judgment level determining means, correct answer rate calculating means, target correct answer rate determining means, and highest correct answer rate calculating means. . The computer 18 statistically processes values relating to a plurality of diagnostic parameters based on the detection signal from the detector 12, and determines a diagnostic level based on the processing result. As the diagnostic parameters, diagnostic parameters relating to waveform characteristics and diagnostic parameters relating to frequency characteristics are used. As the AE parameters relating to the waveform characteristics,
As shown in Table 1, there are an average value, an average peak value, an energy value, and the like. As shown in Table 2, a first peak frequency and its voltage, a rotation primary frequency and its There are a voltage, a secondary rotational frequency and its voltage.

[0007]

[Table 1]

[0008]

[Table 2] For example, if you want to evaluate the detection waveform of the AE signal,
AE parameters relating to waveform characteristics are measured by the method shown in Table 3 based on the detection waveform shown in FIG. The AE parameters relating to the frequency characteristics are measured by the method shown in Table 4 based on the result of the frequency analysis (for example, FFT processing) of the AE detection waveform shown in FIG. 3 (one example is shown in FIG. 4).

[0009]

[Table 3]

[0010]

[Table 4]

In describing the present invention in detail, FIG.
A description will now be given of a statistical processing method and an algorithm construction method using a fuzzy processing method, which are conventionally implemented, using the apparatus configuration shown in FIG. FIG. 5 shows a flowchart of the diagnosis.
First, the plurality of diagnosis targets 10 are divided into normal products and abnormal products,
N normal products and m abnormal products (step 100)
Measurement of various AE parameters is executed for each diagnosis target 10 (step 102). That is, a detector 12 is installed in each diagnosis target 10, a detection signal from each diagnosis target 10 is transferred to the computer 18, and the computer 1
At 8, waveform analysis and frequency analysis of the detection signal are executed, and the results of these analyzes are processed by the arithmetic processing unit 24. The arithmetic processing unit 24 processes the measured value of the AE parameter related to the waveform characteristic and the measured value of the AE parameter related to the frequency characteristic. In performing this arithmetic processing, the following equations (1) to (12) are used. First, the average value of each AE parameter of a normal product is calculated according to equation (1) (step 104).

[0012]

(Equation 1) Next, the variance of the normal product is calculated from the measured value of the normal product and the average value of the normal product according to the equation (2) (step 106). Further, the variance of the abnormal product is calculated from the measured value of the abnormal product and the average value of the normal product according to the equation (3) (step 108). Then, the dispersion ratio is calculated according to the equation (4) as a ratio or a difference between the dispersion of the normal product and the dispersion of the abnormal product (step 110). These processes are performed for each AE parameter, and a plurality of effective parameters are extracted from the dispersion ratio of each AE parameter (step 112). For example, A
Select four E parameters. FIG. 6 shows an example of extracting effective parameters and a display example of the dispersion ratio of each AE parameter.
After the four AE parameters are extracted as effective AE parameters, the deviation amount between the normal product and the abnormal product with respect to the average value of the normal product for each effective parameter is calculated by (5).
Calculation is performed for each diagnosis target 10 according to the equation (6) (step 114). Then, based on the deviation amount from the average value of the normal product, the average deviation amount and the standard deviation of the normal product are calculated according to the equations (7) and (8). Then, the individual certainty factor of each effective AE parameter is determined based on these calculated values (step 116). In determining the individual certainty factor, in the present embodiment, a standard is set for the individual certainty factor R as follows.

[0013]

(Equation 2) FIG. 7 shows the individual confidence for the deviation amount of each of the normal and abnormal samples. The value within the value obtained by the equation shown in Expression 1 is set to 0.1, and its multiple is set to 0.2. Similarly, 0.3,
0.4, and the individual certainty factor is determined. Final total confidence M
_When calculating _S , as a fuzzy inference method,
To calculate the final confidence M _S according Combine function method.
In calculating the final overall confidence M _S is (9) (12) is used from the equation, first, the first overall confidence M ₁ using the individual confidence R ₁ of the first effective parameter (9 ) Determined according to the formula. Next, the second overall confidence M ₂ by using the individual confidence R ₂ of the second effective parameters obtained in accordance with (10). Hereinafter, similarly, by the third overall confidence M ₃ (11) below, obtains the final overall confidence M _S according formula (12). After calculating the final total certainty factor M _S, as shown in FIG. 8, a distribution chart is created in which the horizontal axis represents the final total certainty factor M _S and the vertical axis represents the frequency, and the normal product and the abnormal product are created. If it is separable separating, centered on the determination level of the maximum value and the minimum value of the abnormality article M _S of the normal product M _S. Further, in the case of the mixed type in which the normal product and the abnormal product overlap each other, the intersection of the normal product and the abnormal product is set as the determination level, and the normal product and the abnormal product are determined according to the determination level (step 120). Thereby, a diagnostic algorithm can be constructed. The above is the outline of the conventionally implemented algorithm construction method.

