Diagnostic apparatus, diagnostic system, and diagnostic method for semiconductor manufacturing apparatus

Abstract

The purpose of the present invention is to provide a technique for diagnosing, early, efficiently, and with minimal false alarms, an abnormality caused by aging deterioration of a device in a semiconductor manufacturing apparatus, on the basis of a change in a feature quantity trend obtained from a signal waveform acquired by a sensor mounted on the device.　To this end, a diagnostic apparatus for a semiconductor manufacturing apparatus according to the present invention comprises: a feature quantity computation unit that computes feature quantity from sensor waveform data corresponding to a process ID associated with a sensor item indicating a state of a device to produce a feature quantity trend in a time series; a trend type classification unit that classifies the feature quantity trend by trend type; a trend change amount computation unit that classifies a change in the feature quantity trend by trend change type, applies a trend change amount computation method corresponding to the trend change type on the basis of the trend type, and calculates a trend change amount; and a trend change abnormality diagnosis unit that determines an abnormality of the feature quantity trend in diagnosis from the feature quantity trend in diagnosis and the feature quantity trend in normal condition.

Description

Diagnostic device, diagnostic system, and diagnostic method for semiconductor manufacturing equipment

[0001] The present invention relates to a diagnostic device, a diagnostic system, and a diagnostic method for semiconductor manufacturing equipment.

[0002] Semiconductor manufacturing equipment performs processes to form microstructures on semiconductor wafers. Semiconductor manufacturing equipment typically undergoes periodic maintenance, such as cleaning and replacement, based on the number of wafers processed. However, unplanned maintenance can occur due to equipment degradation caused by aging and the accumulation of reaction by-products depending on usage. To reduce equipment downtime due to unplanned maintenance, continuous monitoring of the equipment's degradation status is required, and early countermeasures, such as cleaning and replacement, are required. To achieve such early countermeasures, a diagnostic system for semiconductor manufacturing equipment first calculates feature values ​​that indicate the characteristics of the sensor waveform data, which are time-series signals consisting of multiple sensor items sequentially acquired for each process from multiple sensors attached to the semiconductor manufacturing equipment. Furthermore, anomalies are diagnosed based on changes in the time-series trend of the feature values ​​(feature value trend), and an alarm is issued. For example, Patent Document 1 states, "The anomaly detection device applies statistical modeling to a summary value that summarizes observed values, thereby inferring a state in which noise has been removed from the summary value, and generates a predicted value that predicts the summary value one period ahead based on the inference. The anomaly detection device detects the presence or absence of an anomaly in the monitored device based on the predicted value." Furthermore, Patent Document 2 states, "A learning-type process anomaly diagnosis device that achieves accurate anomaly detection as well as anomaly diagnosis performance appropriate for practical process monitoring."

[0003] International Publication No. 2018 / 061842 Japanese Patent Application Laid-Open No. 2012-9064

[0004] However, in the case of abnormality diagnosis based on changes in feature value trends of semiconductor manufacturing equipment, prior art techniques can frequently generate false alarms in the following situations, making it difficult to implement efficient countermeasures based on the diagnosis results. First, feature value trends can be classified into multiple trend types depending on their characteristics, such as stationary trends and monotonically increasing trends. Therefore, when detecting changes in feature value trends, applying the same change detection method to feature value trends of different trend types can result in false alarms. For example, applying a level change detection method that is applied to stationary feature value trends to monotonically increasing feature value trends can result in prematurely detecting a level change. Furthermore, when detecting changes in feature value trends, the mere occurrence of a change does not necessarily determine whether an abnormality exists. Therefore, false alarms can occur if all detected changes are interpreted as abnormal. Furthermore, when equipment maintenance is performed, changes occur in the state of the equipment, resulting in changes in the feature value trend. However, the timing of maintenance is not always fixed, and there are situations where the maintenance timing cannot be determined in advance. Therefore, unless the starting point of the feature value trend used for diagnosis is appropriately set, changes in the feature value trend due to maintenance may be detected, resulting in false alarms. The anomaly detection techniques in Patent Documents 1 and 2 do not recognize this issue. Therefore, an object of the present invention is to provide a technology that enables early and efficient diagnosis of anomalies caused by deterioration over time of semiconductor manufacturing equipment, with fewer false alarms, based on changes in the feature value trend obtained from signal waveforms measured by sensors attached to the equipment.

