A method for detecting and controlling atomization abnormality of a polysilicon reduction furnace

Abstract

The present invention relates to a method for detecting and controlling atomization anomaly in a polysilicon reduction furnace, comprising the following steps: determining key influencing factors of atomization anomaly detection and control; batch reading historical furnace data from a DCS system database to form an experience database reserve; adapting anomaly detection rules; correcting anomaly detection rules; constructing detection control processing logic; and finally establishing a unified standardized atomization anomaly detection and processing method. The present invention can solve the problem of standardization of expert experience in traditional control methods, and at the same time solve the problem of universal applicability of anomaly detection and control processing methods in automated improvement methods.

Description

Technical Field

[0001] The invention relates to a method for detecting and controlling atomization anomalies of a polysilicon reduction furnace, and belongs to the field of polysilicon reduction furnace control, in particular to a method for detecting and controlling real-time anomalies in a polysilicon reduction furnace production process. Background Art

[0002] Rod-shaped polysilicon is the raw material for making single-crystal silicon, which is widely used in the electronics industry, such as integrated circuits, semiconductor devices, etc. In the photovoltaic industry, polysilicon is processed into cells and then assembled into photovoltaic modules to convert solar energy into electrical energy. Therefore, polysilicon has become a material with important application value in related fields.

[0003] In the production process of polysilicon reduction furnace, when the balance between the concentration of trichlorosilane in the furnace atmosphere and the surface temperature of silicon rod is destroyed, the silicon generated by the reaction cannot be effectively adsorbed to the surface of silicon rod, but floats in the furnace atmosphere. After a period of accumulation, atomization will occur. The so-called atomization phenomenon refers to the inability of atomized silicon particles to be deposited on the surface of silicon rod in an orderly manner, resulting in uneven crystal structure of silicon rod, which may produce defects and impurities, thereby reducing the quality and electrical properties of silicon rod. For example, silicon rod may have local resistivity anomalies, affecting its application in semiconductor device manufacturing; atomized silicon particles will occupy the space in the furnace, reduce the effective reaction area, reduce the concentration and reaction rate of reactants, and extend the growth cycle of silicon rod; and due to the imbalance of reaction and the occurrence of atomization, in order to achieve the expected production effect, it may be necessary to increase the input energy, such as increasing the heating power, thereby increasing energy consumption and production costs. At present, regarding the production process of polysilicon, taking a 40-pair rod reduction furnace as an example, the single furnace production cycle of improved Siemens solar-grade rod-shaped polysilicon is generally about 100h. In the production process, a pre-planned production formula is usually used as the operation control baseline within 0-100 hours. The production formula mainly includes three parts, namely the hydrogen flow set value, the trichlorosilane flow set value and the current set value. The operation control baseline can also be understood as the ideal optimal operation control path compiled by the reduction system process engineer. The core content is the curve of the above set value changing over time. From the 0th hour of feeding to the 100th hour of feeding, if the state in the furnace is abnormal during this period, it is necessary to give a certain compensation based on the established set value for abnormal handling. When the state in the furnace returns to normal, the set value will return to the control baseline as much as possible.

[0004] Based on the above characteristics of the polysilicon production process, at present, each production unit has two methods for abnormal detection of furnace status (mainly for atomization phenomenon) and control and processing. One is more traditional and more common, that is, through high-frequency on-site inspections and rotations of multiple shifts of personnel, to obtain the furnace status of the reduction furnace, evaluate the surface temperature, diameter and density of the silicon rod, and then feedback to the operator in the central control room, and compensate for the formula setting value given on the DCS based on historical experience to complete the abnormal state processing. The second is an automated improvement based on the previous method, such as a polysilicon reduction furnace fully automatic intelligent control system and control method provided by the application with publication number CN116257004A, the control system includes a control device that cooperates with the reduction furnace, the control device includes a control panel and a voltage detection module electrically connected to the control panel, an exhaust gas temperature detection module, a logic processing module and an execution module; the control method includes dividing the reduction furnace automatic control process into several stages according to the reaction time, and the control strategy of each stage can be set or adjusted separately; for each stage, voltage automatic control, exhaust gas temperature automatic control and exhaust gas temperature maximum limit control are implemented. The second method is to give the limit value of the change on the data collected by the DCS system based on the observation experience of personnel, and then formulate the corresponding formula setting value compensation logic based on the control experience of personnel, and then the program automatically completes some simple exception handling work, thereby reducing the inspection and operation pressure of personnel; however, this control method is only used in a few leading enterprises and has not yet been widely promoted.

[0005] As for the first method, it is very effective in training personnel, accumulating experience and improving operation management level, but its disadvantages are also obvious. The level of automation control is low, and it is highly dependent on experienced personnel. Different operators in different teams and shifts have different abnormal judgment and control processing experience. There is a lack of unified standards, and it is difficult to achieve consistent product quality control. Although the second method has promoted the level of automation control in the field of polysilicon to a certain extent, it is relatively simple in data processing methods, detection standards and even control program content. For the complex chemical reactions in the reduction furnace, it still has disadvantages such as poor adaptability, insufficient intelligent parameter adjustment and long debugging cycle, which can easily lead to missed judgments, over-judgments, failure to operate in time and frequent misoperations during operation.

[0006] To sum up, whether it is possible to adapt to the production and operation characteristics of different reduction furnaces represented by historical data, form a set of standard abnormality detection methods, and then combine them with dynamically changing real-time data to establish a set of standard control processing procedures, is of great practical significance. Summary of the invention

[0007] The present invention provides a method for detecting and controlling the atomization anomaly of a polysilicon reduction furnace, which can solve the standardization problem of expert experience in traditional control methods and solve the universality problem of the anomaly detection and control processing method in the automated improvement method.

