Perovskite thin film defect tracing method and system

Abstract

The invention belongs to the technical field of industrial data processing, and particularly relates to a perovskite thin film defect tracing method and system, and the method comprises the following steps: collecting a real-time optical response data sequence of a perovskite thin film in a preparation process, obtaining a corresponding standard reference sequence, and carrying out the preprocessing of the real-time optical response data sequence; and respectively calculating a first derivative and a second derivative of the real-time optical response data sequence and the standard reference sequence, calculating a rate deviation according to the first derivative, determining a crystallization acceleration weight according to the second derivative, and combining the numerical deviation, the rate deviation and the crystallization acceleration weight to construct a crystallization dynamics inertia mismatch index. According to the method, the contradiction between leak detection in a constant speed period and delay in an outbreak period is solved, and the overall deviation of the system is amplified in a nucleation outbreak period and a logarithmic item weight index level, so that tiny dynamic anomalies can be caught sensitively, and hidden defects are effectively identified.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of industrial data processing, and particularly relates to a perovskite film defect tracing method and system. BACKGROUND

[0002] As a representative of the new generation of photovoltaic technology, perovskite solar cells have attracted much attention due to their high efficiency and low cost. In the process of realizing mass production of more than 100 megawatts, large-area slot coating process is one of the core technologies. This process requires precise control of the synergistic effect of physical field and chemical reaction field when the perovskite wet film undergoes solvent evaporation and crystallization phase transition. Usually, the production line uses air knife blowing or vacuum assistance to accelerate solvent evaporation, and then induces film crystallization through thermal annealing, which is extremely sensitive to environmental parameters.

[0003] However, the existing quality control methods mainly rely on post-detection after the film preparation, such as automatic optical detection or electroluminescence detection. Although this delayed detection method can identify final defects such as pinholes and cracks, it cannot effectively trace the specific process steps where the defects are generated. In actual production, small fluctuations in air knife wind pressure (physical disturbance) and precursor liquid component aging or annealing temperature drift (chemical delay) often lead to crystalline uniformity defects that are extremely similar in appearance. Due to the lack of in-depth analysis of process dynamics in existing technologies, it is difficult to distinguish the root causes of these two types of defects with different properties, making it difficult for process personnel to take timely and appropriate measures, and the yield rate of the production line faces a bottleneck.

[0004] In addition, traditional online monitoring methods usually collect reflectance curves and calculate their Euclidean distance from standard curves. However, this method has serious defects: it is sensitive to the size of the numerical value, but not sensitive to the rate of change. In the nucleation burst period of perovskite crystallization, very small differences in kinetic rate are the root cause of defect generation, and these differences are often masked in the background noise of the overall signal, leading to missed detection or false positives, and failing to accurately guide the adjustment of process parameters on the production line. SUMMARY

[0005] The present application provides a perovskite film defect tracing method and system to solve the technical problems of being unable to distinguish the physical or chemical root causes of perovskite film defects and being insensitive to small abnormalities in crystallization kinetics in the prior art.

[0006] In a first aspect, the present application provides a perovskite film defect tracing method, comprising the following steps: Collecting real-time optical response data sequences of perovskite films during the preparation process, and obtaining corresponding standard reference sequences, and preprocessing the real-time optical response data sequences; The first derivative and the second derivative of the real-time optical response data sequence and the standard reference sequence are calculated respectively, the rate deviation is calculated according to the first derivative, the crystallization acceleration weight is determined according to the second derivative, and the crystallization kinetics inertia mismatch index is constructed by combining the numerical deviation, the rate deviation and the crystallization acceleration weight; According to the slope change characteristics of the standard reference sequence, the preparation process is divided into a solvent evaporation segment and a nucleation burst segment, the crystallization kinetics inertia mismatch index of different stages is calculated by using the fluid dynamics sensitive factor and the chemical reaction sensitive factor, and the wind pressure abnormal confidence is obtained. According to the numerical size of the wind pressure abnormal confidence, the defect source is determined to be a gas knife wind pressure abnormality or an annealing chemical reaction abnormality, and a control instruction is sent to the production line equipment according to the determination result.