Next, a specific embodiment of the present invention will be described with reference to the flowchart shown in FIG. Even if the algorithm is constructed using the algorithm construction method 50 shown in FIG. 5, the correct / abnormal judgment correct answer rate does not always show the highest value depending on the selected four effective parameters and the diagnosis level. For example, the following case applies to such an example. When the variance of the abnormal product (A) is at the position shown in FIG. 9 with respect to the normal product, and the abnormal product (B) is the prominent data of the abnormal product, the dispersion of the abnormal product is calculated as the abnormal product ( A) is a position shown as a virtual abnormal product (A) + (B) in which the abnormal product (B) is superimposed, and the dispersion ratio K _S indicated by the broken line in FIG. Therefore, the dispersion ratio K _S takes a numerically larger value than the other effective parameters, and is selected as the most effective parameter. As a result, the judgment value for discriminating between a normal product and an abnormal product is at the position shown in the figure, and a part of the abnormal product (A) (shaded portion in FIG. 9) is determined to be a normal product, so that the accuracy rate decreases. Will do. In order to avoid using such parameters, in the present invention, the correct answer rate is calculated as shown in FIG. 1 (step 200), and it is determined whether the target correct answer rate is reached (step 201). . If the target accuracy rate has been reached, the algorithm is determined (step 202). However, if the correct answer rate has not reached the target, the algorithm construction method 50 is instructed to select another effective parameter combination different from the effective parameter selected in step 112 of FIG. Similarly, an instruction to automatically change the determination level in step 120 of FIG. 5 so that the accuracy rate is increased is also added. The above operation is repeatedly performed until the target accuracy rate is achieved. It is important to sequentially select the parameters having a high dispersion ratio K _S and change the parameters. The reason is that a parameter with an extremely low variance K _S cannot be effective, and if, for example, four parameters are randomly extracted from 22 parameters to obtain a correct answer rate, the total number of combinations is 175, There are 560 patterns, and the calculation processing time is enormous. Furthermore, there are cases where the target correct answer rate cannot be achieved even in the above-described processing. In the present invention, the maximum accuracy rate calculation means (step 203) shown in FIG. 1 always records the accuracy rate based on the combination of each parameter and the determination level. If the target accuracy rate cannot be obtained, the maximum accuracy rate is calculated. The combination of each parameter indicating the rate and the determination level are used as an abnormality diagnosis algorithm.

FIG. 10 shows a second embodiment of the present invention.
In the above-described embodiment, the case where the normal product and the abnormal product are classified as the diagnosis target has been described. However, when the diagnosis is performed on the actual line of the diagnosis target 10 in which the normal product and the abnormal product are unknown, the computer 18 is used. In FIG. 10, an effective parameter analyzer 30 is used instead of the waveform analyzer 20 and the frequency analyzer 22 as shown in FIG.
Performs a waveform analysis and a frequency analysis on a specific effective parameter, and outputs the analysis result to the arithmetic processing unit 24. In this way, as shown in FIG. 11, a plurality of valid AE parameters P ₁ , P ₂ , P
_3, the P ₄ only perform the measurement of each diagnostic object 10 (step 300), a deviation amount of each sample relative to normal product average value (step 302), further obtains the individual confidence for each valid parameters (step 304 ), And finally obtain the final total certainty factor M _S (step 306). In the process of determining whether the product is normal or abnormal (step 308), the determination of each diagnosis target 10 is performed using the already determined determination level shown in FIG. 8 in the process flow shown in FIG.

