JP2009266291A - Hard disk drive failure prediction device, hard disk drive failure model generation device, and hard disk drive failure prediction method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To highly accurately detect the abnormality of a hydrodynamic pressure bearing spindle motor in a hard disk drive. <P>SOLUTION: Based on history information storing sensor information obtained from a sensor mounted on a hard disk drive (HDD) 100 relating to a learning phase for a desired period, a deterioration estimation model indicating a relationship between a sensor item and the deterioration degree of lubricant oil of a fluid dynamic pressure bearing spindle motor for rotary-driving a disk, and a HDD generation model indicating a relationship between the time-sequential change of the deterioration degree of the lubricant oil and the endurance probability of the HDD are generated. Then, the sensor information obtained from the sensor mounted on the HDD 101 relating to a prediction phase is applied to the generated deterioration estimation model, the degree of the deterioration of the lubricant oil is estimated, and tendency analysis is performed by the HDD endurance model based on the estimated deterioration of the lubricant oil, thereby predicting the failure time of the HDD 101. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a hard disk drive failure prediction device, a hard disk drive failure model generation device, and a hard disk drive failure prediction method for detecting an abnormality of a fluid dynamic bearing spindle motor in a hard disk drive.

  S.M.A.R.T. (Self-Monitoring, Analysis and Reporting Technology) is a function installed in a hard disk drive for the purpose of early detection of failure of the hard disk drive and prediction of failure. This function self-diagnose various inspection items in real time and digitizes their status. The user using the terminal computer can know the numerical value by using various tools. Not all failures can be predicted, but it is very effective to know failures due to aging in a stable usage environment.

As a failure prediction method using this SMART function, (1) a threshold value is set for each item of the SMART function, and a warning is issued when the threshold value is exceeded (for example, Patent Documents 1 and 2), (2) SMART A method to create a failure prediction model from the item start time, start count, and start retry count, diagnose and predict whether a spindle motor has failed due to external vibration or shock, and issue a warning based on the model output result (For example, patent documents 3), (3) The method (for example, patent documents 4-7) which installs a special sensor in data storage device itself, and judges a failure comprehensively from the sensor value etc. are mentioned.
Japanese Patent Laid-Open No. 2007-213670 JP 2007-335016 A JP 2006-236524 A JP 2005-55296 A JP 2002-272172 A JP 2002-131188 A Japanese Patent Laid-Open No. 11-2669 Eduardo.P. Et al, “Failure Trends in a Large Disk Drive Population”, 2007

  However, the above prior art has the following problems. The method (1) is a method of checking the threshold value of each item, and there is a problem that many error warnings and oversights of failures occur. In the case of the method (2), as described in Non-Patent Document 1, the frequency of start-up retries is extremely low, and failure is mostly caused before start-up retries occur, so that the method lacks effectiveness. . In the case of the above (3), since a special sensor is required, there is a problem that it cannot be applied to many hard disk drives currently distributed. Among them, Patent Document 4 uses current, temperature, and specified rotational speed as items. These three items are items that can be acquired even in some of the current hard disk drives, but the items that are standardized only when the true value of the current cannot be acquired and the value is lower than the rated current. Since the specification is subtracted from the value, there remains a problem that the deterioration of the bearing cannot be expressed appropriately.

  In view of the above-described problems of the prior art, the present invention diagnoses and predicts a failure caused by deterioration of a fluid dynamic bearing spindle motor using sensor items that can be acquired by the SMART function or the like in most hard disk drives. An object of the present invention is to provide a hard disk drive failure prediction device, a hard disk drive failure model generation device, and a hard disk drive failure prediction method.

