LED chip etching processing system with multi-process parameter collaborative optimization

Abstract

The application discloses a multi-process parameter collaborative optimization LED chip etching processing system and particularly relates to the field of LED chip etching process control, and comprises a data arrangement module, a risk identification module, a path screening module, an instruction generation module, a process control module and a cycle updating module; the data arrangement module is used for reading the chip identification of the current batch, an etching sub-step set, sub-step sequence restrictions and online data collected by an etching device, performing unified time alignment processing on the online data, and forming current state data; through edge computing, the current state data in the etching process is subjected to risk identification, sub-step sequence re-screening and multi-process parameter joint generation, and dynamic process control is implemented accordingly, so as to solve the problem of how to dynamically adjust the sub-step sequence according to the current processing state to inhibit the amplification of the front-stage risk.

Description

Technical Field

[0001] This invention relates to the field of LED chip etching process control technology, and more specifically, to an LED chip etching processing system with multi-process parameter synergistic optimization. Background Technology

[0002] In the field of LED chip etching, existing technologies mainly focus on achieving stable process control of the processing while ensuring that the etching depth, sidewall morphology, selectivity and endpoint control meet the requirements. The usual practice is to preset the combination of process parameters according to the pretreatment, main etching and over-etching stages, and the equipment executes the segmented formula in a predetermined order, and adjusts the local parameters in combination with online detection results and local feedback from the edge computing unit. Taking the continuous mass production scenario of multi-layer LED chips as an example, when there are areas with significant differences in pattern density on the same wafer, and the equipment only has the ability to execute uniformly at the whole wafer level, and the production line cycle does not allow for frequent insertion of offline measurement and manual rescheduling, the existing practices can maintain the overall process window by shortening the single-segment duration, adding protection segments, or fine-tuning parameters. However, since the process control is still based on a fixed sequence of steps, and edge computing is only used to accelerate parameter correction under the existing path, once the front-end processing prioritizes the amplification of dominant risks such as sidewall damage, residual curing, or end-point margin compression, it is difficult to recover even if compensation is continued in the subsequent stages. In actual production, there will be a verifiable phenomenon that once the front-end deviation is formed, it is difficult to pull back the results in the subsequent stages. Therefore, the technical problem to be solved by this application is: how to dynamically adjust the execution order of sub-steps according to the dominant irreversible risk corresponding to the current processing state in the edge computing-based LED chip etching process control, under the condition of unified execution at the whole chip level and limited cycle time, so as to avoid the continuous compression of subsequent repairable space after the risk in the front end is amplified. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an LED chip etching processing system with multi-process parameter collaborative optimization. This system uses edge computing to identify risks, re-select sub-step sequences, and generate multi-process parameters in conjunction with the current state data during the etching process, and implements dynamic process control accordingly to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an LED chip etching processing system with multi-process parameter collaborative optimization, including a data processing module, a risk identification module, a path selection module, an instruction generation module, a process control module, and a cyclic update module; The data processing module is used to read the chip identifier, etching sub-step set, sub-step sequence restrictions and online data collected by the etching equipment for the current batch, and perform unified time alignment processing on the online data to form the current status data; The risk identification module is used to input the current status data into the identification rules corresponding to each irreversible risk, solve the risk value corresponding to each irreversible risk, and identify the dominant irreversible risk type according to the risk value sorting results. The path filtering module is used to generate rules for the dominant irreversible risk type, the set of etching sub-steps, and the input order of sub-steps, construct a set of candidate sub-step sequences that meet the execution conditions, and filter the target sub-step sequence from the set of candidate sub-step sequences. The instruction generation module is used to input the current status data and the target sub-step sequence into the parameter generation rules, calculate the process parameters corresponding to each sub-step in the target sub-step sequence, and form control instructions corresponding to the target sub-step sequence. The process control module is used to send control commands to the etching equipment, control the etching equipment to perform etching according to the target sub-step sequence and process parameters, and collect stage response data after each sub-step is executed; The cyclic update module is used to merge the stage response data with the current status data to form updated status data. Based on the updated status data, the dominant irreversible risk type is re-identified and the target sub-step sequence is re-selected until the etching control result corresponding to the current batch is formed.