[0005] In order to solve the above-described problems, one representative diagnostic device for semiconductor manufacturing equipment of the present invention includes a feature calculation unit that calculates feature amounts from sensor waveform data of process IDs linked to sensor items that indicate the status of equipment installed in each semiconductor manufacturing equipment of a semiconductor manufacturing equipment group, and creates a feature amount trend in a time series; a trend type classification unit that classifies the feature amount trends into trend types; a trend change amount calculation unit that classifies changes in the feature amount trends into trend change types, and calculates a trend change amount by applying a trend change amount calculation method according to the trend change type based on the trend type; and a feature amount trend change anomaly diagnosis unit that determines an anomaly in the feature amount trend at the time of diagnosis, based on the trend change type or trend change amount of the feature amount trend at the time of diagnosis and the trend change type or trend change amount distribution of the feature amount trend in a normal state.

[0006] According to the present invention, it is possible to diagnose abnormalities caused by the deterioration of semiconductor manufacturing equipment over time early and efficiently with few false alarms based on changes in feature value trends obtained from signal waveforms from sensors attached to the equipment. Problems, configurations, and effects other than those described above will become clear from the description of the following embodiments.

[0007] FIG. 1 is a configuration diagram of a semiconductor manufacturing apparatus and a diagnostic apparatus according to an embodiment. FIG. 2 is a diagram showing a flow until a feature amount trend is acquired from equipment of the semiconductor manufacturing apparatus. FIG. 3 is a flowchart showing an example of a flow of a feature amount trend creation and starting point setting process. FIG. 4 is a flowchart showing an example of a flow of a feature amount trend change analysis process under normal conditions for preparation for diagnosis. FIG. 5 is a diagram showing an example of a trend type of a feature amount trend under normal conditions. FIG. 6 is a diagram showing an example of a trend change type of a feature amount trend under normal conditions. FIG. 7 is a flowchart showing an example of a flow of an abnormality diagnosis process of a feature amount trend during diagnosis. FIG. 8 is a diagram showing an example of a displayed diagnosis result.

[0008] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. However, the present invention is not limited to the embodiment. In addition, in the description of the drawings, the same parts are generally designated by the same reference numerals, and repeated description thereof will be omitted.

[0009] FIG. 1 is a configuration diagram of a semiconductor manufacturing apparatus and a diagnostic apparatus according to an embodiment. <Semiconductor Manufacturing Apparatus> As shown in FIG. 1 , in this embodiment, a semiconductor manufacturing apparatus group 1 includes a plurality of semiconductor manufacturing apparatuses X10, Y11, etc., each installed for each lot. The semiconductor manufacturing apparatuses process a wafer 101, which is a sample, using the installed equipment in accordance with preset process conditions. Each semiconductor manufacturing apparatus also includes a sensor group 102 that measures the equipment status, and can acquire sensor values ​​(e.g., temperature and pressure) measured during processing or idle time as sensor waveform data. Data items in the sensor waveform data are called sensor items. Examples of semiconductor manufacturing apparatus, equipment, and sensors include a plasma processing apparatus, a microwave generator, and a current / voltage sensor, respectively.

[0010] <Diagnostic Device> As shown in FIG. 1 , the diagnostic device 2 includes a group of computers consisting of multiple computers X 30, Y 40, etc., and a server 50. Because the configurations of each computer in the group of computers are similar, only one computer will be described unless necessary. The computer 30 includes an execution unit 31 that acquires sensor waveform data corresponding to each semiconductor manufacturing device in the semiconductor manufacturing device group 1 and performs calculations, and a memory unit 32 that stores information necessary for the processing by the execution unit. The execution unit 31 includes a preprocessing unit 310, a feature value trend starting point setting unit 311, and a feature value calculation unit 312, while the memory unit 32 includes a sensor waveform storage unit 320 and a feature value trend storage unit 321. The server 50 includes an analysis unit 51 that aggregates the results of calculations performed by the execution units of the multiple semiconductor manufacturing devices, performs calculations, and displays the diagnosis results, and a memory unit 52 that stores information necessary for the processing by the analysis unit 51. The analysis unit 51 has a trend type classification unit 510, a trend change amount calculation unit 511, a trend change abnormality diagnosis unit 512, and a diagnosis result display unit 513, and the storage unit 52 has a trend change amount storage unit 520 and a diagnosis result storage unit 521.