  A hard disk drive failure prediction apparatus according to the present invention acquires a sensor data from a sensor mounted on a hard disk drive in a learning phase, and stores the acquired sensor information as sensor information. A deterioration estimation model generation unit that generates a deterioration estimation model that represents a relationship between the sensor information and the degree of deterioration of lubricating oil of a fluid dynamic pressure bearing spindle motor that rotationally drives the disk based on history information accumulated for a period of time, and the generation A hard disk drive survival model generation unit for generating a hard disk drive survival model representing a relationship between a change in the time series of the deterioration degree of the lubricant estimated by the determined deterioration estimation model and a hard disk drive survival probability, and a prediction phase The sensor mounted on the hard disk drive A second data acquisition unit that acquires the data and stores it as sensor information, and applies the acquired sensor information to the generated deterioration estimation model to determine the degree of deterioration of the lubricating oil of the hard disk drive according to the prediction phase. A deterioration estimation unit for estimation, a failure analysis unit that applies a trend analysis by applying the estimated deterioration level of the lubricant to the generated hard disk drive survival model, and predicts a failure time of the hard disk drive according to the prediction phase; It is characterized by comprising.

  A hard disk drive failure model generation device according to the present invention acquires a sensor data from a sensor mounted on a hard disk drive according to a learning phase, stores the sensor data as sensor information, and the acquired sensor information for a desired period A deterioration estimation model generation unit that generates a deterioration estimation model that represents a relationship between the sensor item and the degree of deterioration of the lubricating oil of the fluid dynamic bearing spindle motor that rotationally drives the disk based on the accumulated history information; A hard disk drive survival model generation unit that generates a hard disk drive survival model representing a relationship between a change in the time series of the deterioration degree of the lubricant estimated by the deterioration estimation model and the survival probability of the hard disk drive. And

  A hard disk drive failure prediction method according to the present invention includes a first data acquisition step of acquiring sensor data from a sensor mounted on a hard disk drive according to a learning phase, and storing the acquired sensor information as sensor information. A deterioration estimation model generation step for generating a deterioration estimation model representing a relationship between the sensor item and a degree of deterioration of lubricating oil of a fluid dynamic bearing spindle motor that rotationally drives the disk based on history information accumulated for a period of time, and the generation A hard disk drive survival model generation step for generating a hard disk drive survival model representing a relationship between a change in the time series of the deterioration degree of the lubricant estimated by the determined deterioration estimation model and a survival probability of the hard disk drive, and a prediction phase Installed in hard disk drive A second data acquisition step of acquiring sensor data from the acquired sensor and storing it as sensor information, and applying the acquired sensor information to the generated deterioration estimation model to estimate the degree of deterioration of the lubricating oil A deterioration estimation step, and a failure prediction step of performing a trend analysis by the hard disk drive survival model based on the estimated deterioration degree of the lubricating oil and predicting a failure time of the hard disk drive according to the prediction phase. Features.

  According to the present invention, a hard disk drive failure prediction device for diagnosing and predicting a failure caused by deterioration of a fluid dynamic bearing spindle motor in most hard disk drives by using an acquirable sensor item such as a SMART function, A hard disk drive failure model generation device and a hard disk drive failure prediction method are provided.

(Embodiment 1)
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an example of the overall configuration of a hard disk drive (hereinafter referred to as “HDD”) failure model generation device 1 and HDD failure prediction device 2 according to Embodiment 1 of the present invention. The HDD failure model generation device 1 is a device related to the HDD failure model generation process (learning phase), and the HDD failure prediction device 2 is a device related to the HDD failure prediction process (prediction phase).

  The HDD failure model generation apparatus 1 is a computer such as a general-purpose computer including a data acquisition unit 11, a history information storage unit 12, a model generation unit 13, a deterioration estimation model storage unit 14, and an HDD survival model storage unit 15. The learning phase HDD 100 is connected. Note that it is preferable to provide a plurality of learning phase HDDs and a plurality of types of HDDs for collecting a large amount of history information, but the invention is not limited to this.