[0007] Further, when constructing the crystallization kinetics inertia mismatch index, the calculation formula is: ; In the formula, is the crystallization kinetics inertia mismatch index at time , is the real-time optical response data at time , is the standard reference sequence value at time , and and are the crystallization rates of the real-time optical response data and the standard reference sequence at time , is the crystallization acceleration of the standard reference sequence at time , and and are the numerical weight coefficient and the rate weight coefficient, is a reference constant, and is an inertia gain coefficient. Further, the pre-processing of the real-time optical response data sequence includes: using a moving average filter to smooth the collected real-time optical response data sequence, removing random jitter high-frequency noise caused by gas knife gas flow, retaining low-frequency signals reflecting the trend of crystallization kinetics, and the window size of the moving average filter is set to a preset value.

[0008] Further, the calculation formula of the wind pressure abnormal confidence is:

[0009] ; In the formula, is the wind pressure abnormal confidence, is the time interval of the solvent evaporation segment, is the time interval of the nucleation burst segment, is the crystallization kinetics inertia mismatch index at time , and and are the numerical weight coefficient and the rate weight coefficient.a fluid dynamics sensitive factor, a chemical reaction sensitive factor, a chemical reaction sensitive factor, a normalized smoothing constant, a chemical reaction sensitive factor, a crystallization kinetics inertia mismatch index.

[0010] Further, the fluid dynamics sensitive factor is constructed by a linear increasing function, and the chemical reaction sensitive factor is a constant; The process of dividing according to the slope change characteristics of the standard reference sequence includes: defining a period from the start of monitoring to before the sharp drop of reflectivity as a solvent evaporation segment, and defining a period of sharp drop of reflectivity as a nucleation burst segment.

[0011] Further, the defect source is determined according to the numerical size of the wind pressure abnormality confidence, including: A preset traceability threshold is set, and if the wind pressure abnormality confidence is greater than the traceability threshold, the defect source is determined to be air knife equipment abnormality; If the wind pressure abnormality confidence is less than or equal to the traceability threshold, the defect source is determined to be annealing temperature abnormality or precursor liquid component abnormality.

[0012] Further, according to the determination result, a control instruction is sent to the production line equipment, including: If the air knife equipment is determined to be abnormal, the instruction control module checks the frequency of the air knife gas supply pump and performs compensatory fine tuning; If the annealing temperature is determined to be abnormal, the instruction temperature control module checks the thermocouple reading and performs self-tuning on the heating power.

[0013] In a second aspect, the present application provides a perovskite thin film defect traceability system, comprising a memory and a processor, the memory stores computer program instructions, when the computer program instructions are executed by the processor, the above-mentioned perovskite thin film defect traceability method is realized.

[0014] Further, the system further comprises a non-contact optical monitoring probe, which is installed behind the air knife blowing area and the crystallization starting segment, and is used for real-time acquisition of real-time optical response data.

[0015] Further, the system further comprises a communication module, which is used to connect a production management system to obtain a standard reference sequence, and connect a production line PLC to send a control instruction.

[0016] Beneficial effects are: The present application solves the contradiction between missing detection in the uniform speed period and dullness in the burst period, introduces acceleration as the inertial weight of the system, and constructs a weighted mismatch index containing numerical deviation and rate deviation. In the nucleation burst period, the logarithmic term amplifies the deviation of the whole system by an exponential order, so that the tiny dynamic abnormality can be captured sensitively, and the implicit defects can be identified effectively.

[0017] The present application realizes precise decoupling of physical and chemical causes, uses a wind pressure abnormality confidence formula based on process mechanism to separate and evaluate errors in the time domain, does not need to rely on a large amount of label data for training, and can accurately distinguish whether the physical disturbance is caused by the air knife or the chemical delay is caused by the temperature / formula, and has strong explainability and on-site applicability. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A flowchart of a perovskite thin film defect tracing method.