[0011] In terms of hardware configuration, the execution unit 31 and the analysis unit 51 execute program processing using a computer processor such as a CPU. The diagnostic result display unit 513 may be connected to input devices such as a keyboard and a mouse as an input interface, and may also be connected to a display, printer, or other output interface. The storage units 32 and 52 may be random-access semiconductor memories, storage devices, or storage media (either volatile or nonvolatile) for storing data or programs (applications). The semiconductor manufacturing equipment group 1 is connected to the computer group (computer X30, computer Y40, ...) directly or via a network. The computer group (computer X30, computer Y40, ...) is also connected to the server 50 via a network. This allows each computer to perform high-speed calculations using sensor waveform data acquired from each semiconductor manufacturing equipment using the execution unit 31. The server 50 can analyze and display calculations and diagnostic results across the semiconductor manufacturing equipment group 1. The network can be connected via wire or wireless. Furthermore, the execution unit and storage unit of each computer may be arranged as separate pieces of hardware, or may be shared by switching or selecting functions on a single piece of hardware using software. Similarly, the computer group and server may be arranged as separate pieces of hardware, or may be shared by switching or selecting functions on a single piece of hardware using software. In either case, it is possible to build a system in which individual functional units organically perform their functions, just like running application programs on a common platform.

[0012] (Feature Quantity Trend) FIG. 2 is a diagram showing the process flow up to the acquisition of a feature quantity trend from a device of a semiconductor manufacturing apparatus. As shown in FIG. 2, the semiconductor manufacturing apparatus X10 has one or more devices, and each device is configured with sensor items, which are a series of waveform data generated by one or more sensors. Each sensor item is associated with a process target (wafer ID: e.g., wafer 1, wafer 2, etc.) sequentially supplied with data and the name of the process step (process ID: e.g., process A, process B, etc.) to be sequentially executed. Examples of process steps include a process for adjusting the state within the apparatus, a process for removing material from the apparatus, and a processing process. Furthermore, each process ID of each wafer ID is associated with sensor waveform data, and multiple feature quantities (e.g., feature quantity AP, which is the average value (P) of the waveform data of process A of wafer 1, feature quantity AQ, etc., which is the standard deviation (Q) of the waveform data) can be calculated for each sensor waveform data. When diagnosing an abnormality in a certain device, the time series trend of feature amounts associated with the device, consisting of the correspondence between feature amounts of the same type (e.g., feature amount AP) and wafer IDs, is called a feature amount trend (e.g., feature amount trend AP...). Note that feature amount calculations and feature amount trend acquisition are performed by diagnostic device 2. Although omitted from FIG. 2 , feature amount trends can also be acquired similarly from semiconductor manufacturing equipment Y11,..., etc. When diagnosing an abnormality in the feature amount trend of a specific semiconductor manufacturing equipment, past feature amount trend data of the semiconductor manufacturing equipment can be used as normal state data, but typically the same type of feature amount trend acquired from the same equipment, sensor item, and process ID of another semiconductor manufacturing equipment is used as normal state data.

[0013] <Computer Processing> Each process in the computer 30 will be described in detail below, along with the operation of each functional unit. First, sensor waveform data acquired from the semiconductor manufacturing equipment is temporarily stored in the sensor waveform storage unit 320. Next, the preprocessing unit 310 extracts sensor waveform data over a predetermined interval and converts it into a format that can be calculated, thereby performing data processing as preprocessing. Next, the system proceeds to the feature amount trend creation and starting point setting process, in which feature amounts are calculated to create a feature amount trend and a starting point for the feature amount trend is set. FIG. 3 is a flowchart showing an example of the flow of the feature amount trend creation and starting point setting process. As described below, the system executes step S1 for creating a feature amount trend, step S2 for acquiring a pre-diagnosis feature amount trend, step S3 for calculating a level change point of the feature amount trend, step S4 for determining whether the level change point satisfies the starting point condition, and step S5 for setting the level change point that satisfies the starting point condition as the starting point of the feature amount trend.

[0014] [Feature Trend Creation and Starting Point Setting Process] (S1: Feature Trend Creation Process) In step S1, the feature calculation unit 312 calculates feature quantities that indicate the characteristics of the sensor waveform data for each process to create a feature trend (feature trend creation process). As shown in FIG. 2 , one or more feature quantities are calculated for each sensor item. Statistical quantities such as the average value and standard deviation of the sensor waveform data can be used as feature quantities. For example, for waveform data of a series of current values ​​for a certain sensor item in a plasma processing apparatus, the average value for the entire processing period for each process ID or the average value for a specific time interval (e.g., an interval of several seconds to several minutes) within the processing period can be used as a feature quantity. Furthermore, a feature quantity that is highly likely to be related to deterioration can be calculated by comparing it with domain knowledge of the equipment, or multiple feature quantities can be calculated. Then, the feature quantities are traced in correspondence with the sequentially supplied wafer IDs to obtain a time-series feature trend. The obtained feature trend is stored in the feature trend storage unit 321.

[0015] (S2: Feature Amount Trend Acquisition Process) In step S2, the feature amount trend starting point setting unit 311 acquires the feature amount trend before diagnosis (feature amount trend acquisition process).