  The data acquisition unit 11 includes at least a specified rotation speed, energization time (PowerOnHours), disk temperature (Temperature), spin-up time (SpinUpTime), spin-up current amount ( SpinHighCurrent) is periodically acquired, and the acquired information is converted into history information in a format that can be easily processed by the computer (for example, CSV format), and history information until a failure occurs is stored in the history information storage unit 12. It is a program to memorize. The specified rotational speed is a disk rotational speed (for example, catalog specification) in a steady state. The spin-up time is the time required for the disk to reach a specified rotation speed (spin-up) after starting energization rotation. The spin-up current amount is the maximum amount of current used to spin up the disk. The disk temperature is the temperature inside the disk. The energization time (PowerOnHours), the disk temperature (Temperature), and the spin-up time (SpinUpTime) are acquired in cooperation with the S.M.A.R.T. function provided in the HDD 100 in advance. The specified rotational speed is read from a file (for example, an environment setting file) stored in the disk, and the spin-up current amount is acquired by a dedicated sensor or the like. Hereinafter, items of data acquired by the S.M.A.R.T. function or the like, or items of information stored by the data acquisition unit 11 are referred to as sensor items.

  The model generation unit 13 reads the history information stored in the history information storage unit 12, generates a deterioration estimation model and a HDD survival model for each specified rotational speed, and generates a deterioration estimation model storage unit 14 and a HDD survival model storage unit 15. Are programs to be stored respectively. The deterioration estimation model is a mathematical model that represents the relationship between the history information sensor item and the degree of deterioration of the lubricating oil of the fluid dynamic bearing spindle motor that rotationally drives the disk. The HDD survival model is a mathematical model that represents the relationship between changes in the time series of the deterioration degree of the lubricating oil of the spindle motor obtained from the deterioration estimation model and history information and the survival probability of the hard disk drive. The reason for creating a model for each specified rotational speed from information on a plurality of types of hard disk drives is that the spin-up time varies depending on the specified rotational speed. By classifying in this way, the accuracy of the generated model is improved.

  The HDD failure prediction apparatus 2 includes a data acquisition unit 21, a deterioration estimation unit 22, a failure prediction unit 23, a result notification unit 24, a deterioration estimation model storage unit 14 ′, an HDD survival model storage unit 15 ′, a history information storage unit 25, and A computer such as a general-purpose computer including the deterioration estimation history storage unit 26 is connected to the HDD 101 related to the prediction phase.

  The data acquisition unit 21 has at least a specified rotational speed, energization time (PowerOnHours), disk temperature (Temperature), and spin-up time (which can be acquired by the SMART function) as internal state information from one or a plurality of HDDs 101 to be monitored. SpinUpTime) is periodically acquired, converted into history information (for example, CSV format) that can be easily processed by a computer, and history information until a failure occurs is stored in the history information storage unit 12. It is assumed that the HDD 101 used in the HDD failure prediction process does not have a function that can acquire the spin-up current amount.

  The deterioration estimation model storage unit 14 ′ and the HDD survival model storage unit 15 ′ are storage devices that store the same models as the deterioration estimation model storage unit 14 and the HDD survival model storage unit 15 of the failure model generation device 1, respectively.

  The degradation estimation unit 22 estimates the degradation level of the current HDD 101 by applying the last record of the history information storage unit 25 to the degradation estimation model stored in the degradation estimation model storage unit 14 ′. This is a program for storing an estimation result obtained by pairing a degradation level (hereinafter referred to as “estimated degradation level”) with time in the degradation estimation history storage unit 26 as degradation estimation history information.

The failure predicting unit 23 automatically identifies the failure determination deterioration level at which the disk survival rate is significantly reduced, or presents the HDD survival model on a display (not shown) to inform the user of the threshold θ of the disk drive survival rate. When the failure determination deterioration degree _dθ is determined in advance and new information is added to the deterioration estimation history storage unit 26, the HDD survival model stored in the HDD survival model storage unit 15 ′ is prompted. utilizing, by trend analysis information stored in the degradation estimation history storage unit 17, calculates the remaining service life of up to estimate the degree of deterioration of HDD101 currently monitored reaches the failure determination deterioration degree d _theta It is a program.

  The result notification unit 24 displays the remaining life calculated by the failure prediction unit 23 or the predicted failure time (date and time) that can be calculated from the remaining life when the remaining life falls below a predetermined time (shown in the figure). It is a program that outputs to mail software and notifies the user.