[0008] The technical solution adopted by the present invention to solve its technical problem is:

[0009] A method for detecting and controlling atomization abnormality of a polysilicon reduction furnace, specifically comprising the following steps:

[0010] Step S1: determining the key influencing factors of atomization abnormality detection and control, that is, detecting and controlling the chassis outlet exhaust temperature and its change trend;

[0011] Step S2: batch read historical furnace data from the OPC database of the distributed control system, including reaction time, recipe initial value, recipe final setting value and chassis outlet exhaust gas temperature, locate the time point or time interval when the abnormality occurs by using the difference between the recipe initial value and the recipe final setting value, and then mark the abnormal point on the chassis outlet exhaust gas temperature data according to the located time point or time interval, so as to form a chassis outlet exhaust gas temperature abnormal point database as an experience database reserve;

[0012] The formula includes hydrogen flow, trichlorosilane flow, A1 phase current, C1 phase current, A2 phase current, C2 phase current, B1 phase current and B2 phase current, the initial value of the formula is the original formula entered into the distributed control system by the engineer and not adjusted during operation, and the final setting value of the formula is the initial value of the formula plus the part adjusted by manual compensation during operation;

[0013] Step S3: analyzing and processing the exhaust gas temperature data at the chassis outlet of the polysilicon reduction furnace collected on site, eliminating the fluctuation interference of the data, and statistically analyzing its numerical distribution;

[0014] Step S4: Based on the data acquired in step S3, an absolute abnormal data range, an abnormal data range and a normal data range of the chassis outlet exhaust temperature change are set by an abnormality detection method, and the data are screened;

[0015] Step S5: compare the abnormal data screened out in step S4 with the content in the experience database reserve in step S2. When the matching degree between the two is ≥90%, confirm that the abnormal detection method used in step S4 is effective. If the matching degree between the two is <90%, calibrate the abnormal detection method until the matching degree between the two is ≥90%.

[0016] Step S6: by correcting the abnormality detection method in step S5, the time interval of silicon rod growth, the absolute abnormal interval, adjustment interval and fine-tuning interval of the exhaust temperature at the chassis outlet during production operation are finally confirmed, and the callback strength is determined through the callback boundary, and the compensation control limit is confirmed, so as to build the detection control processing logic;

[0017] Step S7: establishing a unified and standardized atomization abnormality detection and processing method based on the detection control processing logic constructed in step S6;

[0018] As a further preferred embodiment of the present invention, the principle based on which the key influencing factors of atomization abnormality detection and control are determined is the radiation heat transfer principle, and its formula is: In the formula, Q represents the heat transfer from the silicon rod to the inner wall of the furnace, k is the radiation heat transfer coefficient, A represents the emission area of ​​the silicon rod to the inner wall of the furnace, T1 represents the emission temperature of the silicon rod to the inner wall of the furnace, and T2 represents the receiving temperature of the inner wall of the furnace to the silicon rod;

[0019] When the emission area A of the silicon rod to the inner wall of the furnace barrel decreases, the heat transfer from the silicon rod to the inner wall of the furnace barrel decreases. Since the total heat generation at the abnormal moment remains unchanged, the heat discharged from the exhaust hole increases, resulting in a change in the exhaust temperature at the chassis outlet.

[0020] As a further preferred embodiment of the present invention, in step S2, when batch reading historical furnace data from the OPC database of the distributed control system, the data granularity is 1 record / min;

[0021] As a further preferred embodiment of the present invention, in step S3, the exhaust gas temperature data at the chassis outlet of the polysilicon reduction furnace collected on site is averaged to obtain short-time average data and long-time average data, wherein the short-time interval is 3-10 minutes and the long-time interval is 60-120 minutes;

[0022] As a further preferred embodiment of the present invention, the short time is set to 5 minutes and the long time is set to 120 minutes;

[0023] As a further preferred embodiment of the present invention, in step S4, the obtained short-time mean data and long-time mean data are subjected to rolling difference calculation at fixed time intervals, specifically, the short-time mean data is subtracted from the long-time mean data, or the short-time mean data is subtracted from the short-time mean data, and then the distribution of the mean difference data is checked using a histogram;

[0024] Based on the calculated mean difference data of chassis outlet exhaust temperature at different time points combined with the histogram, the anomaly detection method is determined to be the 3σ principle;

[0025] The calculation method of standard deviation σ is: In the formula, n represents the number of chassis outlet exhaust temperature mean difference data samples. represents the average of all mean difference data, x i Represents the value of the i-th mean difference data;

[0026] As a further preferred embodiment of the present invention, in step S4, according to the 3σ principle, the absolute abnormal data range of the mean difference data of the exhaust gas temperature at the chassis outlet is set to [-∞, -3σ] and [3σ, ∞], the abnormal data range is set to [-3σ, -2σ] and [2σ, 3σ], and the normal data range is set to [-σ, σ];

[0027] Screen the mean difference data of chassis outlet exhaust gas temperature with values ​​exceeding 2σ;

[0028] As a further preferred embodiment of the present invention, in step S5, the abnormality detection method is calibrated, that is, the mean difference of the exhaust gas temperature at the chassis outlet is screened until the matching degree between the two is ≥90%;

[0029] Among them, a0 is the empirical compensation for threshold correction in the fine-tuning interval, that is, in the fine-tuning interval, the mean difference of the chassis outlet exhaust gas temperature with a value exceeding σ+a0 is screened; a1 is the empirical compensation for threshold correction in the adjustment interval, that is, in the adjustment interval, the mean difference of the chassis outlet exhaust gas temperature with a value exceeding 2σ+a1 is screened; a2 is the empirical compensation for threshold correction in the absolute abnormal interval, that is, in the absolute abnormal interval, the mean difference of the chassis outlet exhaust gas temperature with a value exceeding 3σ+a2 is screened;

[0030] As a further preferred embodiment of the present invention, in step S6, based on the fact that the production stages of the silicon rods include an accelerated growth stage, a rapid growth stage, and a stable growth stage, the time interval for silicon rod growth is divided into an accelerated growth time interval [0, t2], a rapid growth time interval [t2, t3], and a stable growth time interval [t3, 100];

[0031] The time interval of accelerated growth [0, t2] is further divided into the time interval [0, t1] not involved in regulation and the time interval of accelerated growth involved in regulation [t1, t2];

[0032] The mean difference of the exhaust gas temperature at the chassis outlet was screened, and it was confirmed that the absolute abnormal interval was [3σ+a2,∞], the adjustment interval was 2σ+a1,3σ+a2], and the fine-tuning interval was [σ+a0,2σ+a1];

[0033] Set the callback boundary σ and use it to determine the callback strength. The callback strength includes three levels: [0,σ], [-3σ,0] and [-∞,-3σ];

[0034] The compensation control limit is included in the accelerated growth time interval [t1, t2], The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-1, is the maximum limit of hydrogen compensation in the accelerated growth time interval [t1, t2], ranging from (0, +∞), L TCS-1 is the maximum limit of trichlorosilane compensation in the accelerated growth time interval [t1, t2], and its value range is (-∞, 0);

[0035] In the fast growth time interval [t2, t3], The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-2 , is the maximum limit of hydrogen compensation in the rapid growth time interval [t2, t3], ranging from (0, +∞), L TCS-2 is the maximum limit of trichlorosilane compensation in the rapid growth time interval [t2, t3], and its value range is (-∞, 0);