[0016] (S3: Trend Change Point Calculation Process) In step S3, the feature amount trend starting point setting unit 311 calculates a level change point, which is a trend change point, for multiple feature amount trends created for the same semiconductor manufacturing equipment (trend change point calculation process). A feature amount trend change point (trend change point) is a point at which the tendency of the feature amount trend (e.g., level, slope, variance, period) changes, and a level change point particularly refers to a point at which the tendency of the level (average) changes. There are no particular limitations on the method for calculating the level change point, but for example, in a method based on a control chart commonly used in quality control, a point at which xbar exceeds UCL / LCL (upper and lower limit thresholds set at 3σ) can be considered to be a level change point.

[0017] (S4: Starting Point Condition Determination Process) In step S4, the feature quantity trend starting point setting unit 311 determines whether or not a pre-specified number of feature quantity trends among the multiple feature quantity trends for the same semiconductor manufacturing equipment described above simultaneously have level change points, and whether the time interval between the level change points occurring before and after the level change points is equal to or longer than a specified time (starting point condition) (starting point condition determination process). The specified number and time are not particularly limited, but can be determined appropriately based on past experience. If the determination result is Yes, proceed to step S5; if No, end the feature quantity trend creation and starting point setting process. If No, the starting point is the start point of feature quantity trend measurement. Note that in this embodiment, the starting point condition is determined using a level change point, but it is also possible to specify and determine the starting point condition using other trend change points, such as a slope.

[0018] (S5: Feature Trend Starting Point Setting Process) In step S5, the feature trend starting point setting unit 311 sets the level change point determined in S4 to satisfy the starting point condition as the starting point of the feature trend. Furthermore, the trend change point is prevented from being detected as an anomaly due to maintenance (feature trend starting point setting process). Maintenance work on semiconductor manufacturing equipment typically extends to the entire interior of the equipment, and is expected to uniformly affect the operation of any equipment installed therein, regardless of the sensors or process operations. Therefore, in the case of maintenance, level change points are expected to be detected at approximately the same time in multiple feature trend data, regardless of the type of equipment, process operation, or feature. Once the starting point is set, the feature trend is divided at the starting point and updated so that the feature trend starts from the starting point. The updated feature trend is then stored in the feature trend storage unit 321.

[0019] <Processing on Server> The following describes in detail the processing executed by the server 50, along with the actions of the functional units. The server 50 performs a process of analyzing changes in the feature amount trend during normal times (T1 to T3) based on the feature amount trend data obtained by the computer 30, and then performs an abnormality diagnosis process on the feature amount trend during diagnosis (U1 to U5).

[0020] 4 is a flowchart showing an example of the flow of the process for analyzing changes in feature quantity trends in normal conditions for preparation for diagnosis. As will be described below, the process executes step T1 of acquiring feature quantity trends in normal conditions, step T2 of classifying the trend types of feature quantity trends in normal conditions, and step T3 of classifying the changes in feature quantity trends in normal conditions into trend change types and calculating the distribution of trend change amounts.

[0021] (T1: Feature Amount Trend Acquisition Process) In step T1, the trend type classification unit 510 acquires the feature amount trend during normal times stored in the feature amount trend storage unit 321 of each computer of the multiple semiconductor manufacturing devices (feature amount trend acquisition process). Here, "normal times" is set as a section within a specified time from the starting point of the feature amount trend set in the above-mentioned feature amount trend starting point setting process, or a section prior to the specified time from the time of the abnormality occurrence. Here, the specified time is not particularly limited, but can be set appropriately based on past experience.

[0022] (T2: Trend Type Classification Process) In step T2, the trend type classification unit 510 classifies the feature value trends in the normal state into trend types (trend type classification process). FIG. 5 is a diagram showing an example of trend types for feature value trends in the normal state. As shown in FIG. 5, possible types of feature value trends include a stationary type in which the level (average) and variance remain constant over time, a monotonic type in which the slope increases or decreases in a constant direction, a periodic type in which periodic changes are repeated, and a random fluctuation type in which no regular changes occur. For example, the monotonic type may be further subdivided into a monotonically increasing type or a monotonically decreasing type. Here, the trend type classification method is not particularly limited. For example, the trend type may be determined using an index indicating the trend tendency, such as a stationary test or rank correlation, or may be determined using a machine learning classifier such as a pre-trained decision tree, support vector machine, or k-nearest neighbor method. If the classification results for the trend types of corresponding feature value trends of the same type in multiple semiconductor manufacturing devices are not the same, the classification may be determined by majority vote.