[0005] In a preferred embodiment, the data processing module includes: Read the chip identifier, etching sub-step set, sub-step sequence restrictions, and online data collected by the etching equipment for the current batch; and aggregate the online data according to the current batch and chip identifier to form the raw data. Extract the collection time corresponding to each data item in the raw data, and perform time mapping on data items with different collection times according to a unified sampling time to form aligned data; The missing data items in the aligned data are padded, and the data items that are out of the valid range are marked as invalid, thus forming regularized data. Based on the set of etching sub-steps and the constraints on the order of sub-steps, the regularized data is divided into the corresponding sub-step intervals and organized in chronological order to form the current state data.

[0006] In a preferred embodiment, the risk identification module includes: Read the current status data, extract the corresponding risk data from the current status data according to the data items corresponding to each irreversible risk, and form the judgment data corresponding to each irreversible risk; Input the judgment data corresponding to each irreversible risk into the corresponding identification rules, calculate the current value, the difference between adjacent time points and the duration of continuous change in the judgment data, and form the sub-risk value corresponding to each irreversible risk. The individual risk values ​​corresponding to each irreversible risk are summarized and calculated to form the total risk value corresponding to each irreversible risk, and the risk is ranked from high to low according to the total risk value. Read the first risk type in the risk ranking results to form the dominant irreversible risk type corresponding to the current state.

[0007] In a preferred embodiment, the path filtering module includes: Read the dominant irreversible risk type, the set of etching sub-steps, and the sub-step sequence constraints. Expand the set of etching sub-steps sequentially according to the sub-step sequence constraints to form a candidate sequence set. Read the stage response data and adjacent time change data from the current state data, input each candidate sequence into the sequence generation rule, and determine whether the two adjacent sub-steps in each candidate sequence form a suppression relationship, an amplification relationship, or no effect on the dominant irreversible risk. If a suppression relationship is formed, write a suppression flag; if an amplification relationship is formed, write an amplification flag; otherwise, write a neutral flag to form a layer result corresponding to each candidate sequence.

[0008] In a preferred embodiment, the path filtering module further includes: Read the suppression, amplification, and neutral flags from the first-level results, and combine them with the continuous time-period change data in the current state data to progressively judge the sequence segment composed of three consecutive sub-steps in each candidate sequence. If the suppression relationship formed by the previous sub-step is canceled by the amplification relationship in the next sub-step, write the failure flag; if the suppression relationship formed by the previous sub-step continues to be maintained in the next sub-step, write the retention flag; otherwise, write the pending flag to form the second-level results corresponding to each candidate sequence. Read the step execution conditions from the second-level results and current status data, and judge each candidate order one by one. If there is a sub-step corresponding to the failure mark and the sub-step is restricted by the order of sub-steps and cannot be adjusted, write an exclusion mark. If there is no exclusion mark but there is a sub-step corresponding to the pending mark, write the sub-step to the postpone list. If there is no exclusion mark and no pending mark, write a retain mark to form the filtering results corresponding to each candidate order.

[0009] In a preferred embodiment, the path filtering module further includes: Read the filtering results, retain the candidate order written with the retention mark, and for the candidate order containing the delay list, move the sub-steps in the delay list to the order position after the corresponding release condition according to the order of sub-steps, and then re-execute the first-level judgment and the second-level judgment. If there is still an exclusion mark after re-judgment, delete the corresponding candidate order; otherwise, write the re-judged candidate order into the retention set. Read the candidate order in the hold set, sort them by the number of suppression markers from high to low, the number of amplification markers from low to high, and the number of sub-steps in the delay list from low to high. If the hold set is not empty, read the candidate order corresponding to the first sorted position as the target sub-step order. Otherwise, read the latest stage response data and re-execute the order expansion, judgment, and filtering.

[0010] In a preferred embodiment, the instruction generation module includes: Read the current status data and the target sub-step sequence, extract the status data and step position corresponding to each sub-step according to the target sub-step sequence, and form parameter data; Input the parameter data into the parameter generation rules, calculate the process parameters corresponding to each sub-step according to the target sub-step sequence, and form a parameter sequence. Read the parameter sequence and the target sub-step order, and combine the sub-step identifier with the corresponding process parameter according to the target sub-step order to form the control content corresponding to each sub-step; The control content corresponding to each sub-step is written continuously in the order of the target sub-steps to form control instructions corresponding to the order of the target sub-steps.

