TW584894B - System and method for determining causes causing abnormality of semiconductor equipment - Google Patents

Abstract

A system and a method for determining causes causing the abnormality of semiconductor equipment are disclosed. Diagnostic rules are used to check and correlate the data in the subsystems of computer integrated manufacture (CIM) system to root causes causing equipment abnormality, so as to generate the diagnostic results of diagnosing the abnormal status of equipment. A self-learning scheme is used for automatically integrating decisions, which are made in accordance with the diagnostic results by a user, into the diagnostic rules; and a hit ratio of each decision is provided for the user to know the importance level thereof.

Description

584894 V. Description of the invention (1) [Technical field to which the invention belongs] The present invention relates to a system and method for determining the cause of an abnormality of a semiconductor machine, and particularly to a judgment having a self-learning mechanism. System and method for causing abnormality of semiconductor machine. [Previous Technology] When an abnormal condition occurs in a semiconductor device, the semiconductor device triggers an abnormal state. This abnormal state can be a D0Wn Event, or a critical alarm event. After the equipment engineer learns that the semiconductor machine is in an abnormal state, he must first go to the Alarm Subsystem of the Computer Integrated Manufacture (CIM) system to find all relevant alarms; or go to the Manufacturing Execution Subsystem (Manufacture Execution) System; MES) to find out information about crash events. Next, go to other subsystems of the computer integrated manufacturing system, such as: Statistical Process Control (SPC) subsystem, ReahTime Monitoring (RTM) subsystem, Preventive Maintenance System (PMS) ), Exception handling record subsystem and Particle Map subsystem, etc., to find relevant data one by one. Then, check the data in each subsystem according to some diagnostic rules, and correlate the inspection results to a plurality of abnormal root causes (ROOt-Causes). These root causes are made by semiconductor equipment. Provided by the supplier, such as: · mechanism reasons, particle reasons, process reasons and equipment reasons, etc., for equipment engineers to troubleshoot (shooting-shooting) and restore machine operation. Each abnormal root cause usually has multiple sub-causes. For example,

Page 5 1 ^^ V. Description of the invention (2)

::: The process has a number of subsystems such as pressure, temperature, process time, and exhaust gas. The number of subsystems in the integrated manufacturing system is quite large. Its L-hybrid pair Γ is the root cause of the abnormality and its secondary cause. This is why we found the machine: ΐ = 的 :: = The manufacturer or computer integrated manufacturing system will provide For new engineers, learning Ermai will be useful. However, it is difficult to use the second diagnostic rule to resolve the inefficiency of the diagnostic rule. In addition, according to the experience of the standard, Fang Chu's page y is integrated first, and then, the test method successfully finds the abnormal scabies that cause the abnormality. On the other hand, standard diagnostic rules often fail to help the cause. Adding senior workers: experienced senior workers ... After the problem, due to the lack of special diagnostic rules to resolve the deposit, so that when Fu encounters the same problem next time, = unable to break the insurance family: cognition = =: resignation, then this "For the specific reasons, the engineer must take many correct reasons, which are usually the most common causes of abnormalities and lack of identification of the various diagnostic results. It can be easily found by a teacher, so the experienced project takes a lot of time to check the diagnosis results one by one. Only 584894, the description of the invention, etc. α are recorded. It ’s recorded. It ’s a method to determine the cause of the abnormality of the semiconductor machine. 'Losing force' is more likely to lose valuable expert knowledge, so it develops a misunderstanding to judge the cause of the abnormality of the semiconductor machine; the breaks are used to overcome the aforementioned long learning time; the lack of effective ... Failed to save and share the engineer's expert experience and knowledge,…, and other methods to effectively distinguish the importance of each diagnostic result. [Summary of the invention] The purpose of the present invention is to provide a system and method for judging the cause, so as to provide a systematic diagnosis process :: Come ;: The information of the subsystem of the computer integrated manufacturing system, so that its corresponding association with the root cause of the abnormality , Thus shortening the learning time for engineers. Another object of the present invention is to provide a system and method for judging the causes of abnormal semiconductor devices, so as to effectively save and share expert experience and knowledge. Yet another object of the present invention is to provide a system and method for judging the cause of an abnormality of a semiconductor machine, so as to indicate the importance of each diagnosis result, so that the engineer can smoothly identify which is the most likely cause of the abnormality. According to the above purpose of the present invention, a system for judging the cause of abnormality of a semiconductor machine is proposed, so as to deal with the abnormality triggered by the semiconductor machine. According to a preferred embodiment of the present invention, this judgment causes the abnormality of the semiconductor machine.

