Patent Application
20200159197
The present application claims priority from Japanese application JP 2018-218491, filed on Nov. 21, 2018, the contents of which is hereby incorporated by reference into this application.
The present invention relates to a technique such as information processing, and relates to a technique of controlling or specifying a manufacturing condition in a manufacturing flow.
In a manufacturing system in the manufacturing industry, quality of a manufactured product may vary depending on setting or control of a manufacturing condition in a manufacturing flow. Therefore, an information processing system or the like (sometimes referred to as a manufacturing condition specifying system) for specifying a suitable manufacturing condition is developed so as to maintain or improve the quality of the manufactured product.
PTL 1, PTL 2, and PTL 3 are listed as related-art examples of specifying a manufacturing condition. PTL 1 discloses a method of managing product quality or the like in which a probability model is built from a past manufacturing condition and a manufacturing condition that matches a target value is calculated. PTL 2 discloses a method of predicting an output value or the like in which a plurality of predicted values are output from past performance data. PTL 3 discloses a machine learning system or the like in which a non-parametric expressed class set is generated, in other words, the number of dimensions is reduced, when the number of inputs to be input at the time of using a model is larger than a predetermined number.
PTL 1: JP-A-2013-84057
PTL 2: JP-A-2011-39763
PTL 3: JP-A-2013-205890
The manufacturing state changes every day in a site of the manufacturing industry. The manufacturing state change is, for example, a change in a state of a manufacturing device in a manufacturing process. A state of parameters such as a current, voltage, temperature, pressure, or the like may change in a manufacturing process, for example, by continuing an operation in a manufacturing flow. For example, a state of the manufacturing device may change when a maintenance operation is performed on the manufacturing device.
In order to maintain or improve product quality, it is effective to specify a suitable manufacturing condition according to the manufacturing state change and set the manufacturing condition in the manufacturing flow for operation. The product quality is a predetermined evaluation index value, and is obtained, for example, as a value of an inspection result of a quality inspection process, for example, a yield.
A system for predicting a manufacturing condition using a learning model on a computer is listed as a related-art example of a manufacturing condition specifying system. The system builds a model using performance data including a manufacturing condition and quality of a manufacturing flow, and predicts a suitable manufacturing condition based on learning in a predetermined learning model.
However, there is a room for improvement for the manufacturing condition specifying system as a related-art example in terms of specifying the suitable manufacturing condition according to the manufacturing state change. For example, immediately after the manufacturing state change, the number of data obtained from performance of an operation is small, that is, the number of data for advancing the learning in the learning model is small. Therefore, it is difficult to improve prediction accuracy of the model. In order to improve the prediction accuracy of the model, it is necessary to input the number of data of a certain degree or more to the model so as to advance the learning, but it takes time. When the prediction accuracy of the model is low immediately after the manufacturing state change, a suitable manufacturing condition cannot be specified. As a result, the product quality cannot be maintained or improved.
An object of the invention is to provide a technique capable of specifying a suitable manufacturing condition and maintaining or improving product quality even when the manufacturing condition may change in a manufacturing condition specifying system. Other problems, configurations, effects, and the like will be described in the detailed description of the invention.
A representative embodiment of the invention includes the following configurations. A manufacturing condition specifying system according to an embodiment includes a computer that specifies a manufacturing condition in each manufacturing process of a manufacturing flow. The computer uses manufacturing condition data and quality data at a plurality of time points including a current time point from the manufacturing flow to build a model related to the manufacturing condition and quality; at a time of building the model, builds models each being built for a manufacturing state change as a plurality of models in a case of including manufacturing state changes in the manufacturing process of the manufacturing flow; uses the model and a quality target value to calculate, in the model, a predicted value of manufacturing condition data at a next time point as first data based on learning in a first learning model; uses the model as well as the manufacturing condition data and quality data at the current time point to predict quality data at a next time point and calculate a quality error between the quality data at the next time point and the quality data at the current time point; uses the first data and the quality error to specify manufacturing condition data at the next time point based on learning in a learning model; and stores and outputs information including the specified manufacturing condition data at the next time point.