  Hereinafter, the operation of the HDD failure model generation apparatus 1 will be described with reference to the drawings. FIG. 2 is a flowchart illustrating a specific example of processing in the HDD failure model generation apparatus 1.

  First, the data acquisition unit 11 waits until the HDD power OFF sequence is started (S201). When the HDD power OFF sequence is started, the data acquisition unit 11 acquires at least “data acquisition date” and “disk temperature” from the HDD (S202), and additionally stores them in the history information storage unit 12 in a predetermined format (S203). ). FIG. 3 is a diagram illustrating a specific example of the history information format. FIG. 3A shows a history information schema. Here, data acquisition date and time (Date), hard disk identifier (ID), disk temperature when power is turned off (PowerOffTemp), disk temperature when power is turned on (PowerOnTemp) , A specified rotation speed (RPM), energization time (PowerOnHours), stop time (PowerOffHours), spin-up time (SpinUpTime), spin-up current amount (SpinHighCurrent), and the like. FIG. 3B is a diagram showing the update of the history information during the HDD power OFF sequence. Here, the information in which the last line is updated is shown.

  Next, the data acquisition unit 11 waits until the HDD power ON sequence is started (S204). It is checked whether the HDD power ON sequence is started and one of the connected HDDs is activated (S205).

  In S205, if any HDD has been activated, the data acquisition unit 11 performs at least “acquisition date / time”, “specified rotation speed”, “disk temperature”, “spin-up current amount”, “spin” from all the activated HDDs. The “up time” and “energization time” are acquired (S206), merged with the information stored in S203, and additionally stored in the history information storage unit 12 in a predetermined format (S207). FIG. 3C is a diagram illustrating the update of history information during the HDD power-on sequence. Here, the information in which the last line is updated is shown.

  In S205, when all HDDs are not activated (that is, when a failure occurs), or when a predetermined number of HDDs are not activated, the data acquisition unit 11 stores the history information in the history information storage unit 12. From the obtained information, the “specified rotational speed”, “disk temperature (when ON)”, “disk temperature (when OFF)”, “spin-up current amount”, “spin-up time”, “energization time” and “stop time” are read (S208) .

  Next, the model generation unit 13 categorizes the Bayesian network CPT (conditional probability table) for each specified number of rotations, and stores it as a deterioration estimation model in the deterioration estimation model storage unit 14 (S209).

  Finally, the model generation unit 13 stores in the HDD survival model storage unit 15 an HDD survival model in which the time series information of the “lubricating oil age” node of the degradation estimation model and the HDD survival rate are paired (S210). ), The process is terminated.

FIG. 4 is a diagram for explaining the structure of a deterioration estimation model (Bayesian network). A Bayesian network is a probability model in which a qualitative dependency relationship between a plurality of random variables is represented by a graph structure, and a quantitative relationship between individual variables is represented by a conditional probability. Here, the age (degradation degree) of the lubricating oil of the spindle motor is obtained from three elements of the total operation time (t ₁ ), the disk temperature at stop (T ₁ ), and the total stop time (t ₂ ). The viscosity of the lubricating oil is determined from the age of the oil and the starting disk temperature (T ₀ ). It is shown that the angular velocity of the disk is obtained from the viscosity of the lubricating oil and the amount of spin-up current, and the spin-up time is obtained from this angular velocity. Note that the nodes represented by white circles are items that can be observed in the learning phase, and the nodes represented by black circles are items that cannot be observed, but cannot be observed from items that can be observed by the Bayesian network graph structure. It is possible to estimate items.

  FIG. 5 and FIG. 6 are diagrams showing results of analyzing the tendency of HDD degradation by experiments in giving the dependency shown in FIG. FIG. 5 is a diagram showing the relationship between the spin-up time and the disk temperature. Here, the circles indicate the data in the HDD in the initial state, and the squares indicate the data in the HDD that has deteriorated over time, and when these are connected by data, two quadratic bands are obtained. By comparing these two bands, it is estimated that the deterioration direction is the direction of the arrow. FIG. 6 is a diagram showing the relationship between the spin-up time, the disk temperature, and the spin-up current amount. Here, the same result as in FIG. 5 is shown when the current value is constant, but the result that the spin-up time changes significantly when the current value is changed is shown. That is, it can be seen that the spin-up time is determined not only by the disk temperature but also by the spin-up current amount.