584894 V. Description of the invention (4) The system of the cause includes at least: The abnormal reason reasoning engine (Inference Engine) 'The abnormal reason reasoning engine further includes at least: the abnormal cause diagnosis element, wherein the abnormal cause diagnosis element has a plurality of diagnosis rules 'In order to provide users with multiple diagnostic results, some of the diagnostic rules ^ are used to check the data of multiple subsystems of the computer integrated manufacturing system one by one and correspond to multiple root causes of abnormalities; and self-learning mechanism components' Integrate the user's multiple judgment opinions on the diagnosis results into another part of the diagnosis rules. This self-learning mechanism file provides a reference weight value for each diagnosis result, so that the user recognizes the importance of each diagnosis result. degree. In addition, according to the above-mentioned object of the present invention, a method for judging the cause of the abnormality of the semiconductor and the stage is proposed to deal with the normal state triggered by the semiconductor machine. A preferred implementation includes at least: and the corresponding correlation exists in the diagnosis result as π is ", the diagnosis result is based on the reference proportion of the abnormal root, and the user recognizes each case, and this judgment provides a plurality of check computers integrated into a plurality of different rules, And according to the diagnosis, a reference weight value in the cause is transmitted to the data that causes a diagnosis node to follow the method of the present invention. The diagnosis rules are used to check the data one by one, whether the status has been checked first and each of them is " &Quot;, and one diagnosis result of each diagnosis result is used to make the diagnostic rules that cause the abnormality of the semiconductor machine, some of the multiple sub-systems of the manufacturing system are often the root cause; inspection abnormality produces the first inspection result; If the rule generates a plurality of diagnostic weight values; if the first test result is the most similar, a diagnostic result value is generated; the diagnostic result and each user, among which the importance of the reference specific gravity result; the user

584894 V. Description of the invention (5) Check the diagnosis result, and generate a plurality of judgment sounds to see if the conclusion J is the same as each judgment opinion, and produce the number of :: one diagnosis. The second inspection result is " yes ", Then increase the factory inspection results; re-value; and if the second inspection result is:-the reference ratio of fruit = diagnosis: shot, and give a starting ratio ΐ: mother a judgment judgment procedure. :: Ming 'Can shorten the learning time; provide effective heterogeneous procedures, save and straighten TΘ double abnormal diagnosis to distinguish the importance of each diagnosis result expert experience and knowledge; and effective

[Embodiment] X The judgment of the present invention results in the use of diagnostic rules for semiconductor devices, in order to respond to the semiconductor broadcast, the method of generating electrical power, and the abnormal state of the thousand-conductor device. To know the user's experts: ^ whole. The rule of shifting is provided, and a reference specific gravity is provided for each diagnosis result. According to the Γ diagram, it shows the intent of the system of the present invention that judges that the semiconductor device is abnormal. The component of this judgment that caused the abnormality of the semiconductor machine is the abnormal reason inference engine 200, among which the abnormal cause inference engine 200 to> includes the #common cause diagnosis element 21 and the self-learning mechanism element 220. f First, a plurality of subsystems in the computer integrated manufacturing system 100 are integrated into the abnormal cause reasoning engine 200. The computer integrated manufacturing system includes at least: the alarm subsystem 102, the manufacturing execution subsystem 104, and the statistical process. Control subsystem 106, real-time monitoring subsystem 108, preventive maintenance subsystem, abnormal processing record subsystem 112, and particle map subsystem 114. A major alarm event is triggered when an abnormality occurs on the semiconductor machine 530 (step Η) or

