JPH0765068A - Quality data analyzer - Google Patents

Abstract

(57) [Summary] [Purpose] The purpose is to enable automatic detection of the increasing tendency of the failure rate for each failure mode or failure part, and to predict the number of failures in the future. [Structure] Repair information Shipment information storage unit 1 sums up the number of operating machines and the number of failures by failure month for each failure mode or failure part. Χ ² test section 3 to check whether there is
And a regression analysis unit 4 that analyzes the correlation between the failure month and the failure rate, and a determination unit 5 to which the output signal of the χ ² test unit 3 and the output signal of the regression analysis unit 4 are input. .

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a quality data analyzing apparatus for detecting failure modes of respective parts of an apparatus or increasing tendency of failures for respective parts and predicting future failure occurrence.

[0002]

2. Description of the Related Art A conventional quality data analysis method will be described with reference to FIGS. Figure 5 shows the conventional quality data analysis procedure.
Will be described.

S1. The database stores quality information such as the number of failures and the number of operating units for all the devices shipped so far.
Here, the number of failures refers to the number of devices in which a failure has occurred. S2. From the accumulated information in the database, the number of operating devices and the number of failures, that is, the number of devices in which a failure has occurred are totaled every month to calculate the failure rate for the total number of operating devices every month. Here, the entire device refers to the number of operating devices of the same model. In the following description, the failure rate is referred to as the failure rate of the entire device.

S3. Then, the failure rate of the entire device is compared with a preset target value. S4. Next, with respect to a device in which the failure rate of the entire device exceeds the target value, the failure mode (for example, the tape is twisted or the tape is cut in the case of the magnetic tape device) and parts (in the case of the magnetic tape device, For example, drive motor or key)
For example, by manually inputting data such as the number of failures and the number of operating machines, the bias of failures and the prediction of future failures are analyzed.

FIG. 6 is an explanation of the failure rate for each month, where the horizontal axis is the failure month and the vertical axis is the failure rate. This graph shows a case where the failure rate B of the entire apparatus is equal to or less than the target value C, but the failure rate A of a specific failure mode increases every month.

[0006]

SUMMARY OF THE INVENTION The above-mentioned conventional devices have the following problems. (1) When the failure rate of the entire device exceeds the target value, the failure analysis for each failure mode and each part is performed for the first time. Therefore, it took a long time to collect and input data for this analysis.

(2) Even when the failure rate of the entire apparatus does not exceed the target value, if only a specific failure mode or a failed component increases as shown in FIG. 6, the specific failure mode or the failed component. If the detection of the abnormality is delayed and the specific failure mode or the effect of the failed component is large, it may cause a great deal of trouble to the customer.

In order to solve such a problem, the present invention makes it possible to automatically detect an increase tendency of each failure rate for each failure mode and each part, and to predict the number of future failures. The purpose is to do.

[0009]

[Means for Solving the Problems] The present invention is configured as follows to achieve the above object. FIG. 1 is a diagram illustrating the principle of the present invention. In FIG. 1, 1 is a repair information shipping information storage unit, 2 is an operating unit failure number counting unit by occurrence month, and 3 is χ.
² (Chi-square) test unit, 4 is a regression analysis unit, and 5 is a determination unit.

The repair information shipping information storage unit 1 stores data such as the date of failure, the manufacturing date, the failure mode, and the failed parts as repair information for each shipped device, and the shipping month number as shipping information for all devices. Data on the number of units in operation and the number of units manufactured per month are input and accumulated in a timely manner.

The number-of-operations failure count totaling unit 2 for each month of operation collects the number of operations and the number of failures for each failure mode or for each part in which a failure has occurred for each month in which the failure has occurred. χ ²
The test unit 3 performs a significant difference test by a statistical method, and includes a failure for a certain past period (for example, 6 months) and a failure for a recent certain period (for example, 6 months) consecutive to the past certain period. This is to test whether there is a significant difference in disability for a certain period of time.