  In addition, a minute groove is dug in the bearing part of the spindle motor of the disk drive that is widely used at present. Oil is collected in the direction of narrowing the groove, and the rotating shaft floats up from the sleeve. It is a mechanism that rotates in the form that the disk is involved. Therefore, a viscosity effect of a certain level or more is necessary, but if it is too high, the starting torque cannot be satisfied and a phenomenon of not rotating occurs. From the above, it can be seen that the spin-up time is not directly affected by the temperature, but is closely related to the viscosity of the oil. FIG. 7 is a graph showing the relationship between the disk temperature and the viscosity. As shown in the figure, the viscosity of the oil has a characteristic that it becomes smaller as the temperature is higher, and when the temperature changes by 60 ° C., the viscosity changes by almost one digit. Further, the deterioration of oil proceeds due to factors such as oxidation, contamination of impurities, and high temperature, and the viscosity becomes high. From this, it can be seen that the viscosity at startup is determined by the disk temperature at startup and the current deterioration state of the oil. At this time, oxidation, which is the current state of deterioration of oil, is due to the passage of time. Similarly, the high temperature can be determined by obtaining a temperature in a state where the disk drive is energized. Therefore, it can be seen that the energization time, the stop time, and the disk temperature at the time of stop (that is, the disk temperature in the energized state) cause the oil deterioration.

  Further, as impurities to be mixed in, (1) very fine dust, (2) water droplets due to dew condensation, (3) metal pieces due to damage on the disk surface, etc. are conceivable. (1) can be substituted depending on the elapsed time, and (2) need not be considered in an environment where there is no extreme temperature difference. Further, regarding (3), it is considered that it is not necessary to consider if there is no impact / vibration. Taking the above into consideration, the structure of the deterioration estimation model in FIG. 4 is given.

  FIG. 8 is a diagram illustrating a specific example of the HDD survival model. Here, it is shown that the HDD survival rate (= 1-HDD failure rate) rapidly decreases with the age of the lubricating oil as a boundary.

  Hereinafter, the operation of the HDD failure prediction apparatus 2 will be described with reference to the drawings. FIG. 9 is a flowchart illustrating a specific example of processing in the HDD failure prediction apparatus 2.

  First, the deterioration model and HDD survival model generated by the HDD failure model generation device 1 are copied to the HDD failure prediction device 2 (S901). Note that any means such as a LAN, the Internet, or an external storage medium may be used for copying.

Next, the failure prediction unit 23 automatically identifies the degree of deterioration in which the survival rate is significantly reduced by using the HDD survival model stored in the HDD survival model storage unit 15 ', or as shown in FIG. A hard disk survival model is presented on a display (not shown), and the threshold value θ of the disk drive survival rate is input by the user to determine the failure determination deterioration degree d _θ (S902).

  Next, the data acquisition unit 21 waits until the HDD power OFF sequence is started (S903). When the HDD power OFF sequence is started, at least “data acquisition date” and “disk temperature” are acquired from the HDD (S904), processed and additionally stored in the history information storage unit 25 (S905). FIG. 10 is a diagram illustrating a specific example of a history information format. FIG. 10A shows a history information schema. Here, data acquisition date and time (Date), hard disk identifier (ID), disk temperature when the power is turned off (PowerOffTemp), disk temperature when the power is turned on (PowerOnTemp) , The specified rotation speed (RPM), energization time (PowerOnHours), stop time (PowerOffHours), spin-up time (SpinUpTime), spin-up current amount (SpinHighCurrent), and the like. FIG. 10B is a diagram illustrating the update of history information during the HDD power OFF sequence. Here, the data in which the last line is updated is shown.