According to the representative embodiment of the invention, even when the manufacturing condition may change in the manufacturing condition specifying system, a suitable manufacturing condition can be specified, and the product quality can be maintained or improved.
Hereinafter, embodiments of the invention will be described in detail with reference to the drawings. It should be noted that in all the drawings for describing the embodiments, the same components are denoted by the same reference numerals in principle, and a repetitive description thereof will be omitted.
A manufacturing condition specifying system according to an embodiment of the invention will be described with reference to
The manufacturing condition specifying system according to the embodiment builds one or more models related to a manufacturing condition and quality using manufacturing condition data and quality data at a plurality of time points including a current time point, which includes a manufacturing state change in the manufacturing flow, based on performance of an operation. The system uses manufacturing condition data and models up to the current time point to calculate a predicted value of manufacturing condition data at a next time point based on learning in a learning model. The system further uses quality data and the models up to the current time point to calculate a predicted value of quality at a next time point and calculate a quality error between the quality at the current time point and the quality at the next time point. The system uses the above data to specify suitable manufacturing condition data at the next time point based on the learning in the learning model. Accordingly, the suitable manufacturing condition data at the next time point after the manufacturing state change is obtained. The system stores and outputs information such as specified manufacturing condition data. The manufacturing condition data is reflected, that is, set in the manufacturing flow and the operation is executed. Accordingly, product quality after the manufacturing state change can be maintained or improved.
Further, the manufacturing condition specifying system according to the embodiment converts the manufacturing condition data obtained using the models into a subspace so as to reduce the number of dimensions. The conversion reduces the number of dimensions of data without reducing an amount of information of the model and enables subsequent learning. The system uses subspace data obtained after the conversion to specify a suitable manufacturing condition based on the learning model. At the time of specifying processing, data having a small number of dimensions can be regarded as an input, and the processing can be executed effectively.
The computer 1 includes an input and output unit 11, a communication unit 12, a display unit 20, a control unit 30, a storage unit 40, and the like. These units are connected by a bus or the like (not shown). The input and output unit 11 is connected with an input device (for example, a keyboard or a mouse), a display device (for example, a liquid crystal display, a touch panel), or other output devices (for example, a printer) (not shown), and receives an operation of the user. The communication unit 12 includes a communication interface device for a communication network such as a LAN outside the computer 1, and executes a communication processing between an external server device and a manufacturing system device. The communication unit 12 acquires data such as manufacturing condition data or information from an external device under the control of the control unit 30.
The display unit 20 includes a screen (corresponding screen data or the like) in the manufacturing condition specifying system, and displays the screen on a display screen of a display device. Various types of information such as a manufacturing condition, quality, and a model are displayed on various types of screens which will be described later. The screen functions as a Graphical User Interface (GUI) of the manufacturing condition specifying system. GUI components such as a window, a scroll bar, a list box, a button, and the like are displayed on the screen, and a user operation can be executed through the GUI components.
The user or the computer 1 acquires necessary data such as manufacturing condition data or quality data from each manufacturing processing in a manufacturing flow (will be described in
The user inputs necessary data to the computer 1, and the computer 1 executes a calculation. The computer 1 specifies a suitable manufacturing condition in a target manufacturing flow by a calculation using a model built by using the input data. The user checks the suitable manufacturing condition obtained by the computer 1 on the screen and reflects the suitable manufacturing condition in a manufacturing flow of the target manufacturing system. That is, a parameter value used for control corresponding to the manufacturing condition is set in a manufacturing device in each manufacturing process of the manufacturing flow. The computer 1 may be configured to transmit and set the manufacturing condition or the like to the manufacturing device in each manufacturing process of the manufacturing flow through communication.