[0023] (T3: Trend Change Calculation Process) In step T3, the trend change calculation unit 511 classifies changes in feature value trends during normal times into trend change types, applies a trend change calculation method appropriate for the trend change type based on the trend type, calculates a trend change point and trend change amount, and further calculates a trend change amount distribution from the trend change amounts of multiple feature value trends (trend change calculation process). FIG. 6 is a diagram showing examples of trend change types for feature value trends during normal times. As shown in FIG. 6 , possible types include a level change type in which the trend level trend changes discontinuously, a variance change type in which the trend variance trend changes discontinuously, a trend slope change type in which the trend slope trend changes discontinuously, and a cyclic change type in which the trend cycle trend changes discontinuously. The types may also be further subdivided based on whether the direction of change is positive or negative (for example, in the case of a level change, whether the level change before and after the trend change point is positive (increase) or negative (decrease)). The classification method for trend change types is similar to that for trend types, and therefore a description thereof will be omitted. Then, the amount of change (trend change amount) before and after the trend change point is calculated according to the trend change type (for example, in the case of a level change, the difference between the level values ​​before and after the trend change point). Furthermore, a trend change amount distribution is obtained when the trend change amount values ​​are distributed across multiple feature quantity trends of the same type obtained from the semiconductor manufacturing equipment group. Data related to the trend type, trend change type, trend change amount, and trend change amount distribution of the feature quantity trend obtained in this way is stored in the trend change amount storage unit 520. The trend type data may be stored in step T2.

[0024] When calculating trend change, for example, if the trend type is stationary, a trend change calculation method that detects level changes, variance changes, and trend slope changes is applied because the trend does not inherently have periodicity. Applying a trend change calculation method that detects periodic changes increases the risk of false alarms due to misidentification. By not applying trend change calculation methods for trend change types that are not meaningful to detect based on the trend type, false alarms can be efficiently prevented. While there are no particular limitations on the trend change calculation method, for example, a level change can be detected using an x-R (x-bar-R) control chart, and the point at which x (x-bar) exceeds the 3σ control limit can be determined as the level change point. Furthermore, a calculation method based on the likelihood of a normal distribution can be applied to a variance change; a calculation method based on the likelihood of a piecewise linear model can be applied to a trend slope change; and a calculation method such as k-nearest neighbors or ARIMA, which takes seasonality into account, can be applied to a periodic change.

[0025] [Processing for Diagnosing Abnormalities in Feature Amount Trends During Diagnosis] Next, using the results of the processing for analyzing changes in the feature amount trend during normal times, processing for diagnosing abnormalities in the feature amount trend during diagnosis is performed from the starting point to be diagnosed. Figure 7 is a flowchart showing an example of the flow of processing for diagnosing abnormalities in the feature amount trend during diagnosis. As described below, the process includes Step U1 for acquiring the feature amount trend during diagnosis, Step U2 for classifying the trend type of the feature amount trend during diagnosis, Step U3 for classifying the changes in the feature amount trend during diagnosis into trend change types and calculating the amount of trend change, Step U4 for determining whether the change in the feature amount trend during diagnosis is abnormal, Step U5 for diagnosing the change in the feature amount trend during diagnosis as abnormal if an abnormality is determined, and Step U6 for displaying the diagnosis result. Content that overlaps with the processing for analyzing changes in the feature amount trend during normal times will be omitted as appropriate.

[0026] (U1: Feature Amount Trend Acquisition Process) In step U1, the trend type classification unit 510 acquires the feature amount trend at the time of diagnosis (feature amount trend acquisition process). The feature amount trend to be diagnosed can be selected as appropriate, but it is also possible to perform a diagnosis using all feature amounts in advance and then select a feature amount with good diagnostic accuracy.

[0027] (U2: Trend Type Classification Process) In step U2, the trend type classification unit 510 classifies the diagnostic feature amount trends into trend types, similar to step T2 for the normal feature amount trends (trend type classification process).

[0028] (U3: Trend Change Amount Calculation Process) In step U3, the trend change amount calculation unit 511 classifies the change in the feature amount trend during diagnosis into trend change types, as in step T3 for the feature amount trend during normal operation, and calculates the trend change point and trend change amount based on the trend type by applying a trend change amount calculation method appropriate for the trend change type. Data related to the trend type, trend change type, and trend change amount of the obtained feature amount trend is stored in the trend change amount storage unit 520 (trend change amount calculation process). The trend type data may be stored in step U2. Here too, false alarms can be efficiently prevented by not applying a trend change amount calculation method for trend change types that are not meaningful to detect based on the trend type.