[0019] Figure 2 An effect diagram of the original data acquisition layer.

[0020] Figure 3 An effect diagram of the prior art analysis layer.

[0021] Figure 4 An effect diagram of the present application scheme layer. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0023] Embodiments of the perovskite thin film defect tracing method provided by the present application: The present embodiment is applied to a hundred megawatt perovskite battery slot coating production line. In the production process, after the substrate is coated with a precursor liquid by a coating head, it enters a drying area to receive air knife blowing, and then enters an annealing furnace. The present system is deployed between the drying area and the annealing area, and is used for real-time monitoring and tracing of thin film defects.

[0024] As shown in Figure 1 The perovskite thin film defect tracing method comprises the following steps: S1, collecting real-time optical response data sequences of perovskite thin films in the preparation process, and obtaining corresponding standard reference sequences, and pre-processing the real-time optical response data sequences.

[0025] Specifically, a non-contact optical monitoring probe is installed behind the air knife purging area and at the crystallization initiation stage. Examples include a single-wavelength laser reflectometer or a high-speed spectrometer, with a working wavelength of 635 nm, to obtain better contrast between the liquid film and the crystal. The sampling frequency is set to... The optical response intensity, such as reflectivity, of the thin film surface is collected in real time at a frequency of 50Hz. The real-time optical response data sequence collected for the current production batch is denoted as... ,in Indicates a relative process time point, for example This represents the 0.02nd second. Simultaneously, the standard reference sequence corresponding to the current product model is retrieved from the Production Management System (MES) database. This standard reference sequence is a golden curve obtained by weighted averaging of multiple production data under an ideal process window with extremely high yield.

[0026] Because the airflow from the air knife in the production area causes slight fluctuations in the liquid surface, high-frequency noise is present in the signal. A moving average filter is used to... Smoothing is performed using a moving average filter with WindowSize=5 to remove random jitter and retain low-frequency signals that reflect the crystallization dynamics.

[0027] S2, calculate the first and second derivatives of the real-time optical response data sequence and the standard reference sequence respectively, calculate the rate deviation based on the first derivative, determine the crystallization acceleration weight based on the second derivative, and construct the crystallization dynamics inertial mismatch index by combining the numerical deviation, rate deviation and crystallization acceleration weight.

[0028] Existing technologies only consider numerical differences. This invention introduces speed difference and acceleration weights. The physical logic is that when the system is in a phase of rapid acceleration, its inertia is large and it is extremely sensitive to small disturbances. At this time, the weight of the overall system error should be amplified. However, during the phase of uniform change, the normal weight should be maintained.

[0029] Specifically, the first derivative (velocity) and second derivative (acceleration) of the test sequence and the standard sequence are calculated using the discrete difference method.

[0030] Crystallization rate:

[0031] Crystallization acceleration:

[0032] In a physical sense, it represents the intensity of the phase transition process, when When the value is very large, it means that the thin film is at the point of sudden phase transition from liquid to solid state.

[0033] To precisely capture the anomaly, the following formula of crystallization kinetics inertia mismatch index is constructed :

[0034] Wherein, the meaning and value of each parameter are as follows: is the numerical deviation term, reflecting the difference of absolute reflectivity; is the rate deviation term, reflecting the difference of crystallization speed; is the reference constant, the value constraint is , is a natural constant, about 2.718, the reason for setting is that when the standard acceleration , the logarithmic term , which ensures that in the non-phase transition burst period, the basic error in the parentheses will not be zeroed out, thus avoiding missing detection; is the inertia gain coefficient, for example, 10.0, used to adjust the sensitivity to the nucleation burst period; is the weight coefficient, , exemplarily, such as , to reflect the emphasis on the kinetics process.