  Next, the data acquisition unit 21 waits until the HDD power ON sequence is started (S906). When the HDD power ON sequence is started, at least “acquisition date / time”, “specified rotation speed”, “disk temperature”, “spin-up time”, and “energization time” are acquired from all activated HDDs (S907). The data is merged with the stored data and additionally stored in the history information storage unit 25 according to a predetermined format (S908). FIG. 10C is a diagram illustrating the update of history information during the HDD power-on sequence. Here, the updated data is shown in the last row.

Next, the deterioration estimation unit 22 determines, from the data stored in the history information storage unit 25, the “specified rotational speed (R)”, “disk temperature_ON (T ₀ )”, “disk temperature_OFF (T ₁ )” ) ”,“ Spin-up time (S) ”,“ Energization time (t ₁ ) ”, and“ Stop time (t ₂ ) ”are read (S909).

  Next, the deterioration estimation unit 22 applies the data read in S909 to the deterioration estimation model stored in the deterioration estimation model 14 ′, and estimates an expected value (following a Gaussian distribution) of “lubricating oil age”. The estimation result obtained by pairing the estimated expected value (estimated deterioration level) with the time is stored in the deterioration estimation history storage unit 26 as deterioration estimation history information (S910). FIG. 11 is a diagram illustrating input / output (evidence and query) when estimating the age of the lubricating oil based on the Bayesian network. Here, P (X | e), which is the posterior probability of the age of the lubricating oil, is estimated on the condition of an observable item (e: evidence) represented by a white circle. Note that the spin-up current amount cannot be acquired in the prediction phase, but can be estimated from the characteristics of the Bayesian network if a strong graph structure can be constructed in the learning phase.

Next, the failure prediction unit 23 analyzes the tendency of the deterioration estimation history information stored in the deterioration estimation history storage unit 26 to obtain a failure prediction time (d _θ arrival time) t _b as shown in FIG. The remaining life is calculated based on the difference from t (S911). Figure 12 is a lubricant deterioration curve of the disc and trend analysis is a view for explaining a method of determining the failure prediction time t _b.

  Then, the result notifying unit 24 determines whether or not the remaining life is shorter than the value designated in advance by the user (S912), and if the remaining life is shorter than the designated value, directly gives a warning to the user's display. A warning is notified by display or e-mail (S913), and the process returns to S903. On the other hand, if the remaining life is greater than the specified value, the process returns to S903 without warning.

  By configuring as described above, it is possible to diagnose and predict a failure caused by deterioration of a fluid dynamic bearing spindle motor using sensor items that can be acquired by the S.M.A.R.T. function in most hard disk drives. Further, since the deterioration model and HDD survival model generated by the HDD failure model generation device 1 are copied in advance to the HDD failure prediction device 2 by an external storage medium or the like, it is possible to perform failure prediction even when not connected to the network. There is.

(Embodiment 2)
FIG. 13 is a block diagram illustrating an overall configuration example of the HDD failure model generation device 1 and the HDD failure prediction device 2 according to the second embodiment. The same reference numerals as those in FIG. 1 represent the same components, and thus the description thereof will be omitted. Differences from the first embodiment will be described in detail. As shown in the figure, in this embodiment, unlike the first embodiment, the failure model generation apparatus 1 related to the HDD failure model generation process (learning phase) and the HDD failure prediction related to the HDD failure prediction process (prediction phase). A device 2 is connected via a network 3 such as a LAN.

  Further, instead of copying the deterioration estimation model and the HDD survival model generated by the HDD failure model generation device 1 to the HDD failure prediction device 3, the deterioration estimation unit 22 and the failure prediction unit 23 of the HDD failure prediction device 2 are replaced with the HDD failure model. The degradation estimation model storage unit 14 and the HDD survival model storage unit 15 on the generation device 1 side are respectively accessed and acquired via a network.

  Therefore, the configuration of the HDD failure model generation device 1 is the same as that of the first embodiment, but the HDD failure prediction device 2 has a degradation estimation model storage unit 14 ′ and a HDD survival model storage unit 15 for storing models. Does not have a ´.