The control unit 30 is, in other words, a processor, and includes known elements such as a CPU, a RAM, and a ROM. The control unit 30 has the following configurations as a main processing unit implemented based on program processing. That is, the control unit 30 includes a model building unit 31, a model manufacturing condition specifying unit 32, a quality specifying unit 33, a manufacturing condition determining unit 34, a subspace specifying unit 35, and a manufacturing condition specifying unit 36. The control unit 30 stores various types of data related to the manufacturing condition specifying in the storage unit 40 and manages the information.
The storage unit 40 stores various types of data and information related to the manufacturing condition specifying. The storage unit 40 may be configured with a nonvolatile memory, a storage device, or the like, and may be configured with an external DB server or the like. The storage unit 40 includes a manufacturing condition data storage unit 41, a quality data storage unit 42, a model storage unit 43, a first learning model storage unit 44, a second learning model storage unit 45, a subspace storage unit 46, and a third learning model storage unit 47. The manufacturing condition data storage unit 41 stores manufacturing condition data acquired from the manufacturing flow and manufacturing condition data specified by the computer 1. The quality data storage unit 42 stores quality data and the like acquired from the manufacturing flow. The model storage unit 43 stores data of a model that is built by the computer 1 and related to the manufacturing condition and quality. The first learning model storage unit 44 stores data of a first learning model to be described later. The second learning model storage unit 45 stores data of a second learning model to be described later. The subspace storage unit 46 stores subspace data to be described later. The third learning model storage unit 47 stores data of a third learning model to be described later.
The display unit 20 executes processing of displaying various types of data or information on a screen of a display device based on processing of the control unit 30. The display unit 20 includes a model display unit 21, a manufacturing condition display unit 22, a subspace display unit 23, a quality display unit 24, and a manufacturing condition specifying display unit 25. The model display unit 21 graphically displays a built model on a screen (to be described in
The model building unit 31 executes processing (step S2 to be described in
The model manufacturing condition specifying unit 32 executes processing (step S4 to be described in
The manufacturing condition determining unit 34 executes processing (step S6 to be described in
The quality specifying unit 33 executes processing (step S5 to be described in
The subspace specifying unit 35 uses data including the manufacturing condition data (first data) at the next time point obtained from the model manufacturing condition specifying unit 32, the manufacturing condition data (second data) at the next time point obtained from the manufacturing condition determining unit 34, and the quality error obtained from the quality specifying unit 33 as input data. The subspace specifying unit 35 executes processing (step S7 to be described in
The manufacturing condition specifying unit 36 finally executes processing (step S8 to be described in
Manufacturing condition data is associated with each process. The manufacturing condition data (corresponding data item) is a parameter value for controlling states of the manufacturing device and the sensor. Examples of general parameters include a current, voltage, temperature, pressure or the like. The computer 1 in the manufacturing condition specifying system according to the embodiment acquires manufacturing condition data, quality data, manufacturing flow configuration information, or the like from the manufacturing flow of the manufacturing system. The user or the computer 1 can acquire the manufacturing condition data or the like from the manufacturing device in each process or a control device. The manufacturing condition specifying system can acquire, as monitoring data of performance of an operation, the manufacturing condition data and the quality data at each time point in time series including a manufacturing state change in the manufacturing flow.
In the embodiment, the quality inspection process as a last process of the manufacturing flow is included. The quality data is obtained from the quality inspection process, and a method is applied to associate the quality data and the manufacturing condition data. The invention is not limited thereto, and a method can be applied similarly, for example, even in a case where a quality inspection flow exists independently of the manufacturing flow.