The regression analysis unit 4 analyzes the correlation between the failure occurrence month and the failure rate in the same period as the χ ² test unit 3 (the past fixed period and the recent fixed period). Judgment unit 5
Is for determining whether the failure rate is increasing or decreasing.

The above structure will be described with reference to FIG. From the data in the repair information shipping information storage unit 1, the number of operating failures by month according to the number of failure occurrences is calculated by the counting unit 2 for each failure mode of the device shipped or for each part by the year and month of failure occurrence.

Next, the χ ² test section 3 divides into a failure for a certain past period and a failure for a recent certain period on the basis of the aggregated data of the number-of-operations-failures-by-occurrence-number-by-month basis 2 and divides the faults into a recent certain period. Test whether there is a significant difference in disability.

Further, in the regression analysis unit 4, the correlation between the failure occurrence month and the failure rate in the same target period as the test of the χ ² test unit 3 is based on the collected data of the number of operating units failure count counting unit 2 by occurrence month. Do a relationship analysis.

In the next judging unit 5, if there is a significant difference in the test of the χ ² test unit 3 and there is a positive correlation in the analysis of the regression analysis unit 4, that is, there is an increasing tendency, the failure rate shows an increasing tendency. Therefore, an alarm is output. Further, when there is a significant difference in the test of the χ ² test unit 3 and there is a negative correlation in the analysis of the regression analysis unit 4, there is a tendency that the obstacles are decreasing. In this case, the output of the determination unit 5 is Indicates that the determined failure mode or failure of the failed component has been improved.

[0017]

As a result, the determination section 5 can detect the failure rate that tends to increase, so that the failure countermeasure can be taken early and appropriately.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of the present invention will be described with reference to FIGS. 2 is a block diagram of an embodiment of the present invention, FIG. 3 is an explanatory diagram of an analysis state of the present invention, and FIG. 4 is an explanatory diagram of a predicted state of the number of failures in the present invention.

In FIG. 2, reference numeral 1 is a repair information shipping information storage unit, 2 is an operating unit failure number counting unit for each occurrence month, and 3 is χ.
² (Chi-square) verification unit, 4 regression analysis unit, 5-1 alarm determination unit, 6 operating unit failure count totaling unit by operating month, 7
Indicates a reliability determination unit, 8 indicates a failure number prediction unit, and 9 indicates an operating unit failure number counting unit by manufacturing month. The same parts as those in FIG. 1 are designated by the same reference numerals.

An embodiment of the present invention will be described below with reference to FIG. The repair information shipping information storage unit 1 stores data such as the date of failure, the manufacturing date, the failure mode, and the failed component as repair information for each shipped device, and the shipping month number, operating number, and manufacturing as shipping information for all devices. The monthly data is input and accumulated in a timely manner.

The number-of-operations-failures counting unit 2 for each month of occurrence counts the number of operating machines and the number of failures for each failure mode or for each part in which a failure has occurred for each month in which the failure has occurred. The χ ² test unit 3 performs a significant difference test by a statistical method, and divides into a past certain period and a recent certain period continuous to it, and a past certain period to one month consecutively, and It tests whether there is a significant difference in a fixed period or in the past fixed period for one consecutive month.

The regression analysis unit 4 analyzes the correlation between the failure occurrence month and the failure rate in the same period as the χ ² test unit 3. The alarm determination unit 5-1 outputs an alarm when the χ ² test unit 3 tests the significant difference and the regression analysis unit 4 analyzes that there is a positive correlation. Note that, as will be described later, when it is analyzed that there is a negative correlation, it is possible to output that the failure tends to decrease.

The number-of-operations-failures counting unit 6 for each operating month collects the number of operating machines and the number of failures for each operating month for each failure mode or a part in which a failure has occurred. Reliability determination unit 7
Is to determine the cumulative distribution function by calculating the cumulative hazard of the failure rate for each operating month after shipping the device.