[0029] (U4: Feature Amount Trend Change Anomaly Determination Process) In step U4, when the trend change anomaly diagnosis unit 512 detects a trend change point by calculation for the feature amount trend at the time of diagnosis, it determines whether the trend change type is a trend change type that is not stored in the trend change amount storage unit 520 (i.e., whether the trend change type is a trend change type that was not detected for the normal feature amount trend), or whether the trend change amount of the trend change type deviates from the trend change amount distribution stored in the trend change amount storage unit 520 (feature amount trend change anomaly determination process). If the determination result is Yes, the process proceeds to step U5, and if No, it is determined that no anomaly is present and the process proceeds to step U6.

[0030] (U5: Feature Amount Trend Change Abnormality Diagnosis Processing) In step U5, the trend change abnormality diagnosis unit 512 diagnoses that the feature amount trend change at the time of diagnosis is an abnormal change that cannot be confirmed when multiple devices are normal (feature amount trend change abnormality diagnosis processing).

[0031] (U6: Diagnostic Result Storage and Display Processing) In step U6, the diagnostic result storage unit 521 stores the diagnostic results for the feature quantity trend changes during the diagnosis, and the diagnostic result display unit 513 acquires and displays the diagnostic results from the diagnostic result storage unit 521 (diagnostic result storage and display processing). FIG. 8 is a diagram showing an example of the displayed diagnostic results. The feature quantity trends during the diagnosis of the selected semiconductor manufacturing equipment and device are identified and numbered by the process ID, wafer ID, sensor item, and feature quantity name, and a list of information on the trend change type, trend change point, and trend change amount stored in the trend change amount storage unit 520, the normal trend type obtained from the same type of feature quantity trends of other semiconductor manufacturing equipment, and the result of the abnormality diagnosis (presence or absence of an abnormality) is displayed. Furthermore, by checking the "Trend Display" item, the feature quantity trends of the diagnosis target are also displayed in graph form.

[0032] This allows for efficient countermeasures, for example, when an equipment error occurs, by prioritizing countermeasures for the equipment corresponding to the sensor item linked to the feature amount trend diagnosed as abnormal. Also, by performing maintenance on the corresponding equipment early after an abnormality is diagnosed and before an equipment error occurs, unplanned maintenance can be prevented, leading to efficient countermeasures.

[0033] Although the embodiment has been described above, the present invention is not limited to the embodiment and various modifications are possible without departing from the spirit of the present invention. For example, if it is difficult to install the server 50, the computer 30 may be configured to perform part of the processing of the server 50.