[0011] In a preferred embodiment, the process control module includes: Read the control instructions, write the target sub-step sequence and the corresponding process parameters of each sub-step into the etching equipment to form a sequence of steps to be executed; Read the sequence of steps to be executed, call the process parameters corresponding to each sub-step one by one in the order of the target sub-steps, control the etching equipment to perform the corresponding etching operations, and form the execution results corresponding to each sub-step. Read the execution results of each sub-step and the device data returned by the etching equipment. Extract the corresponding device data after the execution of each sub-step in the order of the target sub-steps to form the stage response data after the execution of each sub-step.

[0012] In a preferred embodiment, the cyclic update module includes: Read the stage response data and current status data, write the stage response data into the corresponding sub-step position in the current status data according to the target sub-step order, and replace the original status data at the corresponding position with the stage response data to form updated status data; The updated status data is input into the irreversible risk identification rules, the risk value corresponding to each irreversible risk is recalculated, and the dominant irreversible risk type is re-identified based on the results of sorting by risk value.

[0013] In a preferred embodiment, the cyclic update module includes: The updated target sub-step order is generated by updating the status data, re-identifying the dominant irreversible risk type, the set of etching sub-steps, and the input order generation rules that restrict the order of sub-steps. Read the execution status of the sub-steps corresponding to the updated target sub-step order. If there are unexecuted sub-steps, output the updated target sub-step order as the basis for continuing etching; if there are no unexecuted sub-steps, output the etching control result corresponding to the current batch.

[0014] The technical effects and advantages of this invention are as follows: 1. Based on the sequence of steps for identifying and rescreening irreversible risks, the risks amplified in the early stages can be suppressed in subsequent process control, thereby relatively alleviating the problem of continuous compression of subsequent repair space due to deviations in the early stages. 2. By performing aggregation, time alignment, completion, and interval sorting on online data, the current status data is kept consistent in terms of batch, chip, time, and sub-step position, thereby providing a unified data foundation for subsequent risk identification and sequential screening; 3. By performing first-level judgment, second-level judgment, delayed rearrangement and sorting on the candidate sub-steps, it is possible to filter out sequences that amplify the dominant irreversible risk or do not meet the execution conditions, thereby improving the adaptability of the target sequence. 4. By inputting the current status data and the target sub-step sequence into the parameter generation rules, the power, gas flow rate, chamber pressure and execution time are calculated in a linked manner, so that the adjustment of multiple process parameters and sequence remains consistent. 5. By writing control commands into the etching equipment and executing them step by step in the order of the target sub-steps, and then extracting the stage response data after the execution of each sub-step, the process control is based on the actual execution results, rather than relying solely on preset values. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the system modules of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Refer to the instruction manual appendix Figure 1 The LED chip etching processing system of the present invention with multi-process parameter collaborative optimization includes a data processing module, a risk identification module, a path screening module, an instruction generation module, a process control module, and a cyclic update module. The data processing module is used to read the chip identifier, etching sub-step set, sub-step sequence restrictions and online data collected by the etching equipment for the current batch, and perform unified time alignment processing on the online data to form the current status data; In this specific implementation, the data processing module organizes the online data generated by the etching equipment during the current batch processing into current status data that can be directly used for subsequent risk identification and path screening, so that the data read by subsequent modules are consistent in batch attribution, chip attribution, time base, and sub-step interval. In specific implementation, the online data is first collected according to the current batch and chip identifier, then the data at different sampling times are processed with unified time alignment, and on this basis, missing data is filled in and abnormal data is marked. Finally, the regularized data is divided into the corresponding sub-step intervals based on the etching sub-step set and the sub-step sequence restrictions to form the current status data. The implementation process includes the following steps: The system reads the chip identifier, etching sub-step set, sub-step sequence restrictions, and online data collected by the etching equipment for the current batch. It then aggregates the online data according to the current batch and chip identifier to form the raw data. Specifically, the chip identifier uses the wafer number, disk position number, or equipment station number within the current batch; the etching sub-step set uses the pre-processing, main etching, and over-etching sub-steps defined in the current recipe; and the sub-step sequence restrictions use the order of sub-steps in the recipe execution relationship table. After reading the online data, the data processing module matches the data from each acquisition channel using the current batch as the first aggregation condition and the chip identifier as the second aggregation condition. Successfully matched data is written to the data buffer of the corresponding chip, while data whose ownership cannot be confirmed is written to the exception record area and not merged into the raw data, thus ensuring that the ownership of the raw data is clearly defined. The data acquisition time corresponding to each data item in the raw data is extracted. Time mapping is performed on the data items with different acquisition times according to the unified sampling time to form aligned data. Specifically, the data processing module first counts the sampling period of each acquisition channel, and then uses the device controller clock period or the shortest sampling period in the current batch as the unified sampling time interval to generate a unified sampling time sequence covering the processing time of the current batch. For continuous data, the unified sampling time is written using the most recent time mapping or interpolation between adjacent two points. For switch quantities, step marker quantities, or alarm quantities, the unified sampling time is written using the most recent previous time retention method, thereby forming aligned data arranged according to the unified time. The process involves filling in missing data items in the aligned data and marking invalid data items that exceed the valid range, thus forming regularized data. Specifically, if a data item has valid values ​​at both the previous and next unified sampling times, interpolation is used to fill in the current missing position; if a valid value exists only at the previous unified sampling time, the previous value is used to continue filling; if the length of consecutive missing data exceeds the set upper limit, a missing value is written; subsequently, the valid range of each data item is checked according to the upper and lower limits of physical quantities given in the equipment manual, the operating range in the process formula, or the upper and lower limits obtained from the statistics of historical stable batches; data items that exceed the corresponding range are marked as invalid while retaining their original values, thereby forming regularized data. Based on the set of etching sub-steps and the sequence constraints of the sub-steps, the regularized data is divided into corresponding sub-step intervals and organized chronologically to form the current state data. Specifically, the data organization module reads the step start marker, step end marker, and recipe execution record from the regularized data to determine the start and end times of each sub-step. When there is a clear step switching signal, the switching signal is used directly as the interval boundary; when there is no clear step switching signal, the adjacent process parameter change point is used as the interval boundary. Then, the data whose acquisition time falls into the corresponding interval is written into the corresponding sub-step interval and arranged chronologically. The order of each interval is then checked according to the sequence constraints of the sub-steps. If the order is abnormal or the interval is missing, a step abnormality marker is written, and finally the current state data is formed. Through the above processing, the data processing module organizes scattered online data collected at different times, which may contain gaps or anomalies, into current status data organized by current batch, chip identifier, unified time base, and sub-step interval. This provides a unified data foundation for subsequent risk identification and path selection, and reduces identification bias caused by unclear data attribution, time misalignment, or outlier contamination. In practical applications: when a batch of LED chips enters the etching equipment, the data processing module synchronously reads the chip number corresponding to each chip in that batch, the pretreatment, main etching, and over-etching sub-step information in the current formula, as well as online data such as RF power, gas flow rate, cavity pressure, temperature, and step switching signals. First, it completes the aggregation by batch and chip number, then completes the time alignment according to a unified clock, performs short-term missing data completion, writes invalid markers for out-of-bounds data, and then divides the organized data into corresponding sub-step intervals according to the step switching signals, finally forming current status data, which is used by subsequent modules to continue to identify dominant irreversible risks and select target sub-step sequences.