Page 9 V. Description of the invention (6) Events (step 62 °), in which major alarm events are created through alarms: human abnormal reason reasoning engine 2GG; crash events are entered into abnormal cause reasoning through the formulation of subsystem 104 Engine 200. The data of the various subsystems of the manufacturing system 100 are input to the cause of the abnormality, and / / engine i is used to diagnose the cause of the abnormality. 21 ○ Establish a plurality of diagnostic rules (name m 9 i), these diagnostic rules are used to provide users Multiple diagnosis 8. The self-learning mechanism element 220 provides each diagnosis result 218 with a specific gravity value 216. The reference specific gravity value 216 is used to indicate the degree of the diagnostic result 21 8 for reference by the user. The reference specific gravity value 216 is a weighted index assigned by the user. The diagnostic result 218 and the reference specific gravity value 216 of the diagnostic result 218 are stored in the result database 56. After receiving the diagnosis result 218, the user will use an obstacle diagnosis and processing device 300 with a graphical user interface (GUI; not shown) to generate a plurality of judgment opinions 318 for the diagnosis result 218. The judgment opinion 318jg is fed to the abnormal cause inference engine 20 (), and the judgment opinion is stored in the abnormal event database 520. Then, the self-learning mechanism element 220 compares each of the #break result 218 with the judgment opinion 318. If they are the same, it means that the user agrees that the diagnosis result 218 is the cause of the abnormality, and thus increases the reference weight value 21 6 thereof. When more users agree with the diagnosis result 2 丨 8, the reference weight value 21 6 is higher, and the user can know the importance of the diagnosis result 2 丨 8 accordingly. In searching for the cause of the abnormal disc, the user will first consider the diagnosis result 218 with a higher reference specific gravity value 216, thereby avoiding wasting time. Conversely, if the diagnosis result 21 8 is different from the judgment opinion 318, it means that the user does not agree that the precipitating result 218 is the cause of the abnormality, and therefore voluntarily

Page 10 584894 V. Description of the invention (7) Establish a new abnormal cause (ie, judgment opinion 318). At this time, Engine 200 will integrate the judgment opinion 318 into the diagnosis rule t, and add the weight value based on an inference. Then, the next time, the semiconductor machine 53 sends the sample port 1a to start the abnormal reason reasoning engine 20. This new anomaly source: different] can be provided, and the user can also use the anomaly score n with a graphical user interface to analyze the anomaly history of the semiconductor machine. ^ 衣 置 4 (i〇, please refer to FIG. 2, which shows the function of the abnormal cause 7G of the preferred embodiment of the present invention is not intended. The abnormal cause diagnosis element 21 of the '=: gauge, 2, where such as As mentioned earlier, some of the diagnostic rules 212 are for the cause of the abnormality (ie, judgment opinions) established by the user. The other diagnostic rules 21 2 are for Xiaolai to check the computer integrated manufacturing system and subsystems one by one. / Data, and correspondingly associated with a plurality of abnormal root causes 214, as shown in Figure 2, the abnormal root causes 214 include institutional reasons, particulate reasons, process reasons, and equipment reasons, and each abnormal root cause 214 has a plurality Sub-causes (not shown), for example, process reasons include pressure, temperature, process time, exhaust, etc. Diagnostic rules 212 can be provided by the machine manufacturer or obtained from interviews with Interview engineers. For example, When an abnormality occurs in the machine, the breaking rule 212 first checks whether the parameter in the database of the real-time monitoring subsystem 丨 〇8 exceeds the limit (〇ut Of c〇ntr〇1; 〇〇c), if the result is "Yes' Then check the sub-causes (such as: pressure, temperature, process time, and exhaust) in the process cause of abnormal root cause 214. In addition, the barrier, interruption and processing device 300 has a machine state of 31. 2. Major alarm status 320 and obstacle handling handover record 33. Among them, diagnosis result 218 was input to the machine downtime status 31. The alarm subsystem 102 entered major alarm information 584894. 5. Description of the invention (8) Material input to major The alarm status is 32, and the manufacturing execution subsystem 104 and the exception handling record subsystem 112 turn the crash data and the exception handling record into the obstacle handling transfer record 330. The exception analysis device 400 has a major alarm. Machine crash history 420 and obstacle handling history Shi Jing is so sigma green 430, 41 (T, f series · ^ 102f major alarm information is entered into the major alarm history 410 manufacturing execution subsystem will crash information Input to the machine report 420, and the exception handling record subsystem 112 records the exception 22 to the obstacle handling history. The reasons for the abnormality of the semiconductor machine judged by the present invention are as follows: W / melon please Referring to FIG. 3, a schematic flowchart of a method for judging the cause of a semiconductor device difference according to the present invention is shown. First, step 700 is performed to provide a diagnostic rule. As mentioned above, some diagnostic rules are used. Come one by one / check the data of the multiple subsystems of the computer integrated manufacturing system and correlate them; multiple root causes of abnormalities, and the other part of the diagnostic rules are obtained by making judgments. When an abnormality occurs on the machine, an abnormal state is triggered. Π0). The abnormal state can be a crash event, a major alarm event, or both. Go to step 712 to check whether the abnormal state already exists in the ^ rule. If the check result of step 712 is " YES ", a plurality of diagnosis results and a reference weight value of each diagnosis result are generated according to the diagnostic rules (half 714). J If the check result of step 712 is " no ", then The diagnosis result and the reference ratio of each diagnosis result are generated according to the most similar ones in the abnormal root cause (step 716). At this time, it is not possible to find the cause of the abnormal state of the machine from the root cause of the abnormality through the diagnosis rule. , So you can only use exceptions at all,