The manufacturing condition specifying system according to the embodiment builds the causality model based on the manufacturing condition data and the quality data acquired from the above-mentioned manufacturing flow. The causality model on a lower side of
The manufacturing condition specifying system according to the embodiment specifies suitable manufacturing condition data based on a built model, and stores and outputs the specified manufacturing condition data. By referring to the manufacturing flow configuration information or the like, the user or the computer 1 can check a corresponding relationship as to whether the specified manufacturing condition data is associated with a corresponding manufacturing device and sensor in a corresponding manufacturing process of the manufacturing flow. For example, the user can output information of associating the manufacturing condition data specified by the computer 1 with a manufacturing flow configuration to the manufacturing system or a person in the manufacturing site. Alternatively, the computer 1 can transmit, to the manufacturing system, information of associating the specified manufacturing condition data with the manufacturing flow configuration, and set the information in each manufacturing device or the like.
(S1) First, in step S1, the control unit 30 acquires, as monitoring data of performance of the operation, the manufacturing condition data from each manufacturing process in the manufacturing flow and the quality data from the quality inspection process. The control unit 30 stores the acquired manufacturing condition data in the manufacturing condition data storage unit 41, and stores the acquired quality data in the quality data storage unit 42. The model building unit 31 refers to acquired manufacturing condition data D1 and quality data D2 from the storage unit 40.
The data (D1 and D2) is data acquired at each time point in time series (corresponding acquisition time point) including time points before and after a manufacturing state change. Time point T is used as a time point for explanation. Manufacturing condition data and quality data obtained at a certain time point T may be expressed as the manufacturing condition data at the time point T and the quality data at the time point T. A current time point may be expressed as a time point (t) and a next time point may be expressed as a time point (t+1).
(S2) Next, in step S2, the model building unit 31 builds the models 50 using the manufacturing condition data and the quality data that is obtained at a past time point and stored in the storage unit 40, in addition to using the manufacturing condition data and the quality data at the latest time point T obtained in step S1. The models 50 include one or more models. In the embodiment, the models 50 are a plurality of (N) models when there is a manufacturing state change. The number of models is set as N. In the embodiment, the causality model as shown in
The model building unit 31 builds each of the models 50 based on the manufacturing condition data D1 and the quality data D2 at each time point according to the manufacturing state change. The model building unit 31 builds a plurality of models (for example,
It should be noted that the models 50 are not limited to the causality model, and other types of models can be applied. In addition, models of a plurality of types may be mixed in the plurality of (N) models.
The model building unit 31 stores data of the built models 50 in the model storage unit 43. The model display unit 21 displays information of the built models 50 on a model building screen in
(S3) In step S3, the control unit 30 sets a quality target value based on a user operation. For example, the display unit 20 provides a setting screen related to quality. A quality target value setting field may be provided in a quality screen in
(S4) In step S4, the model manufacturing condition specifying unit 32 uses the models 50 built in S2 to specify, for each model when there are a plurality of models, a predicted value of manufacturing condition data at a next time point (t+1) with respect to a current time point (t). In
In the processing of S4, the model manufacturing condition specifying unit 32 specifies a manufacturing condition under which quality is good with respect to the quality target value D3 for each of the models 50. The model manufacturing condition specifying unit 32 stores the obtained manufacturing condition data D6 in the storage unit 40. The model manufacturing condition specifying unit 32 stores, in the first learning model storage unit 45, data of the used first learning model including updating. In the embodiment, the first learning model is, for example, a reinforcement learning model.
(S5) In step S5, the quality specifying unit 33 uses the models 50 in S2 to predict quality data D7 at the next time point (t+1) from manufacturing condition data D4 and quality data D5 at the current time point (t). Then, the quality specifying unit 33 compares the predicted quality data D7 at the next time point with the quality data D5 at the current time point (t), and calculates a quality error D8 which is an error between the quality data D7 and the quality data D6. The manufacturing condition data D4 and the quality data D5 at the current time point (t) can use data stored in the quality data storage unit 42. The quality specifying unit 33 stores the data (D7 and D8) calculated in S5 in the quality data storage unit 42.