[0018] The risk identification module is used to input the current status data into the identification rules corresponding to each irreversible risk, solve the risk value corresponding to each irreversible risk, and identify the dominant irreversible risk type according to the risk value sorting results. In this specific implementation, the risk identification module converts the current state data into the dominant irreversible risk type corresponding to the current moment, so that subsequent path selection revolves around the risk that needs to be suppressed first. In specific implementation, risk data is first extracted from the current state data according to the data items corresponding to each irreversible risk, and then the sub-risk values ​​of each irreversible risk are calculated according to the corresponding identification rules. Subsequently, the sub-risk values ​​are summarized to obtain the total risk value, and the dominant irreversible risk type is determined according to the size of the total risk value. The implementation process includes the following steps: The system reads the current status data and extracts the corresponding risk data from the data items corresponding to each irreversible risk to form the judgment data for each irreversible risk. Specifically, the data items corresponding to each irreversible risk are pre-written into the risk item configuration table during the process verification stage. For example, sidewall instability risk corresponds to cavity pressure, bias voltage, RF power, and temperature; residual accumulation risk corresponds to gas flow rate, optical detection value, cavity pressure, and execution time; and endpoint margin compression risk corresponds to endpoint detection value, current step execution time, and the end status of the preceding sub-step. The risk identification module reads the corresponding data items in the current status data according to the risk item configuration table and writes them into the corresponding judgment data in the order of uniform sampling time. When a data item has a missing or invalid mark, the mark is retained and written into the corresponding judgment data. The judgment data corresponding to each irreversible risk is input into the corresponding identification rules. The current value, the difference between adjacent time points, and the duration of continuous change in the judgment data are calculated to form the sub-risk value corresponding to each irreversible risk. Specifically, the current value is the value of the data item corresponding to the current unified sampling time, the difference between adjacent time points is the numerical difference between the current unified sampling time and the previous unified sampling time, and the duration of continuous change is the duration for which the data item continuously maintains the same direction of change or continuously meets the same risk condition. The risk identification module converts the current value, the difference between adjacent time points, and the duration of continuous change into the corresponding sub-risk value according to the identification rules corresponding to the irreversible risk. When a data item has a missing or invalid marker, it is processed according to the corresponding alternative caliber in the identification rules. The individual risk values ​​corresponding to each irreversible risk are summarized and calculated to form the total risk value for each irreversible risk. The risks are then ranked from highest to lowest according to the total risk value. Specifically, the individual risk values ​​under the same irreversible risk are summarized using a risk summary table. When using a weighted summation method, the individual risk values ​​corresponding to the current value, the difference between adjacent time points, and the duration of continuous change are multiplied by their respective weights and then summed to form the total risk value for that irreversible risk. After calculating the total risk value for each irreversible risk, they are ranked from highest to lowest. When two irreversible risks have the same total risk value, the individual risk values ​​corresponding to the duration of continuous change are compared first, and then the individual risk values ​​corresponding to the current value are compared to form the risk ranking result. The first risk type in the risk ranking result is read to form the dominant irreversible risk type corresponding to the current state. Specifically, the risk identification module reads the first risk type in the risk ranking result and writes the risk type into the risk result area corresponding to the current batch, the current chip, and the current unified sampling time. When the first risk types of two adjacent unified sampling times are different and the difference between their total risk values ​​is less than the switching threshold, the first risk type corresponding to the previous unified sampling time remains unchanged. When the difference between their total risk values ​​reaches the switching threshold, the first risk type corresponding to the current unified sampling time is output. Through the above processing, the risk identification module further organizes the current state data into risk results with a ranking relationship and outputs the dominant irreversible risk type corresponding to the current state, so that subsequent path selection revolves around the current dominant irreversible risk. In practical applications: after a certain LED chip completes the main etching, the risk identification module extracts the data items corresponding to the sidewall instability risk, residual accumulation risk, and endpoint margin compression risk from the current state data, and calculates the sub-risk values ​​and total risk values ​​of the three types of risks. If the total risk value corresponding to the sidewall instability risk is the highest, then the sidewall instability risk is determined as the dominant irreversible risk type corresponding to the current state, and is used by the subsequent path selection module to prioritize the target sub-step sequence that has a suppressive effect on the sidewall instability risk.