Page 12 584894 V. Description of the Invention (9) The most similar cause is the most likely cause of the abnormal condition to provide a diagnosis result. " Next, step 718 is performed to transmit the diagnosis result and the reference weight value of each diagnosis result to the user, where the reference weight value is used to let the user recognize the importance of each diagnosis result. The user then checks the diagnosis and generates a plurality of judgments (step 720). The so-called judgment opinion is made for the diagnosis result. When the user judges the diagnosis result to be acceptable, the diagnosis result is used as the judgment opinion; if the user judges the diagnosis result to be acceptable, the user will give it by himself. The reason is judgment. Next, step 722 is performed to check whether each diagnosis result is the same as each judgment of u. If the check result of step 722 is " Yes ", the reference specific gravity value of each #off result is increased. One of the main features of the present invention is a reference specific gravity value for the toilet result 'and is used according to the diagnosis result; The number of times to increase: its reference weight value. If you check $ ", add the opinion to the diagnostic rule and give this judgment: see a starting weight value. Then, the execution ends (step 75). Another =, after step 714 is completed, step 73 may be performed to refer to the specific gravity value to the result of the diagnosis result with the mother-v break ... As a result of the above-mentioned preferred embodiment of the present invention, it is a case of the cattle machine. The system and the cause of the cause of the cause caused half of the learning time; provide effective abnormal diagnosis procedures, and dry: shorten the resources of the unified subsystem # 'to make it relevant ;: J: manufacturing movement preservation and sharing engineer Expert experience and = root cause of abnormality; self-improving knowledge, and effective discrimination of 584894 V. Invention description (10) The importance of each diagnostic result. As a result, manpower and material resources are greatly saved, and the flow of valuable expert knowledge is avoided. Although the present invention has been disclosed as above with a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make various modifications and decorations without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the scope of the appended patent application.

Page 14 584894 Brief description of the drawings [Simplified description of the drawings] The first figure is a diagram showing the system of the present invention for judging the cause of the abnormality of the semiconductor device. FIG. 2 is a schematic diagram illustrating the function of the abnormality diagnosis component of the preferred embodiment of the present invention. Fig. 3 is a flow chart showing a method for judging the cause of an abnormality of a semiconductor device according to the present invention. [Simple description of component representative symbols] 100 computer integrated manufacturing system 102 alarm subsystem 104 manufacturing execution subsystem 106 statistical process control subsystem 108 real-time monitoring subsystem 110 preventive maintenance subsystem 112 exception processing record subsystem 114 particle map subsystem 200 abnormal Cause reasoning engine 210 Abnormal cause diagnosis element 212 Diagnosis rule 214 Abnormal root cause 216 Reference it Weight 218 Diagnosis result

