TWI273361B - System and method for real-time quality detection - Google Patents

Abstract

A process control system capable of real-time-quality-state detection. The system comprises a model analysis unit, a detection unit, and a control unit. The model analysis unit defines model parameters according to historical data of a process. The detection unit determines whether larger offset occurs according to real-time process data and the model parameters. If so, the control unit obtains and compares an offset value with an offset threshold, obtaining a corresponding weighted value according to comparison results, and implements corresponding operations according to the weighted value.

Description

1273361 IX. Description of the Invention: [Technical Field] The present invention relates to a process control system and method, and more particularly to a process control system and method having an immediate quality state detection function. [Prior Art] Optoelectronic and semiconductor processes are quite complicated, and hundreds of different steps are required before and after. The instability of any process control may affect the next many processes, resulting in irreparable damage. The competition in the optoelectronic and semiconductor markets is fierce. How to improve product quality through process monitoring and product quality control to effectively improve product yield is an important key factor for determining the competitiveness of its products. However, in order to effectively improve the product yield, in addition to the pre-production design and post-production inspection and screening of each process, the stability of the process is also very important. In general, due to wear and tear of mechanical equipment, material consumption, regular maintenance, or replacement of a new process, the output of the process often deviates from the preset target value, thus causing variation in process quality, and thus Will reduce the yield of the product. Common quality variations can be divided into two broad categories: Shift and Drift. “Speed shift” refers to the phenomenon that the average value of the process output suddenly increases or decreases. For example, after the preventive maintenance (PM) of the machine, the performance of the machine is quite different from that before the maintenance. At this time, there will be a speed shift, which will cause the output value of the process to suddenly change to Another higher or lower value. "Drift" means that the average value of the process slowly increases or decreases with time, most of which occurs in plasma etching processes such as semiconductors, wall deposition on Furnace machines, or Abrasive pad wear of the Chemical Mechanical Polishing (CMP) process, etc., due to deposition, 0718-A20828TWF (N2); P10930030; alexchen 5 1273361 grinding or corrosion causes process shifts. = Statistic process control (additional (five) p (10) (four) CGntrQi, spc) After the abnormal immortalization of the immortal], there is an abnormal process offset, the force is determined to re-process the mouth to rework (Rework), if If you are unable to use the mouth = remedy, discard it. However, at this time, it has already caused considerable considerableity: In addition, although the abnormal process offset was detected, only a warning was issued to inform the engineer or the operator of the machine that there was a mutation, but the Ministry did not provide an effective solution, only by the past. Experience, the setting letter from the fine-tuning machine. 1 + and b, on the other hand, are more susceptible to the effects of natural variation, thus causing excessive control problems, which in turn increases process instability. Therefore, an effective method of retesting must be provided to quickly process the process to maintain process stability. [Description of the Invention] In view of the above, the object of the present invention is to provide a process control system and method with an immediate quality detection function, which is automatically provided to the controller to adjust the machine when an abnormality occurs in the t process. The set value quickly and effectively controls the output to the target value to maintain process stability. The present invention discloses a process control system with an instant quality status detection function, which includes a model analysis unit, a detection unit and a control unit. The model analysis unit establishes complex model parameters based on historical data of a process. The _ unit generates a large offset in the process according to the process data obtained on the line immediately and the sizing parameter established by the model analysis unit. If the _ unit lin has a large offset, then the control unit obtains the _offset from it, compares the offset with the _offset threshold, and obtains the corresponding first weight according to the comparison result. The value, and 0718-A20828TWF (N2); P1 〇 930030; alexchen 6 1273361 shift interference, can quickly and efficiently maintain the output value of the t-range to the target value. However, in order to ensure a relatively high process quality, in addition to the accuracy of the prediction mode, the weight selection of the EWMA-Based controller also has a considerable impact. Regarding the weight selection of the EWMA-Based controller, the following two important factors should be considered. The first is Sensitivity, which uses the weights to be sensitive to the amount of change in the process. Larger weights are more likely to detect larger offsets, while smaller weights are more likely to detect smaller offsets. The second is Robustness, which is to reduce the impact of noise and not to have too many false alarms. Next, the implementation flow of the process control system of the present invention will be described. FIG. 1 is a schematic diagram showing the architecture of a process control system with an instant quality status detection function according to an embodiment of the present invention. The process control system of the embodiment of the present invention includes an instant parameter computing module 1A, an offline model analysis module 200, a process unit 300, a measurement unit 400, and a database 500. The offline model analysis module 200 further includes a model analysis unit 250 that establishes model parameters (for example, a linear mode) according to historical data of a process or a machine stored in the database 500, and applies the established model parameters to Instant parameter computing module 1〇〇. The instant parameter computing module 100 further includes a detecting unit 110 and a control unit 130. The detecting unit 110 determines whether a large offset occurs in the process according to the process data acquired on the line (On-line) and the model parameters established by the previous model analyzing unit 250, and transmits the determination result to the control unit 13 Hey. The control unit 130 adjusts the process or the set value of the machine by using the EWMA-Based controller that can self-adjust the controller weight according to the judgment result, so as to maintain the process quality at the target value. The detection unit 110 and the control unit 130 will be described in more detail below. 0718-A20828TWF(N2); P10930030; alexchen 9 1273361 Referring to the 2nd, which is a flow chart showing the process control (4) of the function of (4) of the embodiment of the present invention. The heart (d) unit m judges whether the quality of the process is maintained at a steady state (Steady State) based on the model parameter 115 and using the r_StatlStieal estimation method (step S1). That is to say, the R_statistic is a benchmark based on the average value of the process and held at the value of S], and the formula is expressed as: 2, / , where v and " respectively are using the μα method The estimated process noise variation 'is expressed as: where, and 〇<λ^Ι y / = 1,2,3 〇 can be judged by the clever calculation to determine whether process disturbance (drift or speed shift) occurs. (End > Especially?) (Step S2). If yes, go to step S3, otherwise go to step S4. If the process is in steady state (ie, the process average is a fixed value or the added process disturbance is a process independent (Independent) signal and equal (Identically) perturbation), the value of Bay 0 will be approximately equal! . If the average value of the process is offset, or if the noise of the process is a self-correlation (Am〇 c〇rrelati〇n) signal, the value of the formula will be greater than 丨. When the average value of the process changes, at this time, 0718~A20828TWF(N2); P10930030; alexchen 10 1273361, the weight value is greater than the R-statistic control value, then η is incremented by 1, and the preset weight value is equal to the current The larger of the weight value (μΛ) of the batch operation time point and the minimum value of the EWMA-Based controller weight (= ) (step S 5 ). If the weight value of the previous batch operation time point is not greater than the R-statistic control value, the preset weight value is equal to the minimum value of the EWMA-Based controller weight (%=%) (step S6). After the weight value is set, the EWMA-Based controller provided in the control unit 130 adjusts according to the obtained value. Since the process control system of the embodiment of the present invention adjusts the offset amount, it is slowly adjusted in a smooth manner until the process disturbance is reduced to a reasonable range. In addition, when the self-adjustment mechanism is activated, the maximum value (V) of the EWMA-Based controller weights will have different corresponding values as the detection of different degrees of disturbance. In theory, the EWM-Based controller weight can be gradually reduced to the minimum according to the self-adjustment mechanism. If the appropriate EWMA-Based controller weight minimum (%) is set, the controller can have better sensitivity. Interference with a small amount of process. Referring to Figure 3, there is shown a schematic diagram of the adjustment of the controller power parameter of the embodiment of the present invention. In the case of waveform 310, when the process has a large interference, the self-adjustment mechanism is activated to adjust the result according to the controller weight parameter (as described in step S3), and in the case of waveform 330, it is When there is less interference in the process, the self-adjustment mechanism is activated to adjust the results as shown by the controller weight parameters (as described in steps S5 and S6). The ultimate goal is to adjust the controller weight parameter to be equal to the controller weight minimum (%). Next, how to adjust the EWMA-Based controller weight value 0 0718-A20828TWF(N2); P10930030; alexchen 12 1273361 if a process unit is used in the process control system with the instant quality state detection function of the present invention. It can be expressed as: γί=α + βχ^διΛ·ει , where $ is a special variation of the process unit, which can be expressed as:

τ Jr Bu, S, probability of occurrence P

The probability of occurrence is Y - P and, s ~ Ν, μ3, σ3 0 The target value of this process operation is set to 3500. Refer to Figure 4, which shows the results of the EWMA_Based controller process control for self-adjusting EWMA-Based controllers and fixed parameters. It can be clearly seen from Fig. 4 that the process control of the self-adjusting EWMA-Based controller can maintain the quality of the process at the set point faster and more stably than the fixed-parameter EWMA-Based controller, where < The maximum value of the EWMA-Based controller weight when large disturbance occurs, which is adjusted to be equal to the minimum value (%) of the EWMA-Based controller weight, and w; is the EWMA-Based when the previous small interference occurs. The maximum value of the controller weight, which is adjusted to be equal to the minimum (E) of the EWMA-Based controller weight. Figure 5 is a response graph of the self-adjusting EWMA-Based controller weights obtained from Figure 4. In summary, the process control system and method with the instant quality state detection function in the embodiment of the present invention utilizes the R-statistics in the detection unit to detect whether the manufacturing unit has a large offset interference. If so, moderately adjust the controller weights in the control unit to effectively handle the large offset process interference problem, and then gradually reduce the controller weight value. In this way, when there is large offset interference, 0718-A20828TWF{N2); P10930030; alexchen 13 1273361 can be compensated by using larger controller weights. When the process is not greatly disturbed, the control weight can be automatically The reduction is to avoid the interference of the process being improperly controlled by the controller. By the process control system and method of the embodiment of the invention, the manufacturing unit control system can balance the sensitivity and the robustness, thereby sensitively detecting the variation of the process and effectively reducing the influence of the noise, so as not to There are too many error warnings. While the present invention has been described in its preferred embodiments, the present invention is not intended to limit the invention, and the present invention may be modified and modified without departing from the spirit and scope of the invention. The scope of protection is subject to the definition of the scope of the patent application. 0718-A20828TWF(N2); P10930030; alexchen 14 1273361 [Simplified Schematic] FIG. 1 is a schematic diagram showing the architecture of a process control system with an instant quality state detection function according to an embodiment of the present invention. Figure 2 is a flow chart showing the steps of a process control method with an instant quality status detection function according to an embodiment of the present invention. Figure 3 is a schematic diagram showing the adjustment of the weight parameter of the EWMA-Based controller in the embodiment of the present invention. Fig. 4 is a view showing a comparison of process control results of the self-adjusting EWMA-Based controller of the embodiment of the present invention. Figure 5 is a response graph of the self-adjusting EWMA-Based controller weights obtained from Figure 4. [Main component symbol description] 100 to instant parameter calculation module 110 to detection unit 130 to control unit 200 to offline model analysis module 250 to model analysis unit 300 to process unit 400 to measurement unit 500 to database 0718-A20828TWF (N2); P10930030; alexchen 15