The failure number predicting section 8 predicts the number of future failures from the cumulative distribution function and the number of devices shipped. The number-of-operations failure counting unit 9 for each manufacturing month calculates the failure rate for each manufacturing month by totaling the number of operating machines and the number of failures for each manufacturing mode for each failure mode or failed component.

(1) Alarm determination The alarm determination for each failure mode or each failure component is performed in the following order. First of all, the number-of-operating-number-of-failures-number-of-operating-monthly occurrences from the repair information shipping information storage unit 1 counts the number of operating units and the number of failures by month of occurrence of the failure for each failure mode or part of the shipped device.

Next, the χ ² test section 3 and the regression analysis section 4 respectively perform the χ ² test and regression analysis, which will be described later, based on the above aggregated data. In the alarm determination unit 5-1, the χ ² verification unit 3
When it is tested that there is a significant difference, and when the regression analysis unit 4 analyzes that there is a positive correlation, an alarm is output.

Next, the operations of the χ ² test unit 3 and the regression analysis unit 4 will be described in more detail. I. Description chi ² test section 3 of the chi ² test unit 3 performs a chi ² test.

The χ ² test is carried out separately in the following two ways for each failure mode or failed part of the shipped device. Past certain period (eg 6 months) and recent consecutive period (eg 6 months) Past certain period (eg 6 months) and consecutive 1
Months In the above case, a test can be performed when the failure rate is gradually increasing or decreasing, and in the case of, a test can be performed when the failure rate is rapidly increasing or decreasing.

The contingency table for performing the χ ² test is as follows.

[0030]

[Table 1]

In Table 1, r₁Is an obstacle for a certain period in the past
Number of damages, r₂Is the number of failures (or
The number of disability cases for the last 1 month in a row over a certain period), R is the total
Total (r₁+ R₂), A₁Is the number of non-defective products in a certain period in the past,
a₂Is the number of non-defective items in a certain period of time (or past regular period)
The number of non-defective products in one month in a row, A is the total (a₁+
a₂), K₁Is the number of units in operation for a certain period in the past, k₂Recently
Operating units for a certain period of time (or
The number of operating machines for one month), K is the total (k₁+ K ₂)
It

The test formula of the contingency table is as follows.

[0033]

[Equation 1]

Here, the above-mentioned numerical value 2.705 is a reference value when the risk rate is 10% and the degree of freedom is 1 in the χ ² distribution table. The reference value (set value) can be changed depending on the situation.
If the value calculated on the left side of Equation 1 exceeds the reference value (2.705 in the above case), it means that there is a disability during the recent certain period (or one consecutive month in the past certain period). Since there is a significant difference, the χ ² test unit 3 gives an output indicating that there is a significant difference to the alarm determination unit 5-1.

B. Explanation of Regression Analysis Unit 4 The regression analysis unit 4 targets the same failure mode or failure component for the same period as in the case of the above χ ² test, and the correlation between the failure month and the failure rate is correct by the regression analysis. Analyze whether it is negative. Then, when the correlation has a positive correlation,
The failure analysis shows an increasing tendency, and the regression analysis unit 4 gives an output indicating a positive correlation to the alarm determination unit 5-1.

If the correlation has a negative correlation, it indicates that the failure tends to decrease, and the regression analysis unit 4 gives an output indicating the negative correlation to the alarm determination unit 5-1. To be

The alarm determination section 5-1 outputs an alarm when a significant difference is tested by the χ ² test and a positive correlation is analyzed by the regression analysis. Next, regarding the failure mode or the failed part in which the above-mentioned alarm is output, the failure abnormality is detected by performing the following (2) prediction of the number of future failures and the following (3) aggregation of failure rates by manufacturing month. To confirm.

(2) Prediction of the number of future failures The number of operating failures by operating month In the failure counting unit 6, the repair information shipping information sorting section 1 starts operating for the failure mode or failed part for which an alarm was output in (1) above. The monthly number of operating units and the number of failures are totaled.