[0036] In the stable growth time interval [t3,100], the cumulative adjustment of hydrogen flow ≤ The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-3 , is the maximum limit of hydrogen compensation in the stable growth time interval [t3,100], and its value range is (0,+∞); L TCS-3 is the maximum limit of trichlorosilane compensation in the stable growth time interval [t3,100], and its value range is (-∞,0);

[0037] As a further preferred embodiment of the present invention, in step S7, establishing a unified standardized atomization anomaly detection and processing method includes the following steps:

[0038] Step S71, real-time calculation of the mean difference ΔT of the chassis outlet exhaust temperature at time t for 5 minutes t , mean difference ΔT t The rate of change k t , the cumulative adjustment amount x of hydrogen flow and the cumulative adjustment amount y of trichlorosilane flow; where ΔT t =average(T t , T t-1 , T t-2 , T t-3 , T t-4 )-average(T t-5 , T t-6 , T t-7 , T t-8 , T t-9 ), k t =ΔT t -ΔT t-1 ;

[0039] Step S72, for the accelerated growth time interval [t1, t2], the rapid growth time interval [t2, t3] and the stable growth time interval [t3, 100], four modules of hydrogen bias compensation, trichlorosilane feed bias compensation, hydrogen callback, and trichlorosilane feed callback are set in each time interval; the data acquisition frequency of the four modules is 1 acquisition / min, the sending frequency of the hydrogen bias compensation and trichlorosilane feed bias compensation modules is 1 time / min, and the sending frequency of the hydrogen callback and trichlorosilane feed callback modules is 1 time / 5min;

[0040] Step S73, determining whether the current operation time of the polysilicon reduction furnace is in the accelerated growth time interval [t1, t2] or the rapid growth time interval [t2, t3] or the stable growth time interval [t3, 100], if it is in the accelerated growth time interval [t1, t2], proceeding to step S74, if it is in the rapid growth time interval [t2, t3], proceeding to step S75, if it is in the stable growth time interval [t3, 100], proceeding to step S76;

[0041] Step S74: when it is determined that the current time is within the interval [t1, t2],

[0042] The hydrogen bias compensation scheme is: when 2σ+a1≤ΔT t ≤3σ+a2, if k t >0, then increase hydrogen bias compensation 10×(ΔT t -2σ)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ)kg / h; when ΔT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 15kg / h, if k t <0, then increase the hydrogen bias compensation by 10kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts,

[0043] The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-1 The above compensation scheme stops running until t≥2 / 3×L TCS-1 When the system restarts,

[0044] The callback scheme for hydrogen is: when 0≤ΔT t ≤σ, and x>0, then the hydrogen (ΔT t -2σ)kg / h, when -3σ≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when ΔT t When <-3σ and x>0, the hydrogen will be adjusted back to -10kg / h;

[0045] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h;

[0046] Step S75: When it is determined that the current time is within the interval [t2, t3],

[0047] The hydrogen bias compensation scheme is: when 2σ≤ΔT t ≤3σ+a2, if k t >0, then increase hydrogen bias compensation 10×(ΔT t -2σ)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ)kg / h; when ΔT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 15kg / h, if k t <0, then increase the hydrogen bias compensation by 10kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts,

[0048] The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-2 The above compensation scheme stops running until t≥2 / 3×L TCS-2 When the system restarts,

[0049] The callback scheme for hydrogen is: when 0≤ΔT t≤σ, and x>0, then the hydrogen (ΔT t -2σ)kg / h, when -3σ≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when ΔT t When <-3σ and x>0, the hydrogen will be adjusted back to -10kg / h;

[0050] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h;

[0051] Step S76: When it is determined that the current time is within the interval [t3, 100],

[0052] The hydrogen bias compensation scheme is: when 2σ-0.3≤ΔT t ≤3σ+a2-0.3, if k t > 0, then increase hydrogen bias compensation 6×(ΔT t -2σ+0.3)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ+0.3)kg / h; when ΔT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 10kg / h, if k t <0, then increase the hydrogen bias compensation by 6kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts,

[0053] The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-3 The above compensation scheme stops running until y≥2 / 3×L TCS-3 When the system restarts,

[0054] The callback scheme for hydrogen is: when 0≤ΔT t ≤σ, and x>0, then the hydrogen (ΔTt -2σ)kg / h, when -3σ≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when ΔT t When <-3σ and x>0, the hydrogen will be adjusted back to -10kg / h;

[0055] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h.

[0056] Through the above technical solution, compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. The atomization abnormality detection and control processing method of the polysilicon reduction furnace provided by the present invention comprehensively considers the characteristics of different polysilicon reduction furnaces and the different system working conditions in which they are located, forms a set of standard atomization abnormality detection and control processing procedures, improves the accuracy and universality of atomization detection and identification, and the effectiveness of atomization control processing;

[0058] 2. The atomization anomaly detection and control processing method of the polysilicon reduction furnace provided by the present invention realizes that the frequency of manual intervention in the production and operation of the reduction furnace is ≤1 time / 10 furnaces, which basically solves the problems of excessive reliance on personnel experience, poor adaptability of automatic control programs, insufficient intelligence of parameter adjustment, and long debugging cycle, and greatly reduces the labor intensity of personnel observation, inspection, and regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0060] Figure 1 It is a flow chart of the atomization abnormality detection and control processing method of the polysilicon reduction furnace provided by the present invention;

[0061] Figure 2 This is a preferred embodiment control parameter table 1 provided by the present invention;

[0062] Figure 3 This is a preferred embodiment control parameter table 2 provided by the present invention;

[0063] Figure 4 This is the preferred embodiment control parameter table 3 provided by the present invention. DETAILED DESCRIPTION

[0064] The present invention will now be described in further detail with reference to the accompanying drawings.

[0065] As described in the background technology, based on the characteristics of the existing polysilicon production process, if the process of polysilicon automated control is to be promoted, it is necessary to overcome the relatively simple data processing methods, detection standards and control procedures, which lead to poor adaptability, insufficient intelligent parameter adjustment and long debugging cycle when facing complex chemical reactions in the reduction furnace. In order to solve the problems of missed judgment, over-judgment, failure to operate in time and frequent misoperation during the operation of the above process, the present application provides a method for detecting and controlling the atomization anomaly of a polysilicon reduction furnace, which aims to solve the standardization problem of expert experience in traditional control methods on the one hand, and the universality problem of anomaly detection and control processing methods in automated improvement methods on the other hand.