Claims (1)

Amendment date: 95.6.3, '..:i. Ϋ.I, 乂.r. ' 95. 6 . 0 5 1273 汹 "358 „Please modify the patent scope ten, the scope of patent application: - model state detection A functional process control system, including a type parameter; 建立 a complex analog-to-one detection unit based on historical data of a process, a number of processes of the above-mentioned model parameter determined by the green type analysis unit, and the above A large offset occurs in the module; and 1) whether there is a control unit in the above process - the above detection is obtained from the offset - offset #, and θ occurs, and the comparison result is obtained - the offset value of the opposite bank - offset For comparison, the younger brother of the ear 4 has a heavy value, and the heavy value performs the corresponding operation.处 柘 · · · · · · · · · · · · · · · 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如 如The defensive and the above control is based on the above-mentioned first _ weight $ value of the current batch operation time point according to the above offset. Corresponding to the upper port at the upper speed, please refer to the second paragraph of the patent scope for the real-time quality state: the system 'where, if the above-mentioned offset is greater than the above-mentioned offset threshold, the control is earlier than the above-mentioned The weight value is equal to - the first preset weight, and the value is sub- and the size of the offset is adjusted according to the first preset weight value. The invention has the real-time quality state detection function r: minus, wherein if the offset is less than the offset threshold, the control unit then determines that the operation time of the previous batch is obtained. Whether one of the second weight values is greater than a second predetermined weight value. 5. The method of claim 4, wherein if the preceding-to-batch operation time is greater than the second predetermined weight value, Then, the above control unit 0718~A20828TWF1(N2); P10930030; a 丨exchen 16 is. 1273361 The second preset weight value makes the first weight value equal to a larger value. ▲ 6· The patent control system with real-time quality state detection function described in the patent scope 帛 4, wherein the second weight value obtained in the previous batch operation time point is not If the value is greater than the second preset weight value, the control unit determines that the first weight value is equal to the second preset weight value. 7·—a process control method with real-time quality status detection function, the following steps are performed: · 复Based on historical data of a process to establish complex model parameters; ,, according to process data obtained on-line immediately and the above model parameters, judging Whether a large offset occurs in the above process; if the right detection unit determines that a large offset occurs, an offset is obtained. The offset is compared with an offset threshold; The comparison result obtains a corresponding first weight value; and performs a corresponding operation according to the first weight value. 8. If the process control method with real-time energy can be described in item 7 of the patent application scope, in the middle, the 偬 偬 偬 旦 — — 状 状 状 状 状 状 状 中 中 中 中 中 中The second operation time; the servo is taken, and the first weight value is set as the time point according to the above offset. The operation-control method according to item 8 of the claim patent scope, wherein if the offset is greater than the value: the first-weight value is equal to - The preset weight value, and the preset weight value adjusts the magnitude of the offset. x Kang Shang ίο. The process control method with instant energy as described in claim 8 of the patent application, wherein if the upper detection force value is used, the upper offset is -# - j is the upper return offset Critical material reading (4) is obtained at the operating time point of the front-μ. 718-A20828TWF1 (N2); P1〇93〇〇3〇; alexchen 17 1273361 Lu Xiu (more) is replacing the page, one of the second weight Whether the value is greater than a second preset weight value. The process control method as claimed in claim 10, wherein the second weight value obtained in the operation time point of the previous batch is greater than the second preset weight value The control unit causes the first weight value to be equal to a larger value of a third weight value and the second preset weight value. 12. The process control method as claimed in claim 1, wherein if the above-mentioned operating time point of the previous batch is not greater than the second value Presetting the weight value, the control unit causes the first weight value to be equal to the second preset weight value. 0718-A20828TWF1 (N2); P10930030; aiexchen 18 1273ML No. 41358 schema correction page Yongzheng Date: 95.6.3 + for .# rash (more) positive replacement page affavlto 1273361 4^s/^^