  The operation of the HDD failure model generation device 1 of the present embodiment is the same as the operation of the first embodiment (FIG. 2), and the operation of the HDD failure prediction device 2 is performed instead of copying the model in S901 of FIG. The deterioration estimation unit 22 and the failure prediction unit 23 of the HDD failure prediction apparatus 2 access and acquire the deterioration estimation model storage unit 14 and the HDD survival model storage unit 15 on the HDD failure model generation apparatus 1 side through the network, respectively. Is the same.

  By configuring as described above, even when the HDD failure model generation device 1 updates the deterioration estimation model and the HDD survival model based on the latest history information, the HDD failure prediction device 2 can quickly acquire the update. There are significant advantages.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, the constituent elements over different embodiments may be appropriately combined.

  Specifically, the failure model generation device 1 related to the HDD failure model generation process (learning phase) and the HDD failure prediction device 2 related to the HDD failure prediction process (prediction phase) may be a single device instead of separate devices. .

  In addition, the SMART function is used as a method of acquiring various information from the hard disk drive, but there is another sensor function that can acquire the spin-up time, spin-up current amount, disk temperature, energization time, and specified rotation speed. May be used.

  Furthermore, in the above-described embodiment, it is assumed that the spin-up current amount cannot be acquired in the prediction phase. However, when a sensor or the like that can measure the spin-up current amount is provided, the degradation estimation model is used. By adding the measured spin-up current amount to the sensor item to be applied, the accuracy of the estimated deterioration degree can be further improved.

1 is a block diagram showing an overall configuration example of an HDD failure model generation device 1 and an HDD failure prediction device 2 according to Embodiment 1 of the present invention. 4 is a flowchart illustrating a specific example of processing in the HDD failure model generation device 1. The figure which shows the specific example of the format of log | history information. The figure explaining the structure of a degradation estimation model (Bayesian network). The figure which shows the result of having analyzed the degradation tendency of HDD by experiment in giving the dependence relationship of FIG. The figure which shows the result of having analyzed the degradation tendency of HDD by experiment in giving the dependence relationship of FIG. The figure which shows the relationship between disc temperature and a viscosity. The figure which shows the specific example of HDD survival model. 5 is a flowchart showing a specific example of processing in the HDD failure prediction apparatus 2. The figure which shows the specific example of the format of log | history information. The figure explaining the method of estimating the age of lubricating oil based on a Bayesian network. The figure explaining the method which carries out the trend analysis of the lubricating oil deterioration curve of a disk, and calculates | requires failure prediction time. The block diagram which shows the example of whole structure of the HDD failure model production | generation apparatus 1 and HDD failure prediction apparatus 2 which concern on Embodiment 2 of this invention.

Explanation of symbols

1 HDD failure model generation device,
2 ... HDD failure prediction device,
3 ... Network,
11: Data acquisition unit,
12 ... History information storage unit,
13 ... model generation unit,
14, 14 '... deterioration estimation model storage unit,
15, 15 '... HDD survival model storage unit,
21 ... Data acquisition unit,
22 ... deterioration estimation part,
23 ... Failure prediction unit,
24 ... result notification part,
25. History information storage unit,
26: Deterioration estimation history storage unit.

Claims (10)