FIG. 12 shows a third embodiment of the present invention.
This is an example in which the function of the embodiment shown in FIG. 1 is further improved.
Initially, the number of normal products and abnormal products selected at the time of algorithm construction is limited. Therefore, when performing diagnosis, if the data rate is within the data value of the selected sample, the correct answer rate is about 10
Although it is close to 0%, if a sample having different data is input, the accuracy rate will decrease. That is, by storing data of a large number of samples in the system, the accuracy rate of diagnosis is improved. In this embodiment, online or offline diagnosis is performed by the algorithm construction method shown in FIG. 1 (step 400).
Whether the determined result (step 402) is appropriate or not is re-evaluated by overhauling the product (step 40).
4) Then, the reevaluation result and the parameter analysis result 500 of each sample are input to the knowledge base 600. Knowledge base 60
At 0, the diagnostic algorithm is rebuilt again with the database 700 of normal and abnormal products updated based on the re-evaluation result. The above processing flow is repeated to update the diagnosis algorithm. That is, if the number of samples increases,
As the number increases, the correct answer rate of the learning control and the reconstructed diagnostic algorithm increases. Note that the knowledge base 60
0 has the following knowledge. 1. Method of extracting effective parameters In the W-system and F-system parameters shown in Tables 1 and 2, when a plurality of parameters of the same system are selected, the parameters are shifted and the correct answer rate is reduced. A plurality of parameters are selected in descending order, and one or more other parameters are replaced with parameters of another system. Furthermore, when the selected parameter includes a family item, the parameter is replaced with a parameter indicating the maximum dispersion rate in another family. Extracting the above-mentioned knowledge into effective parameters (step 11 in FIG. 5)
The processing to be added to the selection condition in the case of 2) is executed. 2. Deletion of protruding data If there is prominent data in the data group stored in the knowledge base and the accuracy rate does not improve, the standard deviation σ of the data is obtained by the following formula. For example, data exceeding 3σ is prominent abnormal data. And delete it from the data group.

[0017]

(Equation 3) 3. As shown in the introduction 13 of a multi-stepped algorithm techniques, for example, data groups of samples are mixed with different properties, if the map of the final overall confidence M _S normally not determined value that can partition the abnormal product is obtained, When the determination level is determined from the knowledge base 600 (step 120 in FIG. 5), the following command is issued. The maximum M _S value position of the normal product shown in the first processing step of Figure 13 the final overall confidence M _S maps on the by effective parameters calculated on the basis of the data of all samples as a determination level, data shown in FIG. 13 hatched Judge the group as abnormal and delete it from all data.
Next, the database 70 in which the remaining data group is updated
0, to reenter decision processing flow again, the map of final overall confidence M _S is made in the second process shown in FIG. 13. A similar process is performed in the third process. In the first processing step, the hit sound generating product is diagnosed as abnormal in the second processing step, and the bearing sound generating product is deleted from the data group. Since normal product and the brush sound generating product reaches the third process can be completely separated, the maximum M _S value and the minimum M _S value of the brush sound products normal product, for example, the midpoint and the determination level. By using the above-described multi-stage algorithm, it is possible to determine the normality / abnormality of samples having different properties or samples having different abnormalities. In the case of diagnosing a product whose normal or abnormal condition is unknown in the production line, if a judgment process is performed using the first to third diagnostic algorithms as shown in FIG. Judgment becomes possible. Executing the process of step 404 in FIG. 12 online each time may cause a decrease in productivity for a mass-produced product or the like. Therefore, the diagnosis is performed when the diagnosis is started up. However, if the knowledge is increased to some extent and the erroneous diagnosis does not occur for any sample, the diagnosis algorithm has been established. Can be excluded.

[0018]

As described above, according to the present invention,
The values of multiple diagnostic parameters are measured for the diagnostic target group, which is divided into normal and abnormal ones, and the measured values are subjected to statistical processing. The diagnostic parameters to be extracted are extracted, the determination level is determined based on the measured values of the extracted effective diagnostic parameters, and the combination of the effective parameters and the determination level are sequentially updated until the target correct answer rate is obtained. Furthermore, the obtained data is re-edited based on the knowledge base, and a new diagnostic algorithm showing the highest accuracy rate is reconstructed. That is, as the database of the system itself increases, the system itself learns. Therefore, finally, a diagnostic algorithm having a very high normal / abnormal judgment accuracy rate can be constructed.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating the operation of a first embodiment of the present invention.

FIG. 2 is an overall configuration diagram of an apparatus to which the present invention is applied.

FIG. 3 is an AE detection waveform diagram.

FIG. 4 is a frequency characteristic diagram of an AE detection waveform.

FIG. 5 is a supplementary diagram for explaining the present invention.

FIG. 6 is a display example of a calculated dispersion ratio.

FIG. 7 is a data example of a deviation amount and an individual certainty factor of each sample of a normal product and an abnormal product.

FIG. 8 is a diagram for explaining a determination level determination method.

FIG. 9 is a supplementary diagram for explaining the present invention.

FIG. 10 is a configuration diagram showing a second embodiment of the present invention.

11 is a flowchart illustrating the operation of the device shown in FIG.

FIG. 12 is a flowchart illustrating a third embodiment of the present invention.

FIG. 13 is a diagnosis flowchart for explaining the function of FIG. 12;

FIG. 14 is an example of a diagnosis flow when diagnosing a sample using the third embodiment shown in FIG. 12;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 Diagnosis object 12 Detector 14 Amplifier 16 A / D converter 18 Computer 20 Waveform analysis part 22 Frequency analysis part 24 Operation processing part 26 Display part 28 Input / output part 30 Effective parameter analysis part 600 Knowledge base 700 Updated database

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Yuji Taguchi 3-2-2, Sachimachi, Hitachi City, Ibaraki Prefecture Within Hitachi Engineering Services Co., Ltd. (58) Field surveyed (Int.Cl. ⁶ , DB name) G01N 29 / 00-29/28

Claims (12)

(57) [Claims] 1. A method for measuring values of a plurality of diagnostic parameters for a group of diagnostic objects classified into normal and abnormal ones, performing a statistical process on the measured values of the respective diagnostic parameters, A plurality of diagnostic parameters predicted as valid parameters are extracted from the diagnostic parameter group based on the processing result, and the quality of the diagnosis target is determined based on the measured values of the extracted valid diagnostic parameters. Determine the determination level for, and in the algorithm construction method to determine normal, abnormal product based on the determination level, calculate the determined correct, abnormal product determination correct answer rate, if not reach the target correct rate, further, The combination of diagnostic parameters and the judgment level are sequentially updated until the target correct answer rate is obtained, and the correct answer rate is determined from a plurality of parameters and the judgment level that achieve the target value. Automatic construction method of learning type abnormality diagnosis algorithm, characterized in that: 2. The method according to claim 1, wherein the plurality of diagnostic targets are divided into a normal group and an abnormal group, and values of the plurality of diagnostic parameters are measured for each diagnostic target in each group. Calculate the average of
The variance of the normal group is calculated for each diagnostic parameter from the average value and the measured value, then the variance of the abnormal group is calculated for each diagnostic parameter, and the variance calculated for the normal group and the variance calculated for the abnormal group are calculated. The variance of each diagnostic parameter is calculated from the ratio or difference with the calculated variance, and a plurality of effective diagnostic parameters are extracted from the diagnostic parameter group based on the calculated variance, and each diagnostic target is extracted. The deviation amount between the measured value and the average value of the normal group is calculated for the extracted diagnostic parameter, the deviation amount average value and the standard deviation amount of the normal group are calculated from the calculated deviation amount, and Calculate the individual certainty of the effective diagnostic parameter, calculate the final overall certainty based on each individual certainty, and determine the final total certainty based on the calculated value. In the algorithm construction method for determining the quality of the properties of the diagnosis target, the determined correctness rate of the determined normal and abnormal products is calculated, and if the target correctness rate is not reached, the combination of the diagnostic parameters and the determination level are further set as the target. A learning type abnormality diagnosis algorithm automatic construction method characterized by successively updating until a correct answer rate is obtained, wherein the correct answer rate is determined from a plurality of diagnostic parameters for achieving a target value and a determination level of the overall certainty factor. 3. A diagnostic algorithm according to claim 1 or 2, wherein when a target correct answer rate cannot be achieved, a diagnostic algorithm is provided having a combination of a diagnostic parameter indicating the highest correct answer rate and a determination level. Automatic construction method of learning type abnormality diagnosis algorithm. 4. The diagnostic parameter according to claim 1, wherein the diagnostic parameters selected first are in the order of higher variance, and are sequentially updated, and the remaining diagnostic parameters are also in the order of higher variance. A learning type abnormality diagnosis algorithm automatic construction method characterized by the following. 5. The diagnosis according to any one of claims 1 to 4, wherein the diagnosis is performed for a diagnosis target group whose properties of normal or abnormal are unknown by the constructed algorithm, classified as normal or abnormal, and further diagnosed. Is re-evaluated as to whether it is appropriate, corrects the diagnosis results only for the sample to be erroneously diagnosed, stores the data in the knowledge base together with the correct sample for each diagnosis,