[0034] The present invention is not limited to the following embodiments. (Aspect 1) A diagnostic device for semiconductor manufacturing equipment, comprising: a feature calculation unit that calculates feature amounts from sensor waveform data of process IDs associated with sensor items indicating the status of equipment installed in each semiconductor manufacturing equipment in a group of semiconductor manufacturing equipment, and creates a time-series feature amount trend; a trend type classification unit that classifies the feature amount trends into trend types; a trend change amount calculation unit that classifies changes in the feature amount trends into trend change types and calculates a trend change amount by applying a trend change amount calculation method corresponding to the trend change type based on the trend type; and a trend change anomaly diagnosis unit that determines an anomaly in the feature amount trend at the time of diagnosis based on the trend change type or trend change amount of the feature amount trend at the time of diagnosis and the trend change type or trend change amount distribution of the feature amount trend in a normal state. (Aspect 2) The diagnostic device for semiconductor manufacturing equipment according to Aspect 1, wherein trend types that can be classified by the trend type classification unit include a stationary type, a monotonic type, a periodic type, and a random variation type. (Aspect 3) The diagnostic device for semiconductor manufacturing equipment according to Aspect 1 or 2, characterized in that trend change types that can be classified by the trend change amount calculation unit include a level change type, a variance change type, a trend slope change type, and a periodic change type. (Aspect 4) The diagnostic device for semiconductor manufacturing equipment according to any one of Aspects 1 to 3, further comprising a feature amount trend starting point setting unit that sets a starting point of a feature amount trend. (Aspect 5) The diagnostic device for semiconductor manufacturing equipment according to Aspect 4, characterized in that the feature amount trend starting point setting unit determines whether or not a pre-specified number or more of feature amount trends, of multiple feature amount trends created for the same semiconductor manufacturing equipment, simultaneously have level change points, and whether the time interval between the level change points occurring before and after the level change points is equal to or longer than a specified time (starting point condition).(Aspect 6) A diagnostic system for semiconductor manufacturing equipment, characterized in that program instructions are applied to a computer processor to execute: a feature quantity trend creation process that calculates feature quantities and creates a time-series feature quantity trend from sensor waveform data of process IDs linked to sensor items that indicate the status of equipment installed in each semiconductor manufacturing equipment of a group of semiconductor manufacturing equipment; a trend type classification process that classifies the feature quantity trends into trend types; a trend change amount calculation process that classifies changes in the feature quantity trends into trend change types and calculates trend change amounts by applying a trend change amount calculation method corresponding to the trend change type based on the trend type; and a feature quantity trend change anomaly determination process that determines an anomaly in the feature quantity trend at the time of diagnosis from the trend change type or trend change amount of the feature quantity trend at the time of diagnosis and the trend change type or trend change amount distribution of the feature quantity trend in a normal state. (Aspect 7) The diagnostic system for semiconductor manufacturing equipment according to Aspect 6, characterized in that trend types that can be classified by the trend type classification process include a stationary type, a monotonic type, a periodic type, and a random variation type. (Aspect 8) The diagnostic system for semiconductor manufacturing equipment according to Aspect 6 or 7, characterized in that trend change types that can be classified by the trend change amount calculation process include a level change type, a variance change type, a trend slope change type, and a periodic change type. (Aspect 9) The diagnostic system for semiconductor manufacturing equipment according to any one of Aspects 6 to 8, which executes feature amount trend starting point setting process that sets a starting point of a feature amount trend. (Aspect 10) The diagnostic system for semiconductor manufacturing equipment according to Aspect 9, characterized in that it executes starting point condition determination process that determines whether or not a pre-specified number or more of feature amount trends, out of multiple feature amount trends created for the same semiconductor manufacturing equipment, have simultaneous level change points and whether or not the time interval between level change points occurring before and after the level change points is equal to or longer than a specified time (starting point condition).(Aspect 11) A method for diagnosing semiconductor manufacturing equipment, comprising: a feature quantity trend creation process step of calculating feature quantities and creating a time-series feature quantity trend from sensor waveform data of process IDs linked to sensor items indicating the status of equipment installed in each semiconductor manufacturing equipment of a group of semiconductor manufacturing equipment; a trend type classification process step of classifying the feature quantity trends into trend types; a trend change amount calculation process step of classifying changes in the feature quantity trends into trend change types and calculating trend change amounts by applying a trend change amount calculation method corresponding to the trend change type based on the trend types; and a feature quantity trend change anomaly determination process step of determining an anomaly in the feature quantity trend at the time of diagnosis from the trend change type or trend change amount of the feature quantity trend at the time of diagnosis and the trend change type or trend change amount distribution of the feature quantity trend in a normal state. (Aspect 12) The method for diagnosing semiconductor manufacturing equipment according to Aspect 11, wherein trend types that can be classified by the trend type classification process step include a stationary type, a monotonic type, a periodic type, and a random variation type. (Aspect 13) The diagnostic method for semiconductor manufacturing equipment according to Aspect 11 or 12, characterized in that trend change types that can be classified by the trend change amount calculation processing step include a level change type, a variance change type, a trend slope change type, and a periodic change type. (Aspect 14) The diagnostic method for semiconductor manufacturing equipment according to any one of Aspects 11 to 13, including a feature amount trend starting point setting processing step that sets a starting point of a feature amount trend. (Aspect 15) The diagnostic method for semiconductor manufacturing equipment according to Aspect 14, characterized in including a starting point condition determination processing step that determines whether or not a pre-specified number or more of feature amount trends, out of multiple feature amount trends created for the same semiconductor manufacturing equipment, simultaneously have level change points, and whether or not the time interval between level change points occurring before and after the level change points is equal to or longer than a specified time (starting point condition).

[0035] 1...semiconductor manufacturing equipment group, 2...diagnosis device, 10, 11...semiconductor manufacturing equipment, 30, 40...computer, 31...execution unit, 32, 52...storage unit, 50...server, 51...analysis unit, 101...wafer, 102...sensor group, 310...preprocessing unit, 311...feature amount trend starting point setting unit, 312...feature amount calculation unit, 320...sensor waveform storage unit, 321...feature amount trend storage unit, 510...trend type classification unit, 511...trend change amount calculation unit, 512...trend change abnormality diagnosis unit, 513...diagnosis result display unit, 520...trend change amount storage unit, 521...diagnosis result storage unit

Claims

1. A diagnostic device for semiconductor manufacturing equipment, comprising: a feature calculation unit that calculates feature amounts from sensor waveform data of process IDs linked to sensor items that indicate the status of equipment installed in each semiconductor manufacturing equipment of a group of semiconductor manufacturing equipment, and creates a feature amount trend in a time series; a trend type classification unit that classifies the feature amount trends into trend types; a trend change amount calculation unit that classifies changes in the feature amount trends into trend change types and calculates a trend change amount by applying a trend change amount calculation method corresponding to the trend change type based on the trend type; and a trend change anomaly diagnosis unit that determines an anomaly in the feature amount trend at the time of diagnosis based on the trend change type or trend change amount of the feature amount trend at the time of diagnosis and the trend change type or trend change amount distribution of the feature amount trend in a normal state.

2. The diagnostic device for semiconductor manufacturing equipment according to claim 1, wherein the trend types that can be classified by the trend type classification unit include a stationary type, a monotonic type, a periodic type, and a random fluctuation type.

3. The diagnostic device for semiconductor manufacturing equipment according to claim 1 or 2, wherein the trend change types that can be classified by the trend change amount calculation unit include a level change type, a variance change type, a trend slope change type, and a periodic change type.

4. The diagnostic device for semiconductor manufacturing equipment according to claim 1 or 2, further comprising a feature amount trend starting point setting unit that sets a starting point of a feature amount trend.

5. A diagnostic device for semiconductor manufacturing equipment according to claim 4, characterized in that the feature trend starting point setting unit determines whether or not a pre-specified number of feature trend data, among multiple feature trend data created for the same semiconductor manufacturing equipment, simultaneously have level change points, and whether or not the time interval between the level change points occurring before and after the level change points is equal to or longer than a specified time (starting point condition).

6. A diagnostic system for semiconductor manufacturing equipment, characterized in that program instructions are applied to a computer processor to execute: a feature quantity trend creation process that calculates feature quantities and creates a feature quantity trend in a time series from sensor waveform data of process IDs linked to sensor items that indicate the status of equipment installed in each semiconductor manufacturing equipment of a group of semiconductor manufacturing equipment; a trend type classification process that classifies feature quantity trends into trend types; a trend change amount calculation process that classifies changes in feature quantity trends into trend change types and calculates trend change amounts by applying a trend change amount calculation method corresponding to the trend change type based on the trend type; and a feature quantity trend change anomaly determination process that determines an anomaly in the feature quantity trend at the time of diagnosis based on the trend change type or trend change amount of the feature quantity trend at the time of diagnosis and the trend change type or trend change amount distribution of the feature quantity trend in a normal state.

7. The diagnostic system for semiconductor manufacturing equipment according to claim 6, wherein trend types that can be classified by the trend type classification process include stationary type, monotonic type, periodic type, and random fluctuation type.

8. The diagnostic system for semiconductor manufacturing equipment according to claim 6 or 7, wherein the trend change types that can be classified by the trend change amount calculation process include a level change type, a variance change type, a trend slope change type, and a periodic change type.

9. The diagnostic system for semiconductor manufacturing equipment according to claim 6 or 7, further comprising: a feature amount trend starting point setting process for setting a starting point of a feature amount trend.

10. A diagnostic system for semiconductor manufacturing equipment according to claim 9, characterized in that it executes a starting point condition determination process to determine whether or not a pre-specified number of feature quantity trends among multiple feature quantity trends created for the same semiconductor manufacturing equipment simultaneously have level change points, and whether or not the time interval between level change points occurring before and after said level change points is equal to or longer than a specified time (starting point condition).

11. A method for diagnosing semiconductor manufacturing equipment, comprising: a feature quantity trend creation processing step for calculating feature quantities and creating a time-series feature quantity trend from sensor waveform data of process IDs linked to sensor items indicating the status of equipment installed in each semiconductor manufacturing equipment of a group of semiconductor manufacturing equipment; a trend type classification processing step for classifying feature quantity trends into trend types; a trend change amount calculation processing step for classifying changes in feature quantity trends into trend change types and calculating trend change amounts by applying a trend change amount calculation method corresponding to the trend change type based on the trend type; and a feature quantity trend change anomaly determination processing step for determining an anomaly in the feature quantity trend at the time of diagnosis based on the trend change type or trend change amount of the feature quantity trend at the time of diagnosis and the trend change type or trend change amount distribution of the feature quantity trend in normal times.

12. The method for diagnosing semiconductor manufacturing equipment according to claim 11, wherein trend types that can be classified by the trend type classification processing step include a stationary type, a monotonic type, a periodic type, and a random fluctuation type.

13. The method for diagnosing semiconductor manufacturing equipment according to claim 11 or 12, wherein the trend change types that can be classified by the trend change amount calculation processing step include a level change type, a variance change type, a trend slope change type, and a periodic change type.

14. The method for diagnosing semiconductor manufacturing equipment according to claim 11 or 12, further comprising a step of processing for setting a starting point of a feature amount trend.

15. A diagnostic method for semiconductor manufacturing equipment according to claim 14, characterized in that it includes a starting point condition determination process step for determining whether or not a pre-specified number of feature quantity trends, among multiple feature quantity trends created for the same semiconductor manufacturing equipment, simultaneously have level change points, and whether or not the time interval between level change points occurring before and after said level change points is equal to or longer than a specified time (starting point condition).

Images