[0035] In order to more clearly explain the calculation process and effect of the formula, the calculation example is as follows: Assume that in the solvent evaporation stable period, time : The standard value , the previous time , then , and the stage is uniform, assuming ; The observed value , the previous time , then .

[0036] At this time, the numerical deviation , the rate deviation ; the basic error and in the parentheses are , the acceleration weight term , the index , the algorithm degenerates to the conventional numerical detection, reflecting the basic deviation.

[0037] Assume that in the nucleation burst period, time : The standard value , the previous time , then , the previous time ,but (Dramatic change); Observations The moment before ,but .

[0038] At this point, the numerical deviation Rate deviation The sum of the basic errors in parentheses is Acceleration weighting term ;index .

[0039] By introducing acceleration as an overall multiplicative weight, any deviation in the system is amplified during the burst phase, thus keenly capturing minute mismatches in the dynamics.

[0040] like Figure 2 As shown, the abnormal production data curve (red dashed line) and the standard production data curve (black solid line) in the original data are closely intertwined and difficult to distinguish with the naked eye. Figure 3 As shown, the traditional numerical difference method yields results (grey curve) with extremely low amplitudes, making it prone to missed detections. However, as... Figure 4 As shown, after applying this method, an extremely high and sharp main peak (blue solid line) is observed during the nucleation and explosion phase, significantly exceeding the threshold.

[0041] By introducing a nonlinear weighted formula with acceleration as the overall weight, the basic detection accuracy is guaranteed during the steady period, while the logarithmic term is used to exponentially amplify the overall deviation of the system during the nucleation and explosion period. This enables the system to keenly capture dynamic anomalies and solves the problem that traditional methods are not sensitive to dynamic processes.

[0042] S3. Based on the slope change characteristics of the standard reference sequence, the preparation process is divided into solvent evaporation stage and nucleation burst stage. The crystallization kinetic inertial mismatch index at different stages is calculated in segments using fluid dynamics sensitive factors and chemical reaction sensitive factors to obtain the confidence level of wind pressure anomaly.

[0043] The entire process was calculated. Following the curve, the first step is to determine if any anomalies exist. If the maximum peak value of the curve exceeds the initial warning line, such as 0.3, a potential process defect is considered to exist, triggering the source tracing calculation in this step. This step uses segmented decoupling to determine whether the anomaly is caused by fluctuations in air knife pressure, or by abnormalities in the precursor liquid components or thermal environment.

[0044] Specifically, based on the slope variation characteristics of the standard reference sequence curve, the entire process timeline is automatically divided into two intervals: interval (Solvent evaporation phase): Defined as the period from the start of monitoring until the reflectivity begins to decrease sharply. At this moment, this stage is mainly affected by physical wind force, i.e. air knife.

[0045] Interval (Nucleation burst segment): defined as the interval of reflectivity sharply decreasing (R to ), this stage is mainly affected by chemical thermodynamics, such as temperature or formula.

[0046] Build a ratio model to calculate the wind pressure abnormal confidence : ; Wherein: is the fluid dynamics sensitive factor, which is obtained by using a linear increasing function to construct, for example . Its physical meaning is that in the stage, as the solvent volatilizes, the liquid film gradually thins, the surface area volume ratio increases, and the sensitivity of the liquid film to the shear force of the airflow increases linearly with time; is the chemical reaction sensitive factor, which is set to a constant, for example 1.0, in the stage, which is used to count the cumulative error of the nucleation stage; is a normalized smoothing constant, for example 50, which is set because if the nucleation segment error is 0, the denominator will be too small, causing the calculation to diverge; at the same time, as a baseline noise value, it avoids small noise triggering a dramatic fluctuation in confidence.

[0047] Example: Case A: if the fault is caused by air knife wind pressure fluctuation, the error is mainly concentrated in the segment, assuming that the molecular cumulative value is 100, and the segment error is smaller, the denominator cumulative value is 20, plus , then , the value is large.