[0066] The entire detection and control processing method is as follows Figure 1 As shown, the following steps are included:

[0067] Step S1: determining the key influencing factors of atomization abnormality detection and control, that is, detecting and controlling the chassis outlet exhaust temperature and its change trend;

[0068] Before constructing the method, it is necessary to determine the key influencing factors of atomization abnormality detection and control. The principle based on this is the radiation heat transfer principle, and its formula is: In the formula, Q represents the heat transfer from the silicon rod to the inner wall of the furnace, k is the radiation heat transfer coefficient, A represents the emission area of ​​the silicon rod to the inner wall of the furnace, T1 represents the emission temperature of the silicon rod to the inner wall of the furnace, and T2 represents the receiving temperature of the inner wall of the furnace to the silicon rod.

[0069] When atomization anomaly occurs, a large number of fine silicon particles suspended in the furnace atmosphere will reduce the emission area A of the silicon rod to the inner wall of the furnace. At this time, k, T1, and T2 have not changed yet, so the radiation heat transfer will drop significantly with the reduction of the emission area A. However, since the total heat generation at the abnormal moment is relatively unchanged, the heat that cannot be effectively radiated to the inner wall of the furnace will accumulate in the furnace atmosphere and then be discharged as exhaust gas. When it is discharged from the exhaust hole, it will inevitably cause the exhaust gas temperature to rise rapidly. Therefore, we can believe that the key to atomization anomaly detection and control lies in the reasonable and effective detection and control of the chassis outlet exhaust temperature and its changing trend.

[0070] After determining the key to detection and control, proceed to step S2, batch read historical furnace data from the OPC database of the distributed control system (DCS system), the data granularity is 1 item / min, mainly including reaction time, recipe initial value, recipe final setting value and chassis outlet exhaust gas temperature, use the difference between the recipe initial value and the recipe final setting value to locate the time point or time interval when the abnormality occurs, and then mark the abnormal point on the chassis outlet exhaust gas temperature data according to the located time point or time interval, forming a chassis outlet exhaust gas temperature abnormal point database as an experience database reserve;

[0071] Among them, the formula includes hydrogen flow, trichlorosilane flow, A1 phase current, C1 phase current, A2 phase current, C2 phase current, B1 phase current and B2 phase current. The initial value of the formula is the original formula entered into the distributed control system by the engineer and has not been adjusted during operation. The final setting value of the formula is the initial value of the formula during operation plus the part adjusted by manual compensation (that is, the value finally given to the DCS system for issuance).

[0072] The basis of this treatment method is that when the state inside the furnace is abnormal, it will inevitably be accompanied by manual adjustment, and the traces of manual adjustment can be accurately obtained from the difference between the initial value of the recipe and the final set value of the recipe, and all data are related to the reaction time. In this way, the data phenomenon that caused the abnormality can be reversely located through the results of historical data, forming a relatively more valuable experience database, which effectively paves the way for subsequent data analysis work.

[0073] The next step is to adapt the anomaly detection rules, which mainly includes three parts, including data processing, establishing reasonable data statistical rules, and screening out data anomalies according to the standardized anomaly detection method formed in the first two links, which corresponds to step S3-step S4. First, step S3, in the data processing link, considering that the chassis outlet exhaust temperature data collected on site have different degrees of fluctuations, if the collected point data is directly used for anomaly detection, its fluctuation will have a great interference on the final detection effect, so the collected data must be averaged to eliminate the interference of data fluctuations. In the mean processing method, we only consider short-term mean data and long-term mean data. This is because the advantage of short-term mean processing is that it can more sensitively reflect the current trend of exhaust temperature changes, but it is difficult to retain the changes in the past period of time. The advantages and disadvantages of the long-term mean are just the opposite of the short-term mean. It is relatively lagging in reflecting the current trend of exhaust temperature changes, but it can retain the changes in the past period of time. There are multiple options for the short-time mean within the short-time interval of 3-10 minutes, and the average of data within 5 minutes is generally taken. There are also multiple options for the long-time mean within the long-time interval of 60-120 minutes, and the average of data within 120 minutes is generally taken.

[0074] The processed short-term average data and long-term average data of exhaust gas temperature enter the next data statistics link together.

[0075] In step S4, a data statistical rule establishment link is performed based on the data obtained in step S3, and the obtained short-time mean data and long-time mean data are further processed, that is, rolling differences are calculated at fixed time intervals. Generally, the short-time mean data is subtracted from the long-time mean data, or the short-time mean data is subtracted from the short-time mean data, and then the distribution of the mean difference data is checked using a histogram.

[0076] Since each abnormally rapid rise in the exhaust gas temperature at the chassis outlet is a typical data manifestation of abnormal atmosphere conditions in the furnace, the temperature change rate has become a key indicator for personnel observation, regulation, and even automated program control. However, due to the lack of a unified standard for empirical understanding of the temperature change rate and the lack of massive historical data verification, the solutions made are difficult to be universal and are not conducive to improving the overall automation level of the industry.

[0077] The essence of calculating the mean difference data of exhaust gas temperature at different time points is to reflect the rate of change of the exhaust gas temperature at the chassis outlet, and the result of the histogram inspection shows that the mean difference data basically obeys the normal distribution. Therefore, in step S4, the abnormal detection method is determined to be the 3σ principle of statistics, that is, the abnormal data range and the normal data range are screened by the 3σ principle. The 3σ principle means that in the case of normal distribution, about 68% of the data will fall within the range of one standard deviation near the mean, about 95% of the data will fall within the range of 2 standard deviations near the mean, and about 99.7% of the data will fall within the range of 3 standard deviations near the mean. Using the 3σ principle as a guiding ideology can basically determine the distribution of data and the frequency of occurrence of abnormal values, but appropriate adjustments need to be made according to actual conditions during use.

[0078] Here the standard deviation σ is calculated as In the formula, n represents the number of chassis outlet exhaust temperature mean difference data samples. represents the average of all mean difference data, x i Represents the value of the i-th mean difference data. According to the 3σ principle, the absolute abnormal data range of the mean difference data of the chassis outlet exhaust temperature is set to [-∞,-3σ] and [3σ,∞], the abnormal data range is [-3σ,-2σ] and [2σ,3σ], and the normal data range is [-σ,σ].