(S6) In step S6, based on learning in a second learning model LM2, the manufacturing condition determining unit 34 determines a predicted value of manufacturing condition data D9 at the next time point (t+1) from the manufacturing condition data D4 at the latest current time point (t) in performance without using the models 50. The manufacturing condition data D9 is the second data. The manufacturing condition determining unit 34 stores the obtained manufacturing condition data D9 in the storage unit 40. The manufacturing condition determining unit 34 stores, in the second learning model storage unit 45, data of the second learning model LM2 including updating. In the embodiment, similar to the first learning model LM1, the second learning model LM2 is, for example, a reinforcement learning model.
In the embodiment, both the manufacturing condition data D6 which is the first data and the manufacturing condition data D9 which is the second data are used as input data of processing in S7.
(S7) In step S7, data including the manufacturing condition data D6 at the next time point which is the first data obtained in S4, the manufacturing condition data D9 at the next time point which is the second data obtained in S6, the quality error D8 obtained in S5, and the manufacturing condition data D4 at the current time point is input into the subspace specifying unit 35. The subspace specifying unit 35 executes subspace conversion processing in which the manufacturing condition data of the input data including the first data and the second data is converted into subspace data. The number of the manufacturing condition data D6 at the next time point which is the first data obtained in S4 is N corresponding to the number N of the models 50. The number of the manufacturing condition data D9 at the next time point which is the second data obtained in S6 is one. That is, the number of the manufacturing condition data of the input in S7 is (N+1). The subspace conversion processing in S7 is processing of projecting the (N+1) manufacturing condition data into a subspace (in other words, a low-dimensional space) which is a space different from the original space.
The conversion in S7 is a conversion to reduce the number of dimensions of the input data so as to match the number of dimensions of an input format in manufacturing condition specifying processing in S8. In other words, the number of dimensions of the input format of a third learning model LM3 in the manufacturing condition specifying processing in S8 matches the number of dimensions of the subspace data obtained in S7. The subspace data obtained in S7 is data projected to the subspace, and is data whose number of dimensions of parameter is reduced. When the number of dimensions of the (N+1) manufacturing condition data is set as DN1 and the number of dimensions of the subspace data is DN2, DN1>DN2. The number of dimensions DN2 matches the number of dimensions of the input of the third learning model LM3 in the manufacturing condition specifying processing in S8. Details of the processing in S7 will be described later.
The subspace specifying unit 35 obtains subspace manufacturing condition data D10 as the subspace data which is output data. The subspace specifying unit 35 stores the obtained subspace data in the subspace storage unit 46. The subspace display unit 23 displays the subspace data in a subspace data area 230 on a screen in
(S8) In step S8, the manufacturing condition specifying unit 36 uses the subspace manufacturing condition data D10 obtained in S7 to specify optimal manufacturing condition data D11 at the next time point (t+1) based on the learning in the third learning model LM3. The manufacturing condition data D11 is third data. In the processing, the manufacturing condition specifying unit 36 finally specifies optimal manufacturing condition data to be applied to the manufacturing flow from the (N+1) manufacturing condition data in the subspace data. In the processing, the manufacturing condition specifying unit 36 uses past manufacturing condition data and the quality error D8 to build the third learning model LM3. In the embodiment, a reinforcement learning model is used as the third learning model LM3. The manufacturing condition specifying unit 36 stores the specified manufacturing condition data D11 at the next time point (t+1) in the manufacturing condition data storage unit 41. The manufacturing condition specifying unit 36 stores, in the third learning model storage unit 47, data of the third learning model LM3 including updating.