[0048] Case B: if the fault is caused by annealing temperature, the error is mainly concentrated in the segment, assuming that the molecular cumulative value is 10, i.e. the wind pressure segment is normal, and the denominator cumulative value is 150, then , the value is very small.

[0049] By segment integration and ratio calculation based on process mechanism, the dynamic mismatch index of the whole process is mapped to the wind pressure abnormal confidence, realizing the quantitative decoupling of physical defects and chemical defects, and providing a basis for accurate tracing.

[0050] S4, according to the numerical value of the wind pressure abnormal confidence, determine whether the defect source is air knife wind pressure abnormality or annealing chemical reaction abnormality, and send control instructions to the production line equipment according to the determination result.

[0051] After the system calculates the wind pressure anomaly confidence, the following determination logic is executed: Case A, if (pre-set traceability threshold), the determination result is that the defect source is air knife / air knife equipment abnormality; the logic is that the error is mainly concentrated in the solvent evaporation section, and meets the time characteristics of fluid disturbance. The system checks the air knife air supply pump frequency through industrial Ethernet instructions PLC, and performs compensatory fine tuning, such as increasing 0.5 Hz.

[0052] Case B, if , the determination result is that the defect source is annealing temperature abnormality or precursor liquid component aging; the logic is that the error is mainly concentrated in the nucleation section, indicating that the phase transition kinetics is a problem, not a physical purge problem. The system instructs the temperature control module to check the thermocouple reading and performs PID self-tuning on the heating power, or issues an alarm to check the precursor liquid state.

[0053] By setting a clear traceability threshold to determine the wind pressure anomaly confidence, the specific process link where the defect occurs can be automatically locked, and the production line equipment can be directly linked for closed-loop adjustment, realizing full-automatic closed-loop control from defect identification, cause analysis to process optimization, and significantly improving the yield rate and intelligent level of the production line.

[0054] As shown in Figure 4 , after the method of the embodiment is used, the small kinetic differences originally hidden in the noise are significantly amplified as sharp main peaks in the nucleation burst period, clearly indicating the time point and intensity of the anomaly. Combined with the process of Figure 1 , the system can further calculate the confidence and complete the traceability according to the feature. This analysis method based on kinetic inertia characteristics has higher sensitivity and specificity than simple numerical comparison, effectively solving the problem of difficult capture and attribution of small defects in complex industrial environments.

[0055] Embodiments of the perovskite thin film defect traceability system provided by the application: The perovskite thin film defect traceability system includes a processor and a memory, and the memory stores computer program instructions that, when executed by the processor, implement the perovskite thin film defect traceability method described above.

[0056] The perovskite thin film defect traceability system also includes communication interfaces and other components well known to those skilled in the art, the settings and functions of which are known in the art, and therefore will not be described here.

[0057] In this disclosure, a "storage medium" or "computer readable medium" can be any available medium that can be accessed by a general purpose or special purpose computer system. By way of example, and not limitation, such computer readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code means in the form of computer readable instructions or data structures and that can be accessed by a general purpose or special purpose computer system, or a combination thereof. When information is transferred or provided over a network or another communications connection (either hardwired, wireless, or combination thereof) to a computer, the computer properly views the connection as a computer readable medium. Thus, any such connection is properly termed a computer readable medium. Combinations of the above should also be included within the scope of the computer readable media.

[0058] The above are preferred embodiments of the present application, not to limit the protection scope of the present application, therefore: any equivalent changes made according to the structure, shape, principle of the present application, should be covered within the protection scope of the present application.