[0079] In the abnormal data screening stage, we first use the above rules to screen the mean difference data of the chassis outlet exhaust temperature with a value exceeding 2σ, because only when the mean difference data of the chassis outlet exhaust temperature is positive, it indicates that the atmosphere in the furnace has abnormally increased, increasing the probability of atomization abnormality. Then, using the reaction time as a medium, we find the abnormal point corresponding to the chassis outlet exhaust temperature, and then enter the correction stage. That is, step S5 abnormal detection rule correction, compare the abnormal data screened out in step S4 with the content in the experience database reserve in step S2, and when the matching degree between the two is ≥90%, confirm that the abnormal detection method used in step S4 is effective. If the matching degree between the two is <90%, the abnormal detection method is corrected, that is, the mean difference of the chassis outlet exhaust temperature is screened until the matching degree between the two is ≥90%.

[0080] During correction, a0 is the empirical compensation for the threshold correction of the fine-tuning interval, that is, in the fine-tuning interval, the mean difference of the chassis outlet exhaust temperature with a value exceeding σ+a0 is screened; a1 is the empirical compensation for the threshold correction of the adjustment interval, that is, in the adjustment interval, the mean difference of the chassis outlet exhaust temperature with a value exceeding 2σ+a1 is screened; a2 is the empirical compensation for the threshold correction of the absolute abnormal interval, that is, in the absolute abnormal interval, the mean difference of the chassis outlet exhaust temperature with a value exceeding 3σ+a2 is screened.

[0081] Step S6: by correcting the abnormality detection method in step S5, the time interval of silicon rod growth, the absolute abnormal interval, adjustment interval and fine-tuning interval of the exhaust temperature at the chassis outlet during production operation are finally confirmed, and the callback strength is determined through the callback boundary, and the compensation control limit is confirmed, so as to build the detection control processing logic;

[0082] Specifically, the production of solar-grade rod-shaped polysilicon produced by the modified Siemens process is generally considered to be divided into three stages, namely the accelerated growth stage, the rapid growth stage and the stable growth stage. The rapid growth stage is the stage where atomization abnormalities are more likely to occur. Correspondingly, we can divide the growth of silicon rods into three time intervals, namely the accelerated growth time interval [0, t2], the rapid growth time interval [t2, t3] and the stable growth time interval [t3, 100]. Considering that too much intervention at the beginning of feeding will only lead to drastic changes in the airflow in the furnace and increase the risk of the silicon core leaning against the wall, we further divide the accelerated growth stage into a time interval [0, t1] that does not participate in regulation and a time interval [t1, t2] that participates in regulation.

[0083] Taking into account that the transition of the furnace atmosphere from a normal state to an abnormal state during production operation is a process of quantitative accumulation, we introduced the concepts of fine-tuning interval, adjustment interval and absolute abnormal interval in the division of the control interval, and further gave the range of the above intervals around the chassis outlet exhaust temperature mean difference change thresholds σ+a0, 2σ+a1 and 3σ+a2 formed in step S5, confirming that the absolute abnormal interval is [3σ+a2,∞], the adjustment interval is [2σ+a1,3σ+a2], and the fine-tuning interval is [σ+a0,2σ+a1].

[0084] When we deal with an abnormality, the added compensation value will cause the final set value of the recipe to deviate from the initial value of the recipe. The initial value of the recipe, as a preset operating baseline, is a relatively ideal operating path. Therefore, it is necessary to make an orderly callback based on the feedback of the abnormal handling. It is generally believed that when the mean difference of the exhaust gas temperature at the chassis outlet falls within the normal range [-σ,σ], the atmosphere state in the furnace returns to normal. Therefore, σ can be used as the boundary of the callback. The callback intensity can be divided into three levels according to [0,σ], [-3σ,0] and [-∞,-3σ].

[0085] There are three main provisions for compensation control limits: within the accelerated growth time interval [t1, t2], The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-1 , is the maximum limit of hydrogen compensation in the accelerated growth time interval [t1, t2], ranging from (0, +∞), L TCS-1 is the maximum limit of trichlorosilane compensation in the accelerated growth time interval [t1, t2], and its value range is (-∞, 0);

[0086] In the fast growth time interval [t2, t3], The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-2 , is the maximum limit of hydrogen compensation in the rapid growth time interval [t2, t3], ranging from (0, +∞), L TCS-2 is the maximum limit of trichlorosilane compensation in the rapid growth time interval [t2, t3], and its value range is (-∞, 0);

[0087] In the stable growth time interval [t3,100], the cumulative adjustment of hydrogen flow ≤ The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-3 , is the maximum limit of hydrogen compensation in the stable growth time interval [t3,100], and its value range is (0,+∞); L TCS-3It is the maximum limit of trichlorosilane compensation in the stable growth time interval [t3,100], and its value range is (-∞,0).

[0088] Finally, the most important step of this application is step S7, which is to establish a unified and standardized atomization abnormality detection and processing method based on the detection control processing logic constructed in step S6;

[0089] The following steps are involved:

[0090] Step S71, real-time calculation of the mean difference ΔT of the chassis outlet exhaust temperature at time t for 5 minutes t , mean difference ΔT t The rate of change k t , the cumulative adjustment amount x of hydrogen flow and the cumulative adjustment amount y of trichlorosilane flow; where ΔT t =average(T t , T t-1 , T t-2 , T t-3 , T t-4 )-average(T t-5 , T t-6 , T t-7 , T t-8 , T t-9 ), k t =ΔT t -ΔT t-1 ;

[0091] Step S72, for the accelerated growth time interval [t1, t2], the rapid growth time interval [t2, t3] and the stable growth time interval [t3, 100], four modules of hydrogen bias compensation, trichlorosilane feed bias compensation, hydrogen callback, and trichlorosilane feed callback are set in each time interval; the data acquisition frequency of the four modules is 1 acquisition / min, the sending frequency of the hydrogen bias compensation and trichlorosilane feed bias compensation modules is 1 time / min, and the sending frequency of the hydrogen callback and trichlorosilane feed callback modules is 1 time / 5min;

[0092] Step S73, determining whether the current operation time of the polysilicon reduction furnace is in the accelerated growth time interval [t1, t2] or the rapid growth time interval [t2, t3] or the stable growth time interval [t3, 100], if it is in the accelerated growth time interval [t1, t2], proceeding to step S74, if it is in the rapid growth time interval [t2, t3], proceeding to step S75, if it is in the stable growth time interval [t3, 100], proceeding to step S76;

[0093] Step S74: when it is determined that the current time is within the interval [t1, t2],

[0094] The hydrogen bias compensation scheme is: when 2σ+a1≤ΔT t ≤3σ+a2, if k t >0, then increase hydrogen bias compensation 10×(ΔT t -2σ)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ)kg / h; when ΔT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 15kg / h, if k t <0, then increase the hydrogen bias compensation by 10kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts,