Page 15 584894 Brief description of the diagram 2 2 0: Self-learning mechanism element 30 0: Obstacle diagnosis and processing device 3 1 0: Machine crash status 3 1 8: Judgment opinion 320: Major alarm status 330: Obstacle handling transfer record 400 : Anomaly analysis device 41 0: Major alarm history 420: Machine crash history 430: Obstacle handling history 520: Anomaly event database 530: Semiconductor machine 5 6 0: Results database 61 0: Trigger major alarm event 620: Trigger a crash event 700: Provide a diagnostic rule 71 0: Trigger an abnormal state 712: Check whether the abnormal state already exists in the diagnostic rule? 71 4: Generate diagnostic results and its reference weight value according to the diagnostic rules 71 6: Generate diagnostic results and its reference weight value according to the most similar ones among the root causes of the abnormality 718: Send to the user 720: The user checks the diagnostic result and generates a judgment Opinion 722: Check whether each diagnosis result is the same as each judgment?

Page 16 584894 Brief description of the diagram 724 726 728 730 750 Increase the reference weight value of the diagnosis result Add the judgment opinion to the diagnosis rule and give the initial weight value Save to the abnormal event database Save to the result database End execution #

Page 17

Claims (1)

584894 VI. Scope of Patent Application 1 · A system for determining the cause of an abnormality of a semiconductor machine, to handle an abnormal state triggered by a semiconductor machine. The system for determining the cause of an abnormality of a semiconductor machine includes at least: An abnormality reasoning engine (Inference Engine), wherein the abnormality reasoning engine further includes at least: an abnormality diagnosis element, wherein the abnormality diagnosis element has a plurality of diagnosis rules, so as to provide a user with a plurality of diagnosis results, part of which These diagnostic rules are used to check the multiple subsystems of a computer integrated manufacturing (CIM) system one by one. (Sub-systems) data, and corresponding to a plurality of abnormal root causes (Root-Causes); and a self-learning mechanism element, wherein the self-learning mechanism element generates the user for the diagnosis results The plurality of judgment opinions are integrated into the other part of the diagnostic rules, the self-learning mechanism element and provide-the, the diagnosis result-the reference weight A, so that the importance of the diagnosis results. mother 2. According to the judgment in Item 丨 of the scope of patent application, the cause of semi-conducting cause "" further includes at least a result database for storing the machine; this "judgment result and the reference weight value of each and every diagnosis result I. Isugi 3. The system for determining the cause as described in item 1 of the scope of the patent application, further includes at least an anomalous event database, a machine, and some judgments of the anomaly. It is used to store the Page 18 584894 6. Application scope 4. The system for judging the cause of the abnormality of the semiconductor machine as described in item 1 of the scope of application for the patent, wherein the user uses an obstacle diagnosis and processing device for the diagnosis results. To generate those judgments. ] The system for determining the cause of the abnormality of the semiconductor device as described in item 4 of the scope of the patent application, wherein the obstacle diagnosis and processing device has a Graphic User Interface (GUI). 3. The system for determining the cause of the abnormality of the semiconductor device as described in item 4 of the scope of the patent application, wherein the obstacle diagnosis and processing device has a machine down state, a critical alarm state, and an obstacle Handle transfer records. 7. The system for determining the cause of the abnormality of the semiconductor device as described in item 1 of the scope of patent application, wherein the user uses an abnormality analysis device to divide, a major alarm history record of the semiconductor device, a device Station crash history and an obstacle handling history. 8. The system for determining the cause of the abnormality of the semiconductor device as described in item 7 of the scope of the patent application, wherein the abnormality analysis device has a graphical user interface. 9 · As determined by item 1 of the scope of patent application Page 19 584894 6. The scope of the patent application Reason system, where the abnormal state is selected from a group consisting of a down event and a major alarm event. 10 · The system for determining the cause of the abnormality of the semiconductor machine as described in item 9 of the scope of the patent application, wherein the subsystems of the computer integrated manufacturing system include at least: an alarm subsystem, wherein the major alarm event is transmitted through The alarm subsystem enters the abnormal reason reasoning engine; a manufacturing execution subsystem (Manufacture Execution System; MES) 'Where the crash event entered the reasoning engine of the abnormal cause through the manufacturing execution subsystem; a statistical process control (SPC) subsystem; a real-time monitoring (RTM) subsystem A preventive maintenance subsystem (PMS); an exception handling record subsystem; and a particle map subsystem. 