Claims

1. A method for defect tracing of a perovskite thin film, characterized by, The method comprises the following steps: a real-time optical response data sequence of the perovskite film in the preparation process is collected, a corresponding standard reference sequence is obtained, and the real-time optical response data sequence is preprocessed; first and second derivatives of the real-time optical response data sequence and the standard reference sequence are calculated respectively, a rate deviation is calculated according to the first derivative, a crystallization acceleration weight is determined according to the second derivative, and a crystallization kinetics inertia mismatch index is constructed by combining the numerical deviation, the rate deviation and the crystallization acceleration weight; the preparation process is divided into a solvent evaporation segment and a nucleation burst segment according to the slope change characteristics of the standard reference sequence, the crystallization kinetics inertia mismatch index of different stages is calculated by using a fluid dynamics sensitive factor and a chemical reaction sensitive factor, and an air pressure abnormality confidence is obtained; the source of the defect is determined to be air knife air pressure abnormality or annealing chemical reaction abnormality according to the numerical value of the air pressure abnormality confidence, and a control instruction is sent to the production line equipment according to the determination result.

2. The perovskite thin film defect tracing method according to claim 1, wherein, When the crystallization kinetics inertia mismatch index is constructed, the calculation formula is: ； wherein, is the crystallization kinetics inertia mismatch index at time is the real-time optical response data at time is the standard reference sequence value at time and are the crystallization rates of the real-time optical response data and the standard reference sequence at time is the crystallization acceleration of the standard reference sequence at time and are the numerical weight coefficient and the rate weight coefficient, respectively, is the reference constant, is the inertia gain coefficient.​​​​​ 3. The perovskite thin film defect tracing method according to claim 1, wherein The preprocessing of the real-time optical response data sequence comprises the following steps: a sliding average filter is used to smooth the collected real-time optical response data sequence, random jitter high-frequency noise caused by air knife gas flow is removed, low-frequency signals reflecting the trend of crystallization kinetics are retained, and the window size of the sliding average filter is set to a preset value.

4. The perovskite thin film defect tracing method according to claim 1, wherein The calculation formula of the air pressure abnormality confidence is: ； In the formula, The confidence level for wind pressure anomalies. This refers to the time interval during solvent evaporation. The time interval of the nucleation and explosion phase. For a moment Fluid dynamics sensitive factors, For a moment Chemical reaction sensitive factors The normalized smoothing constant is For a moment The inertial mismatch index of crystallization kinetics.

5. The perovskite thin film defect tracing method according to claim 4, wherein The fluid dynamics sensitive factor is constructed by using a linear increasing function, and the chemical reaction sensitive factor is a constant; The process is divided according to the slope change characteristics of the standard reference sequence, which comprises the following steps: a period from the start of monitoring to before the reflectivity starts to sharply decrease is defined as the solvent evaporation segment, and a period in which the reflectivity sharply decreases is defined as the nucleation burst segment.

6. The perovskite thin film defect tracing method according to claim 1, wherein The source of the defect is determined according to the numerical value of the air pressure abnormality confidence, which comprises the following steps: A preset traceability threshold is set, if the air pressure abnormality confidence is greater than the traceability threshold, the source of the defect is determined to be air knife equipment abnormality; If the air pressure abnormality confidence is less than or equal to the traceability threshold, the source of the defect is determined to be annealing temperature abnormality or precursor liquid component abnormality.

7. The perovskite thin film defect tracing method according to claim 6, wherein According to the determination result, a control instruction is sent to the production line equipment, which comprises the following steps: If it is determined that the air knife equipment is abnormal, an instruction control module checks the frequency of the air knife gas supply pump and performs compensatory fine tuning; If it is determined that the annealing temperature is abnormal, an instruction temperature control module checks the thermocouple reading and performs self-tuning on the heating power.

8. A perovskite thin film defect tracing system, characterized in that, The system further comprises a memory and a processor, and the memory stores computer program instructions, which, when executed by the processor, implement the perovskite film defect traceability method of any one of claims 1-7.

9. The perovskite thin film defect tracing system of claim 8, wherein, The system further comprises a non-contact optical monitoring probe, which is installed behind the air knife blowing area and at the crystallization starting segment, and is used for collecting real-time optical response data in real time.

10. The perovskite thin film defect tracing system of claim 8, wherein, The system further comprises a communication module, which is used for connecting a production management system to obtain a standard reference sequence, and connecting a production line PLC to send a control instruction.

Images