Next, the reliability determining unit 7 obtains a cumulative hazard (failure rate) after the device is shipped based on the above-mentioned aggregated data, and determines the reliability as follows. The cumulative hazard is calculated by accumulating the failure modes of the devices shipped in the past in which the alarms have been output or the failure rates of the parts in each month for each elapsed month. An example of a graph of this cumulative hazard is the upward-sloping cumulative distribution curve in FIG. Then, this cumulative distribution curve can be easily expressed by a mathematical expression.

Cumulative distribution function F (t) indicating cumulative hazard
Can be expressed, for example, by the following Expression 2.

[0041]

[Equation 2]

Here, m is a shape parameter, γ is a position parameter indicating the time when a failure starts to occur, t ₀ is a scale parameter (for example, 63% of the time is a failure),
t indicates time.

The function F (t) of the above equation 2 is called a Weibull function, and the reliability can be determined by the value of the shape parameter m as follows. FIG. 3B shows a change curve of the failure rate depending on the value of m.

When m <1, it indicates an initial failure type failure, and the failure rate decreases as the operating time elapses. This shows that the reliability increases as the operating time elapses.

When m = 1, an accidental failure is shown, the failure rate is always constant, and the reliability does not change. m
When> 1, the wear-type failure was indicated, and the failure rate increased as the operating time increased. This indicates that the reliability decreases as the operating time elapses.

In the failure number predicting unit 8, the future cumulative hazard for each failure mode or each failure unit is obtained from the cumulative distribution function F (t) of the equation 2. Then, the number of future failures can be predicted from the obtained cumulative hazard and the number of shipped devices (or the same device scheduled to be shipped in the future).

As an example, FIG. 4 shows January and 2 respectively.
For the devices shipped in the month and March, the expected number of failures due to the passage of time (time) is calculated from the number of units shipped each month and the cumulative distribution function F (t) for January shipment b and February shipment c. Three
Predicted like monthly shipment d, and total failure count e (e =
b + c + d) shows the obtained prediction.

In FIG. 4, the number of failures is predicted when m of the cumulative distribution function F (t) determined by the reliability determination unit 7 is m> 1. Here, the curves b to d of the number of failures for January to March shipments increase first and then decrease as compared with the failure rate increasing with time when m> 1, as described above. . This decrease indicates that when a failure occurs, the number of operating machines decreases accordingly, so that the number of failures decreases conversely even if the failure rate increases.

(3) Aggregation of Failure Rate by Manufacturing Month In the operating quantity failure number accumulating section 9 by manufacturing month, the failure mode or the failure part for which an alarm is output from the repair information shipping information storage section 1 in the above (1), The operating rate and the number of failures for each manufacturing month are totaled and the failure rate for each manufacturing month is output. Accordingly, it is possible to investigate an abnormality (bias) in the failure rate in the manufacturing month, and analyze the abnormality in the manufacturing month.

As described above, according to the present invention, by inputting the repair information and the shipping information data into the repair information shipping information storage unit 1 in a timely manner, the alarm determination in (1) above and the future failure in (2) are performed. Quality data analysis such as forecasting the number of occurrences and summarizing the failure rate by manufacturing month in (3) is automatically performed.

Although one embodiment has been described above, the present invention is not limited to this, and the present invention can be implemented as follows. In the above-described embodiment, the failure mode or the failed part in which the alarm is output from the alarm determination unit 5-1 is estimated in the future number of failures and the failure rate for each manufacturing month is aggregated, but regardless of the alarm output. Forecasting the number of future failures,
And the failure rate by manufacturing month can be aggregated independently.

The operating-number failure count totaling unit 2 for each month of occurrence, the operating-unit failure count totaling unit 6, and the operating-unit failure count totaling unit 9 for each manufacturing month can each have a graph creating function.