A first data acquisition unit that acquires sensor data from a sensor mounted on a hard disk drive according to a learning phase, and stores the sensor data as sensor information;
Deterioration that generates a deterioration estimation model that represents the relationship between the sensor information and the degree of deterioration of the lubricating oil of the fluid dynamic bearing spindle motor that rotationally drives the disk, based on history information obtained by accumulating the acquired sensor information for a desired period. An estimation model generation unit;
A hard disk drive survival model generating unit that generates a hard disk drive survival model representing a relationship between a change in the time series of the deterioration degree of the lubricant estimated by the generated deterioration estimation model and the survival probability of the hard disk drive;
A second data acquisition unit that acquires sensor data from a sensor mounted on the hard disk drive according to the prediction phase, and stores the sensor data as sensor information;
A deterioration estimation unit that applies the acquired sensor information to the generated deterioration estimation model to estimate the degree of deterioration of the lubricating oil of the hard disk drive according to the prediction phase;
Applying the estimated degree of deterioration of the lubricant to the generated hard disk drive survival model to perform a trend analysis, and predict a failure time of the hard disk drive according to the prediction phase;
A hard disk drive failure prediction apparatus comprising:   The degradation estimation model generation unit includes at least the energization time to the disk, the specified rotational speed in a steady state of the disk, and the predetermined rotational speed after power is supplied to the disk, among the sensor items of the history information The hard disk drive failure prediction apparatus according to claim 1, wherein the deterioration estimation model is generated using a required time, a temperature of the disk, and a maximum amount of current required to reach the specified rotational speed.   The deterioration estimating unit includes at least the energization time to the disk, the specified rotational speed in a steady state of the disk, and the specified after the power is applied to the disk, among the sensor information acquired by the second data acquiring unit. The time required to reach the rotational speed and the temperature of the disk are applied to the deterioration estimation model to determine the current deterioration degree of the lubricating oil of the hard disk drive related to the prediction phase. Item 3. The hard disk drive failure prediction apparatus according to Item 2.   4. The hard disk drive failure prediction apparatus according to claim 1, wherein the deterioration estimation model is a Bayesian network in which the sensor items and the deterioration degree of the lubricant are nodes. 5. A data acquisition unit that acquires sensor data from a sensor mounted on a hard disk drive related to the learning phase, and stores it as sensor information;
Deterioration that generates a deterioration estimation model that represents the relationship between the sensor items and the degree of deterioration of the lubricating oil of the fluid dynamic bearing spindle motor that rotationally drives the disk, based on the history information obtained by accumulating the acquired sensor information for a desired period. An estimation model generation unit;
A hard disk drive survival model generation unit that generates a hard disk drive survival model representing a relationship between a change in the time series of the deterioration degree of the lubricant estimated by the generated deterioration estimation model and the survival probability of the hard disk drive;
A hard disk drive failure model generation device comprising:   The degradation estimation model generation unit includes at least the energization time to the disk, the specified rotational speed in a steady state of the disk, and the predetermined rotational speed after power is supplied to the disk, among the sensor items of the history information 6. The hard disk drive failure model generation device according to claim 5, wherein the deterioration estimation model is generated using a required time, a temperature of the disk, and a maximum amount of current required to reach the specified rotational speed. .   7. The hard disk drive failure model generation device according to claim 5, wherein the deterioration estimation model is a Bayesian network having the sensor items and the deterioration degree of the lubricant as nodes. A first data acquisition step of acquiring sensor data from a sensor mounted on a hard disk drive according to a learning phase and storing it as sensor information;
Deterioration that generates a deterioration estimation model that represents the relationship between the sensor item and the degree of deterioration of lubricating oil in a fluid dynamic bearing spindle motor that rotationally drives the disk, based on history information obtained by accumulating the acquired sensor information for a desired period. An estimation model generation step;
A hard disk drive survival model generation step for generating a hard disk drive survival model representing a relationship between a change in the time series of the deterioration degree of the lubricant estimated by the generated deterioration estimation model and the survival probability of the hard disk drive;
A second data acquisition step of acquiring sensor data from a sensor mounted on the hard disk drive according to the prediction phase and storing it as sensor information;
A deterioration estimation step for estimating the degree of deterioration of the lubricating oil by applying the acquired sensor information to the generated deterioration estimation model;
A failure prediction step of performing a trend analysis by the hard disk drive survival model based on the estimated deterioration degree of the lubricating oil, and predicting a failure time of the hard disk drive according to the prediction phase;
A hard disk drive failure prediction method comprising:   In the deterioration estimation model generation step, at least the energization time to the disk, the specified rotational speed in the steady state of the disk, and the specified rotational speed after the power supply to the disk is reached among the sensor items of the history information 9. The hard disk drive failure prediction method according to claim 8, wherein the deterioration estimation model is generated using a required time, a temperature of the disk, and a maximum amount of current required to reach the specified rotational speed.   The hard disk drive failure prediction method according to claim 8 or 9, wherein the deterioration estimation model is a Bayesian network having the sensor items and the deterioration degree of the lubricating oil as nodes.