[0095] The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-1 The above compensation scheme stops running until y≥2 / 3×L TCS-1 When the system restarts,

[0096] The callback scheme for hydrogen is: when 0≤ΔT t ≤σ, and x>0, then the hydrogen (ΔT t -2σ)kg / h, when -3σ≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when ΔT t When <-3σ and x>0, the hydrogen will be adjusted back to -10kg / h;

[0097] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h;

[0098] Step S75: When it is determined that the current time is within the interval [t2, t3],

[0099] The hydrogen bias compensation scheme is: when 2σ≤ΔTt ≤3σ+a2, if k t >0, then increase hydrogen bias compensation 10×(ΔT t -2σ)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ)kg / h; when ΔT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 15kg / h, if k t <0, then increase the hydrogen bias compensation by 10kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts,

[0100] The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-2 The above compensation scheme stops running until y≥2 / 3×L TCS-2 When the system restarts,

[0101] The callback scheme for hydrogen is: when 0≤ΔT t ≤σ, and x>0, then the hydrogen (ΔT t -2σ)kg / h, when -3σ≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when ΔT t When <-3σ and x>0, the hydrogen will be adjusted back to -10kg / h;

[0102] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h;

[0103] Step S76: When it is determined that the current time is within the interval [t3, 100],

[0104] The hydrogen bias compensation scheme is: when 2σ-0.3≤ΔT t ≤3σ+a2-0.3, if kt > 0, then increase hydrogen bias compensation 6×(ΔT t -2σ+0.3)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ+0.3)kg / h; when ΔT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 10kg / h, if k t <0, then increase the hydrogen bias compensation by 6kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts,

[0105] The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-3 The above compensation scheme stops running until y≥2 / 3×L TCS-3 When the system restarts,

[0106] The callback scheme for hydrogen is: when 0≤ΔT t ≤σ, and x>0, then the hydrogen (ΔT t -2σ)kg / h, when -3σ≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when ΔT t When <-3σ and x>0, the hydrogen will be adjusted back to -10kg / h;

[0107] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h.

[0108] Finally, this application provides a specific case study on establishing a unified and standardized atomization anomaly detection and processing method. Figure 2 , Figure 3 as well as Figure 4 The control parameter tables provided are for illustrative purposes only.

[0109] The first step is to calculate in real time the mean difference ΔT of the exhaust temperature at the chassis outlet at time t for 5 minutes t , ΔT t =average(T t , T t-1 , T t-2 , T t-3 , T t-4 )-average(T t-5 , T t-6 , T t-7 , T t-8 , T t-9 ); mean difference ΔT t The rate of change k t , k t =ΔT t -ΔT t-1 ; Cumulative adjustment of hydrogen flow x, cumulative adjustment of trichlorosilane flow y;

[0110] The second step is to Figure 2 Control parameter table 1 time interval [2,10], Figure 3 The time interval [10,40] of the control parameter table 2 and Figure 4 The time interval of control parameter table 3 is [40,100]. In each time interval, four modules are set: hydrogen bias compensation, trichlorosilane feed bias compensation, hydrogen callback, and trichlorosilane feed callback. The data acquisition frequency of all modules is 1 acquisition / min, the sending frequency of the compensation module is 1 time / min, and the sending frequency of the callback module is 1 time / 5min.

[0111] The third step is to determine if the current time is in the interval [2,10].

[0112] The hydrogen bias compensation scheme is: when 2.3≤ΔT t ≤3, if k t >0, then increase hydrogen bias compensation 10×(ΔT t -2)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2)kg / h; when ΔT t >3, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 15kg / h, if k t <0, then increase the hydrogen bias compensation by 10kg / h. When the cumulative adjustment of hydrogen flow rate x≥40, the above compensation scheme stops running until x≤20, then restarts running;

[0113] The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3, add trichlorosilane bias compensation -60kg / h. When the cumulative adjustment amount of trichlorosilane flow y≤-600, the above compensation scheme stops running until y≥-400, then restarts running;

[0114] The callback scheme for hydrogen is: when 0≤ΔT t ≤1, and x>0, then the hydrogen (ΔT t -2)kg / h, when -3≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2*(ΔT t -2)kg / h; when ΔT t <-3, and x>0, the hydrogen will be adjusted back to -10kg / h;

[0115] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h;

[0116] The fourth step is to determine if the current time is in the interval [10,40].

[0117] The hydrogen bias compensation scheme is: when 2≤ΔT t ≤3, if k t >0, then increase hydrogen bias compensation 10×(ΔT t -2)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2)kg / h; when ΔT t >3, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 15kg / h, if k t <0, then increase the hydrogen bias compensation by 10kg / h. When the cumulative adjustment of hydrogen flow rate x≥72, the above compensation scheme stops running until x≤48, then restarts running;

[0118] The feed bias compensation scheme for trichlorosilane is: when ΔT t≥3, add trichlorosilane bias compensation -60kg / h. When the cumulative adjustment amount of trichlorosilane flow y≤-600, the above compensation scheme stops running until y≥-400, then restarts running;

[0119] The callback scheme for hydrogen is: when 0≤ΔT t ≤1, and x>0, then the hydrogen (ΔT t -2)kg / h, when -3≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2)kg / h; when ΔT t <-3, and x>0, the hydrogen will be adjusted back to -10kg / h;

[0120] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h;

[0121] Step 5: When the current time is in the interval [40,100],

[0122] The hydrogen bias compensation scheme is: when 1.7≤ΔT t ≤2.7, if k t > 0, then increase hydrogen bias compensation 6×(ΔT t -1.7)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -1.7)kg / h; when ΔT t >3, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 10kg / h, if k t <0, then increase the hydrogen bias compensation by 6kg / h. When the cumulative adjustment of hydrogen flow rate x≥60, the above compensation scheme stops running until x≤40, then restarts running;

[0123] The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3, add trichlorosilane bias compensation -60kg / h. When the cumulative adjustment amount of trichlorosilane flow y≤-600, the above compensation scheme stops running until y≥-400, then restarts running;

[0124] The callback scheme for hydrogen is: when 0≤ΔT t ≤1, and x>0, then the hydrogen (ΔT t -2)kg / h, when -3≤ΔT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2)kg / h; when ΔT t <-3, and x>0, the hydrogen will be adjusted back to -10kg / h.

[0125] The callback scheme of trichlorosilane is: ΔT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h.