11. The system for judging the cause of the abnormality of the semiconductor device as described in item 丨 of the scope of patent application, wherein each and every abnormal root cause has a plurality of sub-causes. 12 · —A method of judging the cause of the abnormality of the semiconductor machine, by a different Scope of patent application ^ Reason reasoning engine to handle half ^ ^ 'Sad' This judgment caused an abnormality triggered by semiconductor AA σ to provide a plurality of diagnostic rules. The methods of reporting and the reasons for at least include: One inspection and one computer integration of these diagnostic rules are used to correlate the data of multiple subsystems to each subsystem, and the root cause of the six cranes; Already exists in these diagnostics # generate the first inspection result; and if the first inspection result is " yes ", each diagnosis result and each > I diagnosis rule produces a plural number One of the diagnostic results is referred to 4 if the test result is " No ", then according to the weight value; a reference, a result, and a reference to each of these diagnostic results will be The results of these 5 surveys and the paper 1 itk fiff JL J »r the mother of the card shirt is broken, and, the reference weight value is transmitted to a user 'where the reference weight value is used for the user Recognize the importance of every decisive result Degree; the tester should check the diagnosis results and generate a plurality of judgment opinions; check whether each of the diagnosis results is the same as each of the judgment opinions, and generate a second inspection result; if the second inspection result If "Yes", then increase the reference weight value of each of the diagnosis results; and if the second inspection result is "No", add each of the judgment opinions to the diagnosis rule of the wife and give A starting specific gravity value. 13. Semiconductor equipment threatened by judgment as described in item 12 of the scope of patent application 584894 VI. Method of applying for patent scope = Original =, where the abnormality reasoning engine includes at least two Γ cause diagnosis elements 'where the abnormal cause diagnosis element has these removal rules; and: 1 learning mechanism element' of which The self-learning mechanism element adds the judgment opinions to the rules, and provides the test ratio of each diagnosis result. The method for judging the abnormality of the semiconductor device as described in item 12 of the patent scope, including at least storing the diagnostic results and the reference weight values of each of the shirt break results to a result database. 15 * The original & Tian Shenming's judgment in the patent scope of item 12 caused abnormal semiconductor device data. The t method at least includes storing these judgments to an anomaly. The said judgment caused the abnormality of the semiconductor machine method, wherein the user uses an obstacle diagnosis and processing device to generate the reference opinions on the diagnosis results. The original reason is that the user of the semiconductor machine is abnormal because of the judgment described in item 16 of the scope of the patent application. The method for diagnosing and processing the obstacle has a pattern that makes the machine abnormal. The judgment described in item 8 leads to semiconducting Coffee 584894 6. The method of applying for the scope of the patent, wherein the obstacle diagnosis and processing device has a machine down state, a major alarm state, and a handover record of obstacle handling. 19. The method for judging the cause of the abnormality of a semiconductor device as described in item 12 of the scope of the patent application, wherein the user uses a 〆 anomaly analysis device to analyze a major alarm history of the semiconductor device, a device Crash history and an obstacle handling history. 20 · The semiconductor device is abnormal as judged in item 19 of the scope of patent application The method for determining the cause of the abnormality, wherein the abnormality analysis device has a graphical user interface 0 21. The method for determining the cause of the abnormality of the semiconductor device as described in item 12 of the scope of patent application, wherein the abnormal state is selected from the crash Events and major alarm events. 22. The method for judging the cause of the abnormality of the semiconductor device as described in the application patent ## 21, wherein the subsystems of the computer integrated manufacturing system include at least: an alarm subsystem, wherein the major alarm event Access to the abnormal reason reasoning engine through the alert subsystem; build an execution subsystem, where the crash event is entered into the reason cause of the abnormal reason through the manufacturing execution system;-statistical process control subsystem; Page 23 584894 VI. Scope of patent application An instant monitoring subsystem; a preventive maintenance subsystem; an exception handling record subsystem; and a particle map subsystem. 2 3. The method for judging the cause of the abnormality of the semiconductor device as described in item 12 of the scope of the patent application, wherein each and every abnormal root cause has a plurality of sub-causes. Page 24