Although the fixed period in the χ ² test is set to 6 months, the fixed period is not limited to this, and can be set to 3 months, 1 year, or every fixed day. In the χ ² test, the past certain period and the recent certain period that was continuous to it were set as the past certain period, but it is not necessary to make the past certain period and the recent certain period consecutive, and the χ ² test is performed by opening these periods. It can be carried out.

[0054]

As described above, the present invention has the following effects. By examining the increasing tendency of the failure for each failure mode or failure component, it is possible to take an improvement measure for the failure early.

By predicting the number of future failure occurrences, it is possible to surely confirm the abnormality. By collecting the failure rates for each manufacturing month, it is possible to confirm the deviation of failures in the manufacturing month.

If there is a significant difference in the χ ² test and there is a negative correlation in the regression analysis, it is possible to confirm the tendency of the failure mode or the countermeasure for improving the failed component. Repair information Shipment information storage unit 1 database that enables automatic input of data to the monthly operating device failure count counting unit 2, operating month operating unit failure count counting unit 6 and manufacturing month operating unit failure count counting unit 9 By doing so, each analysis can be automated.

[Brief description of drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a configuration diagram of an embodiment of the present invention.

FIG. 3 is an explanatory diagram of an analysis state of the present invention.

FIG. 4 is a diagram illustrating a predicted state of the number of failures in the present invention.

FIG. 5 is an explanatory diagram of a conventional example.

FIG. 6 is an explanatory diagram of a conventional monthly failure rate.

[Explanation of symbols]

1 Repair information Shipment information storage section 2 Occurrence number of troubles counting section by occurrence month 3 χ ² Test section 4 Regression analysis section 5 Judgment section 5-1 Alarm judgment section 6 Operation number of troubles counting section by operation month 7 Reliability judgment section 8 Failure Number of incidents forecasting section 9 Total number of operating failures by manufacturing month

Claims (4)

[Claims] 1. A repair information shipping information storage unit (1) for storing repair information and shipping information, and a malfunction mode or a malfunctioning number for each malfunctioning part for each malfunctioning part from the repair information shipping information storage unit (1). Whether there is a significant difference in the number of operating failures by number of occurrences per month (2) and whether there is a significant difference in the failure mode or failure parts during the recent fixed period for the past fixed period and the recent fixed period. assaying chi ² test unit (3) and, the chi ² test unit regression analysis section for analyzing the correlation between the failure mode addition or failure month for each failed component and failure rates of the period of (3) ( 4), a determination unit (5) to which the output signal of the χ ² test unit (3) and the output signal of the regression analysis unit (4) are input, the quality data analysis device. 2. An alarm determination unit that outputs an alarm when the χ ² test unit determines that there is a significant difference and the regression analysis unit analyzes that there is a positive correlation, and an output of the alarm determination unit. A total number of operating units for each operating month and the number of failures by operating month for each of the failure modes or failed parts from the repair information shipping information storage section; and a cumulative hazard for each failure mode or each failed part. The quality data analyzer according to claim 1, further comprising: a reliability determination unit that determines the reliability, and a failure number prediction unit that predicts a future failure number from the cumulative hazard of the determined reliability determination unit. 3. An alarm determination that outputs an alarm when the χ ² test unit (3) tests for significant difference and the regression analysis unit (4) analyzes for positive correlation; According to the output of the determination unit, the repair information shipping information storage unit is provided with an operating unit failure number counting unit for each manufacturing month for counting the number of operating units and the number of failures for each of the failure modes or failed parts by manufacturing month. The quality data analyzer according to claim 1. 4. A repair information shipping information storage section (1) for storing repair information and shipping information, and a working number and a failure number for each failure mode or a failure part or an operating month for each failed part from the repair information shipping information storage section (1). And a reliability determination unit (7) that determines the cumulative distribution of the failure rate for each operating month for each failure mode or for each of the failure modes, and the reliability determination From the cumulative distribution determined by the department (7), the number of failures in the future is calculated for each shipping time, and the number of failures calculated is added up for each time to predict the number of failures. A quality data analyzer characterized by: