KR100376802B1 - Expert system for discriminating fault of hard disk drive and method thereof - Google Patents

Abstract

The expert system of the present invention includes a rule file that formalizes the step-by-step procedure on the repair judgment processor that the repairman wants to grasp and infer, and analyzes the rule file to perform each procedure to perform the defect determination process of the actual repairman. Build an inference engine that can be disguised. It is also provided with a pattern recognition module for recognizing the shape of the defect distribution on the disk medium providing information important for determining the repair determination result. Reading and testing MC information in the HDD required by the rule file, the inference engine, and the pattern recognition module may use the existing Defect information extraction module (V-tool) and the defect verification module (gemini).

Description

Expert system for failure determination of hard disk drive and its method {EXPERT SYSTEM FOR DISCRIMINATING FAULT OF HARD DISK DRIVE AND METHOD THEREOF}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a manufacturing process of a hard disk drive (HDD), and more particularly, to a technique of analyzing a cause of failure of an HDD determined to be defective at the time of manufacturing a HDD.

These days, data storage and reading devices are rapidly increasing in capacity in response to the multimedia implementation of the system, and are being developed to enable high-speed data access. An example of such data storage and reading devices is a hard disk drive (HDD). The HDD is widely used as an auxiliary memory device of a computer system due to the advantage of enabling high-speed data access and high capacity.

The HDD is shipped as a complete product through a series of manufacturing and testing procedures as shown in FIG. 1 is a diagram illustrating a general main procedure in the HDD manufacturing process, which is divided into six stages.

Referring to each step with reference to Figure 1, first, the first step (I) of the manufacturing process HDA (Head Disk Assembly) assembly process is a process of assembling the head disk assembly (HDA) of the HDD of the manufacturing process clean ( in the room. The second step (II) of the manufacturing process is a servo writer which writes a servo pattern for servo control of a magnetic head that reads and writes data to and from a disk, which is a recording medium. Is performed by. The function test, which is the third step (III) of the manufacturing process, is a test performed before combining the HDA made in the HDA assembly process and the PCBA made in the printed circuited board assembly (PCBA) process. Connect to the test PCBA and test whether it works properly. HDA, considered normal in this functional test process, is then combined with PCBA. The burn-in test process, which is the fourth stage (IV) of the manufacturing process, is the process that takes the longest time (usually 8 to 16 hours) of the HDD manufacturing process, and does not require a separate test system. It is placed on a rack in the burn-in room of and then performed according to its own program (firmware). In this burn-in test process, the defect part existing on the disk is found in advance in order to allow the consumer to use the HDD normally, and a process of preempting the defect part when the actual drive is used. The final test process, which is the fifth step (V) of the manufacturing process, is a process for checking whether the HDD set passed in the burn-in test process is normally defected. Test the defect state for each set. After the final test process, the HDD set is shipped as a finished product through the shipment inspection process, packaging and shipment process, which is the sixth step (VI) of the manufacturing process.

In the above process, the HDD may be treated as defective because the defect is more than a prescribed value or inoperable. The HDD treated as defective is sent to a professional repairman later to find out the cause of the defect, that is, to identify and prescribe the type of defect. The repairman analyzes the defects of the sent HDD to determine the cause of the defective HDD.

2 is a flowchart of a general defect determination work performed by a repairman for analyzing a bad HDD. Referring to FIG. 2, a repairman first connects a bad HDD to a tester using a personal computer (PC). The tester is equipped with various programs that can read various defect information recorded in the HDD. The repairer first checks the header information of the HDD through the tester in step 20, and then in step 22, the repairer determines the distribution of the defect location on the disk, and infers the defect type through this. This is done by operating a specific Defect Information Extraction Module (aka V-Tool) that graphically displays the Defect location on the disc.

The results of the burn-in test or the final test during the manufacturing process of each drive are recorded in a maintenance cylinder (MC) which is a system area of the disc. At this time, information on the location of each defect is recorded. The Defect information extraction module, that is, the V-tool, reads the information on the location of the defect in the MC of each disk, and stores the information on the defect in A of FIG. As shown in the figure, the graphic is displayed on the PC screen. In this case, the marking is performed by applying a separate color and contrast to the circular disk and the portion where the disk defect occurs. The repairman infers the approximate cause of the defect by grasping the defect distribution form of the disk displayed on the screen. Defect distribution forms may have a variety of forms, the defect distribution forms directly or indirectly indicate the cause of the defect. In other words, the form shown in A of FIG. 2 is a so-called "sticking" form, which may indicate that the disk is likely to be defective.

When the repairman grasps the shape of the defect distribution and infers the cause of the defect, it is verified in step 24 that the cause of the defect is correct. This can be done by driving a separate failure verification module (aka Gemini). The defect verification module Gemini verifies the defect by performing a read / write test on the error inferred disk and the head. The execution result of the failure verification module is output as text on the screen of the PC, through which the repairman determines the type of failure in step 26 and instructs the prescription accordingly. For example, the type of failure may be determined to be a disk failure, and the replacement of the disk may be instructed as a prescription.

This operation can be used to analyze the cause of the bad HDD for the bad HDD and take measures accordingly. In addition to the graphic information on the distribution of the defect location, the repairman can read the various defect information recorded in the MC to analyze the defect more precisely. .

However, as described above, since the conventional failure determination method is performed by the repairman, a separate human resource for the failure determination is required, and the consistency of the failure determination is difficult to be maintained.

Accordingly, an object of the present invention is to provide an HDD failure determination expert system and method for performing a failure determination of the HDD without requiring any human resources.

Another object of the present invention is to provide an HDD failure determination expert system and method for maintaining consistency of failure determination and improving hit rate.

In order to achieve the above object, the expert system of the present invention includes a rule file that formalizes a step-by-step procedure on a repair judgment processor that a repairman wants to grasp and infer, and analyzes the rule file to perform each procedure. Construct an inference engine that simulates the failure determination process of the actual repairer. It is also provided with a pattern recognition module for recognizing the shape of the defect distribution on the disk medium providing information important for determining the repair determination result. At this time, the shape of the Defect distribution is atypical, and many forms are mixed. Therefore, various pattern recognition techniques using artificial intelligence and statistical techniques are introduced. In addition, the MC information reading and testing in the HDD required by the rule file, the inference engine, and the pattern recognition module can use the existing Defect information extraction module (V-tool) and the defect verification module (gemini).

1 is a flowchart of a general main procedure in the manufacturing process of a hard disk drive.

2 is a flowchart of a general failure determination work of a repairman for analyzing a bad hard disk drive.

3 is a functional block diagram of a hard disk drive failure determination expert system according to an embodiment of the present invention.

Figure 4 is an exemplary view of the defect distribution form and the kind of failure according to the present invention.

5 is a block diagram of a hard disk drive failure determination expert system according to an embodiment of the present invention;

6 is a view for explaining the operation of the inference engine through the rule file of FIG.

FIG. 7 illustrates an example of a rule file format and some rule routines of FIG. 5; FIG.

8 is a flowchart of an example of one rule routine of the rule file of FIG. 5 and a failure determination procedure accordingly;

9 is a flowchart illustrating a pattern recognition process of the pattern recognition unit of FIG. 5;

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, specific matters such as configuration of specific rule files and rule routines are shown, which are provided to help a more general understanding of the present invention, and these specific matters may be modified or changed within the scope of the present invention. It will be obvious to those of ordinary skill in the art.

The present invention provides an expert system for automating the HDD failure determination work procedure that is being performed manually by a conventional repairman. The expert system of the present invention is configured to have knowledge of repair judgment of a repairman. Accordingly, the HDD failure determination expert system of the present invention reads the information of the HDD, that is, the information of the MC to grasp the information on the defect including the information on the distribution of the defect location of the HDD, and thus to determine the distribution of the defect, etc. Determine the type of defect. Referring to FIG. 3, the HDD failure determination expert system according to an exemplary embodiment of the present invention uses a defect information extraction module (defect information extraction module) 32 that reads information of an HDD, that is, information of an MC, and through defect information. The pattern recognition module 36 for recognizing the shape of the defect distribution and the defect inference module 30 for determining the type of defect through the information on the defect and the information on the pattern recognition are provided. In addition, the failure verification module 34 may be further provided to more clearly verify the determination of the type of failure. The defect information extraction module 32 may use an existing V-tool, and the defect verification module 34 may also use an existing gemini.

On the other hand, the shape of the Defect distribution is atypical, many forms are often mixed, and it is often difficult to define exactly which distribution form corresponds to what kind of defects. Accordingly, the present invention proposes a defect distribution type and a defect type determined accordingly as shown in FIG. 4.

4 is an exemplary diagram of a defect distribution form and a kind of defects according to the present invention. Defect distribution form proposed in the present invention may be called "annular: minute hand: total distribution: straight line: stamp: circumference: block". According to each type of defect distribution, the type of defect is regarded as "head: head: head: disk: disk: disk or servowriter: servowriter or mechanism" respectively.

In more detail, the "annular" of the above-described Defect distribution form means that the Defect region is formed in a ring shape along the track direction on the disc, which is caused by contact with the disc rotating at high speed. To be considered. Judging from this, the "minute hand" shape is similar to the "annular shape" but the defect part is formed in the shape of an entire ring without being connected along the track direction, and is considered to be a head failure. A form that occurs randomly throughout the phase, which is considered "head failure." "Straight" refers to the defect formed on the disc to some length across the track and is considered a disc failure. "Ticking" refers to the fact that the defect area is formed by spotting at some point on the disk and is considered to be a defective disk. "Circumference" means that the defect occurs in the shape of a circumference with a certain length along some track direction on the disc, and is regarded as a disc or servo writer failure. A "block" refers to a defect formed in a relatively wide block at any part of the disc, which is considered to be a failure of the servowriter or instrument. As described above, the type of distribution of defects and the types of defects may be distinguished, and the type of distribution of defects and types of defects may be further set. For example, it can be said that symmetry is formed in two areas of the defect generating region symmetrically about the center of the disk, which can be regarded as a defect of the spindle motor. Hereinafter, the configuration of the expert system for identifying the defect distribution form as described above and determining the type of defect will be described in more detail with reference to FIG. 5.

5 is a block diagram of a hard disk drive failure determination expert system according to an exemplary embodiment of the present invention. The expert system according to an embodiment of the present invention formalizes a step-by-step procedure of repair decision that the repairman wants to grasp and infer, and includes a rule file 51 that describes the signal / branch relationship of each procedure. By analyzing the file 51 and performing each procedure, the inference engine 50 that simulates the defect determination process of the actual repairman and the defect coordinate file obtained through the defect information extraction module 52 are analyzed and the defect distribution on the disc medium is analyzed. The pattern recognition unit 56 which recognizes the shape and the inference engine 50, which are driven by the inference engine 50, extract the HDD data (MC information), which is input data of the defect determination procedure executed by the inference engine 50, from the target HDD. The defect verification module 54, which is driven by the information extraction module 52 and the inference engine 50, performs a test program for diagnosing and verifying the HDD status among the failure determination procedures executed by the inference engine 50. And a medium file 55 for temporarily storing the information of the defect information extraction module 52 and the defect verification module 54 so that the inference engine 50 and the pattern recognition unit 56 can be used. That is, the defect information extraction module 52 and the failure verification module 54 interface with the inference engine 50 and the pattern recognition unit 56 through the parameter file 55. The user may instruct the inference engine 50 to start the operation through a key input device (not shown) and the like, and display information on the determination result output from the inference engine 50 such as a monitor (not shown). Can be seen through.

In the expert system configured as described above, the parameter file 55 will be described in more detail. The parameter file 55 serves as an interface with the inference engine 50, the defect information extraction module 52, and the defect verification module 54. It is a text file. When the inference engine 50 drives the defect information extraction module 52 or the failure verification module 54, the result is stored in the parameter file 55, and the inference engine 50 reads it to obtain the necessary result. The parameter file 55 is largely referred to as "MC.DAT" file which stores basic information of the initial MC, "ERROLOG.DAT" file which is an error log file, and "OUTPUT.LOG" file which is a file about Gemini test results. Is done.

When the "MC.DAT" file and the "ERROLOG.DAT" file are checked for a new HDD, all existing data is deleted and replaced with the contents of the new HDD. Also, the "OUTPUT.LOG" file is updated with new contents during the failure determination process. That is, the defect information extraction module 52 may burn-in program, burn-in state, serial number, F / T (F / T) information, servo defect list, head skew table, error rate from MC information recorded in the HDD. It reads the result value and the like and writes it to the "MC.DAT" file, and also reads the error log and writes it to the "ERRORLOG.DAT" file.

Hereinafter, the structure and operation of the inference engine 50, the rule file 51, and the pattern recognition unit 56, which are main parts of the expert system as shown in FIG. 5, will be described in detail with reference to the accompanying drawings. Shall be.

FIG. 6 is a view for explaining the operation of the inference engine 50 through the rule file 51 in FIG. 5 in more detail, and FIG. 7 is a diagram showing the format of the rule file 51 and an example of some rule routines. The inference engine 50 is the main control unit of the expert system, which starts the entire system and controls the information exchange and operation of each module. The rule file 51 consists of a plurality of rule routines, each of which describes a repair determination procedure in the form of a text file. The inference engine 50 loads the appropriate rule routine from the rule file 51 and executes the action accordingly. The configuration of the rule file 51 and the inference engine 50 separately as described above is to allow flexible replacement in future technology development and model change. In other words, the expert system in various fields of the related art has a difficulty in that all components must be newly renewed when contents should be added and modified due to the development of new technology. In the present invention, in order to overcome such a problem, a rule file describing each repair discrimination procedure is separately configured, and when the contents are modified and added, it is possible to do this simply by changing and modifying the rule file.

Many rule routines that make up the rule file 51 are, for example, "POWER.RUL", "CONT.RUL", "END.RUL", "BURN.RUL", "LED31.RUL", "LED32.RUL". ... and so on. Each rule routine is configured to follow the failure determination procedure of the actual repairman in accordance with the defective state checked during the manufacturing process of the HDD. That is, when the HDD is determined to be defective during various tests of the HDD during the manufacturing process, information on the occurrence of the failure is recorded in the MC of the HDD. This is a so-called "CONT" error that means freezing in progress, a so-called "END" error that means stopping after the test is completed, and "BURN", "FAIL LED31", "FAIL LED32", "FAIL LED33" ... etc. There is this. In the conventional repairman to determine the cause of the failure of the bad HDD, instead of performing the same operation for any HDD, it is to determine the cause of the failure by performing different operations (although they are similar to each other) through the failure occurrence information. .

Accordingly, the expert system according to an embodiment of the present invention models the conventional repairman's work, reads the failure occurrence information of each HDD, and performs an appropriate operation according to each failure occurrence state. Different procedures of different operations are implemented as a plurality of routines in the rule file. For example, when the cause of the failure of the conventional HDD is a "CONT" error, the "CONT.RUL" routine in the rule file of the present invention is configured to model the procedure of the work performed by the repairman.

6 and 7, each line of each rule routine in the form of a text file may have a structure of a ternary operation consisting of a sentence number, a predicate, and a sentence number to branch according to the true and false of the predicate. have. The inference engine 50 first drives the defect information extraction module 52, and grasps the failure occurrence information of the corresponding HDD through the media file 55, and then loads the corresponding rule routine of the rule file 51. At this time, the inference engine 50 loads the corresponding rule routine of the rule file 51 in units of lines (statements) to analyze and execute the instructions of the corresponding line, that is, the conditional part. Subsequently, according to the analysis and execution result, branching is made to the corresponding sentence number and the row of the sentence number is loaded. Since each row of each rule routine follows a failure determination procedure of a conventional repairman, the inference engine 50 automatically performs the task of the conventional repairman.

In FIG. 7, examples of the “LED33.RUL” routine and the “POWER.RUL” routine among the rule routines of the rule file 51 are disclosed as A. B, respectively. The routine "POWER.RUL" describes the initialization that the expert system will perform first. The routine "LED33.RUL" describes a failure determination operation corresponding to the case where the failure occurrence information of the HDD is "FAIL LED33". This rule routine and the operation according to this rule routine will be described in more detail with reference to FIG. 8.

8 a, b show a flowchart of the " LED32.RUL " rule routine and accordingly the failure determination procedure. 8A shows a rule routine, and FIG. 8B shows a failure determination procedure of the inference engine 50 accordingly. An operation corresponding to the reference number of FIG. 8A corresponds to the operation corresponding to each sentence number of FIG. 8B. Referring to FIG. 8A, the inference engine 50 starts “LED32.RUL” in step 100 (ie, sentence number 100 of LED32.RUL in FIG. 8B), and then, in step 200, the MC information (ERROLOG.DAT) is obtained. The analysis determines whether there is continuity in the servo or gray code. Analyzing whether there is continuity in the servo or gray code is to determine whether there are consecutive sectors of defects, in which case, proceed to step 300 and check that the bad type is likely to be a disk, and proceed to step 400. Otherwise, proceed directly to step 400.

In step 400, it is determined whether SPEC OUT has occurred among the defect checks in the head unit per zone of the disk for each step in the MC information. In this case, the specification may be set to 10, for example. In this case, spec out occurs when there are more than 10 head defects in each zone. The specification setting is an additional additional option of the expert system of the present invention, and may be configured so that a user inputs a value, that is, a specification, as a reference for all error checks in advance. For example, it is possible to set, as a specification, a value as a reference in advance if it is determined that there is continuity in the case where there is more or more continuous in checking the continuity of the servo or gray code in step 200. To this end, the expert system of the present invention includes a user interface module for outputting a graphic or text to a screen for the user to easily set the operation, and may include a menu window for specification setting in the user interface module. have.

In step 400, if the depec spec out occurs in the head unit per zone per step, the process proceeds to step 500; In operation 500, the inference engine 50 drives the pattern recognition unit 56 to determine whether the pattern recognition value is calculated. If the pattern recognition unit 56 has not calculated any pattern recognition value, the process proceeds to step 900, and after calculating a specific pattern recognition value, the process proceeds to step 600.

In step 900, the defect verification module 54, for example, the Gemini is driven to perform a read / write (R / W) test and check whether the result is calculated. If the current process up to step 900 has passed through steps 100-200-300-400-500, it is considered that there is a possibility of a defect in the disk and the head. If the result value that reveals the cause of the failure is not calculated, the process proceeds to step 1000 and checks to be re-introduced, and in step 1300, it is displayed on the screen. Of course, if the result value of the Gemini is calculated in step 900, the process proceeds to steps 1100 and 1300 and then outputs the result of the Gemini on the screen.

On the other hand, if the pattern recognition value is calculated in step 500 and proceeds to step 600, in step 600 it is determined whether a head castle exists in the pattern. That is, when the pattern recognition value includes the annular, minute hand, or total distribution among the defect distribution forms as shown in FIG. 4, it is considered that the kind of defect is likely to be the head, and thus the head property exists in the pattern. To be considered. In this way, if the pattern is regarded as having a head castle, the process proceeds to step 700, and if it is considered that there is no head castle in the pattern, the process proceeds to step 800 later. In step 700, it is determined whether the result according to the pattern recognition value indicates that all heads are defective. In general, since the heads in one HDD are almost never defective, when all the heads are found to be defective, it is determined in step 1200 that the PCBA is defective, and then in step 1300, the heads are output on the screen.

On the other hand, if all heads do not appear to be defective in step 700, the process proceeds to step 800, and in step 800, the Gemini test for determining whether the determination of the failure type according to the Gemini test result value and the pattern recognition value is identical. In step 1300, the result is output on the screen. As described above, the failure determination operation by the “LED32.RUL” rule routine may be performed by performing each step as shown in FIG. 8. The inference engine 50 performs the defect determination operation using the defect information extraction module 52, the defect verification module 54, the pattern recognition unit 56, and the parameter file 55 according to the rule file 51. Done.

Next, with reference to FIG. 9, the operation of the pattern recognition unit 56 of the expert system according to an embodiment of the present invention as shown in FIG. 5 will be described in more detail. 9 is a flowchart illustrating a pattern recognition operation of the pattern recognition unit 56 of FIG. 5. The pattern recognition unit 56 first reads the error log (ERROLOG.DAT) file recorded in the intermediate file 55 by the defect information extraction module 52 in step 91. At this time, the error log file records information on the position of the defect generated on the disc for each head, and the position information on the defect is represented by the polar coordinate using the radius r and the angle θ. In operation 92, a memory mapping operation is performed in which the polar coordinates of the radius r and the angle θ are converted into rectangular coordinates having the horizontal value as the θ value and the vertical value as the r value. This coordinate transformation is intended to make the calculation process easier, and to speed up the performance of subsequent operations. Referring to FIG. 9, in the memory mapping process, a circular disk image represented by a radius r as indicated by A and an angle θ on the disk graphic image is deformed into a rectangular shape having horizontal and vertical r and θ as indicated by B. Is shown. When this process is performed, for example, when the shape of the diff distribution is annular, it appears as a ring in polar coordinates, but in the rectangular coordinates, it appears in the form of a band extending horizontally on a radius r value.

Subsequently, in step 94, a filtering operation is performed to remove the minute defects. Filtering is performed using the "K-mean clustering" algorithm. As a result of the filtering operation, as shown in FIG. 9C, the fine defect part is removed from the disk image displayed in the rectangular coordinates, and the meaningful defect part remains.

Subsequently, in step 96, each segment of the defects remaining in step 94 is cut out and analyzed for segmentation. When the segmentation operation is performed, each defect portion is classified into a separate area as indicated by D in FIG. 9. This segmentation work is performed to facilitate pattern analysis in a later pattern analysis work. For example, even if the distribution form of the defect contained in one segment is a circumferential form as shown in FIG. 4, when the segments having such a circumferential defect portion are formed relatively equidistantly on a similar radius r value, such a segment is formed. In consideration of the relationship between the entire Defect distribution form can be regarded as the minute hand form. The overall pattern analysis between these segments is then performed in step 98.

Various algorithms may be applied to the pattern analysis technique in step 98. For example, when the defect distribution form is an annular or circumferential form as shown in FIG. 4, the defect shape included in each segment is represented by a band extending horizontally or slightly shorter on a similar radius r value. At this time, the defect distribution shape is annular in consideration of the comparison between the deviation between the vertical axis component (r value) and the horizontal axis component (θ value) between the rectangular coordinates of the continuous parts of the defect in the band form, and the continuous range of the horizontal axis. Or circumferential. In addition, various types of defect distributions are recognized by comparing the regularity of the distribution of each segment and comparing the similarity between the defect distribution form and the virtual straight line.

The result of the pattern analysis is then output to the inference engine 50 in step 99. In this case, information about each recognized pattern (defect distribution form) and the position of the pattern may be output. The operation of the pattern recognition unit 50 is performed through the process as shown in FIG. 9.

The expert system according to the features of the present invention can be achieved by the configuration and operation as shown in FIG. 5 to FIG. 9. This expert system can be implemented on a single PC that can be connected to the HDD, and furthermore, it is applied to operate in conjunction with burn-in test process or final test process equipment during the manufacturing process of HDD. It can be implemented to be able to determine the failure immediately. Such expert system may also be implemented to provide information about failure determination of each HDD to the Manufacturing Execution System (MES) that oversees the entire manufacturing process of the HDD for systematic failure management.

Meanwhile, in the above description of the present invention, specific embodiments have been described, but various modifications may be made without departing from the scope of the present invention. Therefore, the scope of the present invention should not be defined by the described embodiments, but by the claims and equivalents of the claims.

As described above, the expert system according to the present invention can automatically perform the failure determination of the HDD without requiring a separate human resource, and also maintain the consistency of the failure determination and improve the hit rate. Furthermore, in the expert system according to the present invention, a rule file describing the failure determination procedure of each HDD is separately configured, and when the contents are modified and added, this is possible only by changing and modifying the rule file. Will be able to cope flexibly during development.

Claims (16)

In the expert system for determining the failure of the hard disk drive, An information extraction module for reading related information from the hard disk drive; A pattern recognition module for recognizing which one of a plurality of preset types the defective form of the hard disk drive corresponds to through the extracted related information; And an inference module that receives the recognition information of the pattern recognition module and infers the failure type of the corresponding hard disk drive based on information of a failure type preset according to each failure type and related information extracted by the information extraction module. Expert system characterized in that. The expert system according to claim 1, wherein the pattern recognition module recognizes which form of the distribution of defects for each disc among the extracted related information is. 2. The expert system of claim 1, further comprising a failure verification module for verifying the inferred failure type or upon inferring the failure type. In the expert system for determining the failure of the hard disk drive, An information extraction module for reading related information from the hard disk drive; A pattern recognition unit recognizing a defective form of the corresponding hard disk drive corresponding to any one of a plurality of preset forms through the extracted related information; A storage unit in which a rule file describing the step-by-step procedure of the defect determination is recorded; A failure determination procedure for controlling the driving of each function unit and inferring the type of failure of the corresponding hard disk drive by receiving the information obtained by the information extraction module and the recognition information of the pattern recognition unit is included in the procedure described in the rule file. Expert system comprising a reasoning engine to perform according to. The expert system according to claim 4, wherein the pattern recognition module recognizes which type of defect distribution for each disc corresponds to the extracted related information. The inference engine and the pattern recognition unit of claim 4, wherein the failure verification module performs a test for diagnosing a condition of the hard disk drive and a failure verification, and temporarily stores information of the information extraction module and the failure verification module. Expert system, characterized in that it further has a storage in which the intermediate file to use it. The storage unit of claim 6, wherein the storage unit in which the intermediate file is recorded has a file storing basic information of a corresponding hard disk drive, a file storing information on an error log, and a file storing a test result of the failure verification module. Expert system characterized in that. 5. The expert system according to claim 4, wherein the storage unit in which the rule file is recorded has a description of a step-by-step procedure of failure determination to be inferred and a signal / branch relationship of each procedure. The expert system according to claim 4, wherein the storage unit in which the rule file is recorded comprises a plurality of rule routines describing different work procedures according to the failure occurrence information of the hard disk drive. 10. The expert system according to claim 9, wherein each row of the plurality of rule routines has a ternary operator structure consisting of a sentence number, a predicate, and a sentence number to branch according to the true and false of the predicate. In the failure determination method of the expert system for determining the failure of the hard disk drive, Reading related information from the hard disk drive; Recognizing a bad shape of the corresponding hard disk drive through the related information read; And analyzing the related information read and the recognized defect type to infer a defect type of the corresponding hard disk drive. In the failure determination method of the expert system for determining the failure of the hard disk drive, Reading related information from the hard disk drive; Recognizing a bad shape of the corresponding hard disk drive through the related information read; Reading the contents of the rules file describing the step-by-step procedure of the failure determination; Controlling the operation of each functional unit, and analyzing the read related information and the recognized bad shape to infer the bad type of the corresponding hard disk drive according to the procedure described in the read rule file. Defective determination method. A method for recognizing a defect distribution form of an expert system for determining a defect distribution form of a hard disk drive for each disc and determining a defect of the hard disk drive, Reading an error log file that records the defect location information generated on the disc for each head; A memory mapping process of converting the error log file into an image of a corresponding disk; A segmentation process for cutting out and analyzing the defects of the image of the disc one by one; Defect distribution shape recognition method characterized in that it has a process of analyzing the defect distribution form by analyzing the defect shape in the segments separated by the segmentation and the distribution between the segments. 15. The method of claim 13, further comprising a filtering process for removing a minute defect portion of the disc image converted into the rectangular coordinates. The method of claim 13, wherein the defect location information of the error log file is recorded in polar coordinate format. The method of claim 13, wherein the memory mapping process converts polar coordinates of the error log file into rectangular coordinates and converts the image into a disk image.