[0126] In summary, this application has formed a set of standard atomization anomaly detection and control processing procedures through historical data analysis and comprehensive consideration of the characteristics of different polysilicon reduction furnaces and their different system operating conditions, which has improved the accuracy and universality of atomization detection and identification, as well as the effectiveness of atomization control processing. It has basically solved the problems of excessive reliance on personnel experience, poor adaptability of automatic control programs, insufficient intelligence in parameter adjustment, and long debugging cycles, and has achieved a manual intervention frequency of ≤1 time / 10 furnaces during the production and operation of the reduction furnace, greatly reducing the labor intensity of personnel observation, inspection, and regulation.

[0127] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless defined as herein.

[0128] The meaning of "and / or" described in this application means that the situations where each exists alone or both exist at the same time are included.

[0129] The term “connection” as used in this application may mean a direct connection between components or an indirect connection between components via other components.

[0130] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for detecting and controlling atomization abnormality in a polysilicon reduction furnace, characterized in that: The specific steps include: Step S1: determining the key influencing factors of atomization abnormality detection and control, that is, detecting and controlling the chassis outlet exhaust temperature and its change trend; Step S2: batch read historical furnace data from the OPC database of the distributed control system, including reaction time, recipe initial value, recipe final setting value and chassis outlet exhaust gas temperature, locate the time point or time interval when the abnormality occurs by using the difference between the recipe initial value and the recipe final setting value, and then mark the abnormal point on the chassis outlet exhaust gas temperature data according to the located time point or time interval, so as to form a chassis outlet exhaust gas temperature abnormal point database as an experience database reserve; The formula includes hydrogen flow, trichlorosilane flow, A1 phase current, C1 phase current, A2 phase current, C2 phase current, B1 phase current and B2 phase current, the initial value of the formula is the original formula entered into the distributed control system by the engineer and not adjusted during operation, and the final setting value of the formula is the initial value of the formula plus the part adjusted by manual compensation during operation; Step S3: analyzing and processing the exhaust gas temperature data at the chassis outlet of the polysilicon reduction furnace collected on site, eliminating the fluctuation interference of the data, and statistically analyzing its numerical distribution; Step S4: Based on the data acquired in step S3, an absolute abnormal data range, an abnormal data range and a normal data range of the chassis outlet exhaust temperature change are set by an abnormality detection method, and the data are screened; Step S5: compare the abnormal data screened out in step S4 with the content in the experience database reserve in step S2. When the matching degree between the two is ≥90%, confirm that the abnormal detection method used in step S4 is effective. If the matching degree between the two is <90%, calibrate the abnormal detection method until the matching degree between the two is ≥90%. Step S6: by correcting the abnormality detection method in step S5, the time interval of silicon rod growth, the absolute abnormal interval, adjustment interval and fine-tuning interval of the exhaust temperature at the chassis outlet during production operation are finally confirmed, and the callback strength is determined through the callback boundary, and the compensation control limit is confirmed, so as to build the detection control processing logic; Step S7: Establish a unified and standardized atomization anomaly detection and processing method based on the detection control processing logic constructed in step S6.

2. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 1, characterized in that: The key influencing factors for determining abnormal atomization detection and control are based on the principle of radiation heat transfer, and the formula is: In the formula, Q represents the heat transfer from the silicon rod to the inner wall of the furnace, k is the radiation heat transfer coefficient, A represents the emission area of ​​the silicon rod to the inner wall of the furnace, T1 represents the emission temperature of the silicon rod to the inner wall of the furnace, and T2 represents the receiving temperature of the inner wall of the furnace to the silicon rod; When the emission area A of the silicon rod to the inner wall of the furnace tube decreases, the heat transfer from the silicon rod to the inner wall of the furnace tube decreases. Since the total heat generation at the abnormal moment remains unchanged, the heat discharged from the exhaust hole increases, causing the exhaust temperature at the chassis outlet to change.

3. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 1, characterized in that: In step S2, when historical furnace data is read in batches from the OPC database of the distributed control system, the data granularity is 1 record / min.

4. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 1, characterized in that: In step S3, the exhaust gas temperature data at the chassis outlet of the polysilicon reduction furnace collected on-site is averaged to obtain short-time average data and long-time average data, wherein the short-time interval is 3-10 minutes and the long-time interval is 60-120 minutes.

5. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 4, characterized in that: The short time setting is 5 minutes and the long time setting is 120 minutes.

6. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 5, characterized in that: In step S4, the obtained short-time mean data and long-time mean data are subjected to rolling difference calculation at fixed time intervals, specifically, the short-time mean data is subtracted from the long-time mean data, and then the distribution of the mean difference data is checked using a histogram; Based on the calculated mean difference data of chassis outlet exhaust temperature at different time points combined with the histogram, the anomaly detection method is determined to be the 3σ principle; The calculation method of standard deviation σ is: In the formula, n represents the number of chassis outlet exhaust temperature mean difference data samples. represents the average of all mean difference data, x i Represents the value of the i-th mean difference data.

7. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 6, characterized in that: In step S4, according to the 3σ principle, the absolute abnormal data range of the mean difference data of the chassis outlet exhaust gas temperature is set to [-∞, -3σ] and [3σ, ∞], the abnormal data range is set to [-3σ, -2σ] and [2σ, 3σ], and the normal data range is set to [-σ, σ]; The mean difference data of chassis outlet exhaust gas temperature with values ​​exceeding 2σ are screened.

8. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 7, characterized in that: In step S5, the abnormality detection method is calibrated, that is, the mean difference of the exhaust gas temperature at the chassis outlet is screened until the matching degree between the two is ≥90%.

9. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 8, characterized in that: In step S6, based on the fact that the production stages of the silicon rod include an accelerated growth stage, a rapid growth stage, and a stable growth stage, the time interval for silicon rod growth is divided into an accelerated growth time interval [0, t2], a rapid growth time interval [t2, t3], and a stable growth time interval [t3, 100]; The time interval of accelerated growth [0, t2] is further divided into the time interval not involved in regulation [0, t1] and the time interval of accelerated growth involved in regulation [t1, t2]; The mean difference of the chassis outlet exhaust temperature was screened and it was confirmed that the absolute abnormal interval was [3σ+a2,∞], the adjustment interval was [2σ+a1,3σ+a2], and the fine-tuning interval was [σ+a0,2σ+a1]; Among them, a0 is the empirical compensation for threshold correction in the fine-tuning interval, that is, in the fine-tuning interval, the mean difference of the chassis outlet exhaust gas temperature with a value exceeding σ+a0 is screened; a1 is the empirical compensation for threshold correction in the adjustment interval, that is, in the adjustment interval, the mean difference of the chassis outlet exhaust gas temperature with a value exceeding 2σ+a1 is screened; a2 is the empirical compensation for threshold correction in the absolute abnormal interval, that is, in the absolute abnormal interval, the mean difference of the chassis outlet exhaust gas temperature with a value exceeding 3σ+a2 is screened; Set the callback boundary σ and use it to determine the callback strength. The callback strength includes three levels: [0, σ], [-3σ, 0] and [-∞, -3σ]; The compensation control limit includes the cumulative adjustment of the hydrogen flow rate during the accelerated growth time interval [t1, t2] The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-1 , is the maximum limit of hydrogen compensation in the accelerated growth time interval [t1, t2], ranging from (0, +∞), L TCS-1 is the maximum limit of trichlorosilane compensation in the accelerated growth time interval [t1, t2], and its value range is (-∞, 0); The cumulative adjustment of hydrogen flow rate during the rapid growth time interval [t2, t3] The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-2 , is the maximum limit of hydrogen compensation in the rapid growth time interval [t2, t3], ranging from (0, +∞), L TCS-2 is the maximum limit of trichlorosilane compensation in the rapid growth time interval [t2, t3], and its value range is (-∞, 0); Cumulative adjustment of hydrogen flow rate within the stable growth time interval [t3, 100] The cumulative adjustment of trichlorosilane flow rate ≥ L TCS-3 , is the maximum limit of hydrogen compensation in the stable growth time interval [t3, 100], and its value range is (0, +∞); L TCS-3 is the maximum limit of trichlorosilane compensation in the stable growth time interval [t3, 100], and its value range is (-∞, 0).

10. The method for detecting and controlling atomization abnormality of a polysilicon reduction furnace according to claim 9, characterized in that: In step S7, establishing a unified standardized atomization anomaly detection and processing method includes the following steps: Step S71, real-time calculation of the mean difference AT of the chassis outlet exhaust temperature at time t for 5 minutes t , mean difference AT t The rate of change k t , the cumulative adjustment amount x of hydrogen flow and the cumulative adjustment amount y of trichlorosilane flow; where AT t =average(T t , T t-1 , T t-2 , T t-3 , T t-4 )-average(T t-5 , T t-6 , T t-7 , T t-8 , T t-9 ), k t =AT t -AT t-1 ; Step S72, for the accelerated growth time interval [t1, t2], the rapid growth time interval [t2, t3] and the stable growth time interval [t3, 100], four modules of hydrogen bias compensation, trichlorosilane feed bias compensation, hydrogen callback, and trichlorosilane feed callback are set in each time interval; the data acquisition frequency of the four modules is 1 time / min, the sending frequency of the hydrogen bias compensation and trichlorosilane feed bias compensation modules is 1 time / min, and the sending frequency of the hydrogen callback and trichlorosilane feed callback modules is 1 time / 5min; Step S73, determining whether the current operation time of the polysilicon reduction furnace is in the accelerated growth time interval [t1, t2] or the rapid growth time interval [t2, t3] or the stable growth time interval [t3, 100]; if it is in the accelerated growth time interval [t1, t2], proceed to step S74; if it is in the rapid growth time interval [t2, t3], proceed to step S75; if it is in the stable growth time interval [t3, 100], proceed to step S76; Step S74: when it is determined that the current time is within the interval [t1, t2], The hydrogen bias compensation scheme is: when 2σ+a1≤AT t ≤3σ+a2, if k t >0, then increase hydrogen bias compensation 10×(ΔT t -2σ)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ)kg / h; when AT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 15kg / h, if k t <0, then increase the hydrogen bias compensation by 10kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts, The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-1 The above compensation scheme stops running until y≥2 / 3×L TCS-1 When the system restarts, The callback plan for hydrogen is: when 0≤AT t ≤σ, and x>0, then the hydrogen (ΔT t -2σ)kg / h, when -3σ≤AT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when AT t <-3σ, and x>0, the hydrogen will be adjusted back to -10kg / h; The callback plan for trichlorosilane is: AT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h; Step S75: when it is determined that the current time is within the interval [t2, t3], The hydrogen bias compensation scheme is: when 2σ≤ΔT t ≤3σ+a2, if k t >0, then increase hydrogen bias compensation 10×(ΔT t -2σ)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ)kg / h; when AT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 15kg / h, if k t <0, then increase the hydrogen bias compensation by 10kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts, The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-2 The above compensation scheme stops running until y≥2 / 3×L TCS-2 When the system restarts, The callback plan for hydrogen is: when 0≤AT t ≤σ, and x>0, then the hydrogen (ΔT t -2σ)kg / h, when -3σ≤AT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when AT t <-3σ, and x>0, the hydrogen will be adjusted back to -10kg / h; The callback plan for trichlorosilane is: AT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h; Step S76: When it is determined that the current time is within the interval [t3, 100], The hydrogen bias compensation scheme is: when 2σ-0.3≤AT t ≤3σ+a2-0.3, if k t > 0, then increase hydrogen bias compensation 6×(ΔT t -2σ+0.3)kg / h, if k t =0, then increase the hydrogen bias compensation by 4kg / h. If k t <0, then increase hydrogen bias compensation 4×(ΔT t -2σ+0.3)kg / h; when ΔT t >3σ+a2, if k t ≥1, then add hydrogen bias compensation 20kg / h, if 0≤k t <1, then increase hydrogen bias compensation 10kg / h, if k t <0, then increase the hydrogen bias compensation by 6kg / h. When the cumulative adjustment of hydrogen flow The above compensation plan will stop running until When the system restarts, The feed bias compensation scheme for trichlorosilane is: when ΔT t ≥3σ+a2, add trichlorosilane bias compensation -60kg / h, when the cumulative adjustment amount of trichlorosilane flow y≤L TCS-3 The above compensation scheme stops running until y≥2 / 3×L TCS-3 When the system restarts, The callback plan for hydrogen is: when 0≤AT t ≤σ, and x>0, then the hydrogen (ΔT t -2σ)kg / h, when -3σ≤AT t ≤0, and x>0, if k t >0, then call back hydrogen (ΔT t -2σ)kg / h, if k t =0, then the hydrogen will be adjusted back to -4kg / h. If k t <0, then call back hydrogen 2×(ΔT t -2σ)kg / h; when AT t When <-3σ and x>0, the hydrogen will be adjusted back to -10kg / h; The callback plan for trichlorosilane is: AT t ≤0, and y<0, the trichlorosilane feed is adjusted back to 20kg / h.

Images