(S9) In step S9, the display unit 20 displays, on a screen, various types of data or information such as the manufacturing condition data, the quality data, and the models, which are obtained as results of the above processing, so as to update screen display content. For example, the manufacturing condition specifying display unit 25 displays information of the manufacturing condition data D11 at the next time point (t+1) specified in S8 in a manufacturing condition data area 251 on a result screen in
The specified optimal manufacturing condition data D11 can be reflected in the manufacturing flow of the manufacturing system according to an operation of the user. For example, the manufacturing condition data D11 can be set to the manufacturing flow by pressing an OK button on the result screen in
(S10) First, in step S10, the subspace specifying unit 35 acquires the manufacturing condition data D4 at the current time point (t) in
(S11) In step S11, the subspace specifying unit 35 calculates a difference between the manufacturing condition data at the current time point (t) and the manufacturing condition data at the next time point (t+1) from the manufacturing condition data (D6 and D9) obtained in S10, and keeps the difference as a vector value (in other words, a difference vector). Details of S11 will be described in
(S12) In step S12, the subspace specifying unit 35 connects the vector values obtained in S11 to form a vertical vector as shown
(S13) In step S13, the subspace specifying unit 35 expresses the vertical vector obtained in S12 by a sum of vectors for items of the manufacturing condition data, as shown in
(S14) In step S14, the subspace specifying unit 35 calculates a coefficient value of vectors such that each component of a vector decomposed in the sum of the vectors obtained in S13 is 1. Details of S14 will be shown in
For the Model #1 (M1), manufacturing condition data t+i at the time point (t+l) is expressed by a formula 1 when the manufacturing condition data is expressed by a matrix of two rows and one column, in other words, by a column vector with the number of dimensions of 2. In the formula 1, a row vector is expressed by x1, t+1=(4, 4), a first data item value is 4, and a second data item value is 4. When being expressed similarly, the manufacturing condition data x1, t at the time point (t) is expressed by a formula 2. In the formula 2, a row vector is expressed by x1, t=(2, 3), a first data item value is 2, and a second data item value is 3. The subspace specifying unit 35 calculates a difference between the manufacturing condition data x1, t+1 in the formula 1 and the manufacturing condition data x1, t in the formula 2, and creates a difference vector in a formula 3. In the formula 3, a row vector is expressed by Δx1=x1, t+1−x1, t=(2, 1).
Similarly, for the Model #2 (M2), manufacturing condition data x2, t+1 at the time point (t+1) is expressed by a formula 4 when the manufacturing condition data is expressed by a matrix of two rows and one column, in other words, by a column vector with the number of dimensions of 2. In the formula 4, a row vector is expressed by x2, t+1=(5, 4). The manufacturing condition data x2, t at the time point (t) is expressed by a formula 5. In the formula 5, the row vector is expressed by x2, t=(3, 3). The subspace specifying unit 35 calculates a difference between the manufacturing condition data x2, t+1 in the formula 4 and the manufacturing condition data x2, t in the formula 5, and creates a difference vector in a formula 6. In the formula 6, a row vector is expressed by Δx2=x2, t+1−x2, t=(2, 1).
Under an actual manufacturing condition (i=0), manufacturing condition data x0, t+1 at the time point (t+1) is expressed by a formula 7 when the manufacturing condition data is expressed by a matrix of two rows and one column, in other words, by a column vector with the number of dimensions of 2. In the formula 7, a row vector is expressed by x0, t+1=(5, 4). The manufacturing condition data x0, t at the time point (t) is expressed by a formula 8. In the formula 8, a row vector is expressed by x0, t=(3, 3). The subspace specifying unit 35 calculates a difference between the manufacturing condition data x0, t+1 in the formula 7 and the manufacturing condition data x0, t in the formula 8, and creates a difference vector in a formula 9. In the formula 9, a row vector is expressed by Δx0=x0, t+1−x0, t=(2, 1).
The manufacturing condition specifying unit 36 in step S8 of
An example of a manufacturing state change will be described as follows. A worker may perform a maintenance operation on a manufacturing device or the like of each manufacturing process that forms the manufacturing flow of the manufacturing system. In this case, before and after a maintenance event (corresponding event time point), an internal or an external physical state of the manufacturing device or the like, for example, a value of a parameter such as a current may change instead of being a constant value. Depending on the manufacturing state change, an optimal manufacturing condition may be changed. When an operation is executed by continuing to apply the same manufacturing condition before and after the manufacturing state change, the quality of a product to be manufactured after the change may be deteriorated. That is, depending on the manufacturing state change, a suitable manufacturing condition may be changed internally, and it is necessary to specify a suitable manufacturing condition after the change. Therefore, corresponding to such a manufacturing state change, the manufacturing condition specifying system according to the embodiment has a function of building a model and a learning model and specifying a suitable or optimal manufacturing condition corresponding to a next time point after the change.
It should be noted that a unit of time point and time used in the manufacturing condition specifying system according to the embodiment is of a unit with a size corresponding to manufacturing time of a product in a target manufacturing system, for example, a day, an hour, a minute and the like, and can be set appropriately.
Parameters that form the manufacturing condition data correspond to a manufacturing flow of the target manufacturing system, can be set appropriately, and are not particularly limited. Examples of the parameters include currents, voltage, temperature, pressure, and the like. These parameters can be controlled, measured or the like. For example, the parameters include a current or voltage at a predetermined position in a manufacturing device, pressure in a predetermined space in a manufacturing device, temperature of an environment near the inside or outside of a manufacturing device, and the like. Examples of an applicable target manufacturing system and manufacturing flow include, but not limited to, at least a semiconductor manufacturing system and manufacturing flow.
The model building screen in
In the example of
When the user presses the setting button 211 on the model building screen in
After setting a condition or the like, the user presses an OK button. Accordingly, the control unit 30 reflects the condition or the like in the manufacturing condition specifying system. Then, according to the processing of the model building unit 31, the causality model can be built under the condition or the like. When the user presses a Cancel button, the condition or the like is not reflected, and the screen returns to a state before the setting screen is opened (the screen in
The model display unit 21 includes one or more model display areas 210 for displaying one or more causality models built by the model building unit 31, and displays the model display areas 210 in the model building screen in
The manufacturing condition screen in
The subspace display area 230 is an area where the subspace display unit 23 displays the subspace data obtained by the subspace specifying unit 35. A time difference (that is, a difference vector) of each model of the manufacturing condition data is displayed as the subspace data in the subspace display area 230. The subspace data related to the manufacturing condition data X1 is shown in the example. The subspace display area 230 is in, for example, a table format, and includes a model ID and a plurality of time periods as item columns. The model ID is an ID for each model, and corresponds to a label. The plurality of time periods are time periods between the time points T. For example, for the manufacturing condition data X1, and Model #1, a difference between data item values is 2 during a time period from a time point T1 to a time point T2, and a difference between data item values is 0 during a time period from the time point T2 to a time point T3.
The quality screen in
In the product yield display area 240, it is also possible to display an event time point 242 which is data representing a time point when an event is generated in a manufacturing process of a manufacturing flow. The event includes an event such as maintenance corresponding to a manufacturing state change. For example, an event time point Tx indicates a time point when a maintenance event occurs in a certain manufacturing device Dx.
As shown in the quality screen, the product quality varies over time. In the example, the yield tends to decrease at approaching time points before and after a certain event time 242 (Tx). Thereafter, a suitable manufacturing condition is specified by the manufacturing condition specifying system according to the embodiment, and when the suitable manufacturing condition is applied to the manufacturing flow, the yield tends to increase and is improved.
The result screen in
In the table of the manufacturing condition data as shown in
A lower side of
A table 1802 on a lower side of
As described above, the subspace specifying unit 35 processes data of the difference vectors shown in the table 1801 on the upper side, calculates the coefficient values shown in the table 1802 on the lower side, and stores the subspace data which is the calculated coefficient values in the subspace storage unit 46.
Learning methods that can be applied to each of the above-mentioned learning models are as follows.
First learning model LM1: state space model, reinforcement learning model
Second learning model LM2: state space model, reinforcement learning model
Third learning model LM3: reinforcement learning model
Known reinforcement learning is a kind of machine learning that addresses a problem in which a current state of an agent in certain environment is observed and then an action to be taken is determined. The reinforcement learning learns a strategy to obtain the largest reward through a series of actions. A known deep learning model or the like can be applied as a reinforcement learning model. A known state space model is a kind of a time-series analysis model. The state space model includes a state model and an observation model.
As described above, according to the manufacturing condition specifying system in the embodiment, even when there is a manufacturing state change, a suitable manufacturing condition can be specified, and the product quality can be maintained or improved. In particular, according to the embodiment, an optimal manufacturing condition corresponding to a quality target value can be specified according to the manufacturing state change. According to the embodiment, even when the number of operation data is small immediately after the manufacturing state change, a suitable manufacturing condition can be specified in a short period. In particular, according to the embodiment, a countermeasure can be easily made and a suitable manufacturing condition can be specified even when there is a large number of manufacturing state changes or a large number of parameters of a manufacturing condition and a model by using a subspace conversion. According to the embodiment or a modification, it is possible to improve the quality at an early stage after the manufacturing state change or improve a prediction accuracy of the model according to a priority policy.
As described above, the manufacturing condition specifying system according to the embodiment builds a plurality of models based on manufacturing condition data acquired at each state or time point of a past manufacturing state change. The system estimates a manufacturing condition at a next time point from each model of the plurality of models. The system calculates a difference for each time period between the time points from the plurality of estimated manufacturing conditions, and embeds the differences into a subspace as a base. The system uses subspace data which is the data embedded in the subspace as an input, and specifies an optimal manufacturing condition at a next time point based on learning.
Another embodiment can be implemented as follows. First, as a manufacturing condition specifying system according to a modification, the subspace conversion processing in step S7 in
In the manufacturing condition specifying system according to the modification, with regard to the subspace conversion processing in step S7 in
On the other hand, in the modification, the second data of the result of step S6 is not used during the processing of steps S7 and S8. The modification is effective in a case of a policy that places emphasis on increasing the prediction accuracy of the model more than improving, in an early stage, the product quality immediately after the manufacturing state change. In the modification, after the manufacturing state change, learning in a certain time period is advanced to increase the prediction accuracy of the model, and then optimal manufacturing condition data can be specified with high accuracy.
When a system of a related-art example is used as a comparative example with respect to the manufacturing condition specifying system according to the embodiment, the difference or the like is as follows. In PTL 1, a probability model is built from past manufacturing conditions in order to specify a manufacturing condition under which a product is good, and a manufacturing condition that matches a target value is calculated. However, when there is a manufacturing state change that cannot be expressed by the past manufacturing conditions, it is necessary to correct a built model in this technique.
In PTL 2, prediction is executed by outputting a plurality of predicted values from past performance data (for example, manufacturing conditions or quality) and weighting the output values according to similarity. However, a manufacturing condition cannot be specified when a prediction target (here, quality) is a good product in this technique.
PTL 3 explores a condition (for example, a manufacturing condition) that satisfies a certain condition (for example, a condition under which a product is good) in an environment (for example, a manufacturing process) by using reinforcement learning. However, when there are a plurality of models to be used, the number of items of the manufacturing condition to be specified is reduced by class classification in this technique. Therefore, it is regarded that the manufacturing condition under which a product is good may not be obtained in this technique.
On the other hand, the manufacturing condition specifying system according to the embodiment can build a plurality of models using a manufacturing condition and quality obtained from past manufacturing including a manufacturing state change, and can explore an optimal manufacturing condition using a manufacturing condition and a quality error obtained from each model.
Although the invention has been described in detail based on the embodiment, the invention is not limited to the embodiment described above, and various modifications can be made without departing from the scope of the invention.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2018-218491 | Nov 2018 | JP | national |