An automatic construction method of a learning-type abnormality diagnosis algorithm, comprising reconstructing an algorithm based on an updated database. 6. The learning-type abnormality diagnosis according to claim 5, wherein when data that is more prominent than other data is added to the knowledge base as data, the data is deleted from the database based on a predetermined criterion. Algorithm automatic construction method. 7. A method for measuring a plurality of diagnostic parameters with respect to a group of diagnostic subjects classified into normal and abnormal ones, and performing a statistical process on the measured values of the respective diagnostic parameters. A plurality of diagnostic parameters predicted as valid parameters are extracted from the diagnostic parameter group based on the processing result, and the quality of the diagnosis target is determined based on the measured values of the extracted valid diagnostic parameters. In the algorithm construction apparatus for determining a determination level for determining, normal, abnormal products based on the determination level, means for calculating the determined correct, abnormal product determination correct answer rate, and when the target correct answer rate is not reached, Further, there is provided a means for sequentially updating the combination of the diagnostic parameters and the determination level until a target accuracy rate is obtained. A learning type abnormality diagnosis algorithm automatic construction apparatus characterized by being determined from a constant level. 8. A measuring means for measuring values relating to a plurality of diagnostic parameters with respect to a group of diagnostic objects including those having normal and abnormal properties, and a measuring means for measuring values relating to a plurality of diagnostic parameters. Mean value calculating means for calculating the average value of each of the obtained diagnostic parameters, first variance calculating means for calculating the variance of normal diagnostic targets for each diagnostic parameter, and the measured value of the measuring means and the average value calculating means A second variance calculating means for calculating a variance of an abnormal diagnosis target for each diagnostic parameter from the calculated value of the variance calculating means; Rate calculating means, parameter extracting means for extracting a plurality of effective diagnostic parameters from the diagnostic parameter group based on the calculated value of the dispersion rate calculating means, and a measurement value and an average value calculating means. To each diagnosis Deviation amount calculating means for calculating a deviation amount between the measured value of the elephant and the average value for each effective diagnostic parameter extracted by the parameter extracting means; and a deviation amount of a normal diagnosis target group from the calculated value of the deviation amount calculating means. Calculating means for calculating an average value and a standard deviation; individual confidence calculating means for calculating individual certainty of each valid diagnostic parameter from the calculated value of the calculating means; and final calculating based on the calculated value of the individual certainty calculating means. In an algorithm construction apparatus comprising a final overall certainty calculating means for calculating the total certainty factor, and a determination level determining means for determining a determination level for determining the quality of the property of the diagnosis target according to the calculated value of the final certainty factor calculating means,
Correction rate calculation means for calculating the correct answer rate for normal / abnormal product judgment by the created algorithm, and judge whether the correct answer rate has reached the target correct answer rate. If the correct answer rate does not reach the target correct answer rate, the target correct answer rate is determined. An automatic learning type abnormality diagnosis algorithm constructing apparatus comprising target correct answer rate determining means for issuing a command for sequentially updating the obtained combination of diagnostic parameters and the final overall certainty determination level. 9. The highest accuracy rate calculation means according to claim 7 or 8, wherein when a target accuracy rate cannot be achieved, a diagnosis algorithm is provided by a combination of a diagnostic parameter indicating the highest accuracy rate and a determination level. A learning type abnormality diagnosis algorithm automatic construction apparatus characterized by having: 10. A method according to any one of claims 7 to 9, wherein the means for executing diagnosis of a diagnosis target group whose normal or abnormal property is unknown by the constructed algorithm,
Means for classifying the product as abnormal and re-evaluating the validity of the diagnostic result, and for correcting the diagnostic result only for the erroneous diagnosis target sample and storing data in the knowledge base together with the correct sample for each diagnosis are updated. A learning-type abnormality diagnosis algorithm automatic construction device, comprising means for reconstructing an algorithm based on a database. 11. A learning method according to claim 10, further comprising means for deleting data from the database based on a predetermined criterion when data that is more prominent than other data is added to the knowledge base as data. Automatic construction system for pattern abnormality diagnosis algorithm. 12. The knowledge base according to claim 10, wherein
When the judgment level for judging the quality of the property cannot be determined from the obtained final overall certainty, the maximum value of the final overall certainty of the normal product is set as the determination level, and the first diagnosis for discriminating between a normal product and an abnormal product. Algorithm, and then delete the diagnostic target group determined to be abnormal by the first diagnostic algorithm, construct a diagnostic algorithm again from the remaining diagnostic target group, and similarly calculate the maximum value of the final total certainty factor of the normal product. A learning-type abnormality diagnosis algorithm automatic construction apparatus characterized in that it has, as knowledge, a second diagnosis algorithm that determines a normal or abnormal product as a determination level, and a multi-stage diagnosis algorithm that performs the same processing.