[0019] The path filtering module is used to generate rules for the dominant irreversible risk type, the set of etching sub-steps, and the input order of sub-steps, construct a set of candidate sub-step sequences that meet the execution conditions, and filter the target sub-step sequence from the set of candidate sub-step sequences. In this specific implementation, the path filtering module is used on the edge computing side to filter the target sub-step sequence for process control from the etching sub-step set according to the dominant irreversible risk type corresponding to the current state. In specific implementation, the etching sub-step set is first expanded sequentially according to the sub-step sequence constraints to form a candidate sub-step sequence set that satisfies the sequence constraints. Then, the phase response data, adjacent time change data, continuous time change data and step execution conditions in the current state data are combined to determine the effect of each candidate sub-step sequence on the dominant irreversible risk layer by layer. After that, the candidate sub-step sequence that does not meet the conditions is excluded, the adjustable candidate sub-step sequence is postponed and rearranged, and finally the target sub-step sequence is selected from the retained candidate sub-step sequence. The implementation process includes the following steps: The system reads the dominant irreversible risk type, the set of etch sub-steps, and the sub-step sequence constraints. It then sequentially expands the set of etch sub-steps according to these constraints to form a candidate sub-step sequence set. Specifically, the etch sub-step set uses all sub-steps in the current recipe, and the sub-step sequence constraints are based on the pre- and post-constraint relationships in the recipe execution relationship table. The edge computing unit first determines the starting sub-step that is not subject to the pre-constraints of other sub-steps, and then sequentially connects the sub-steps that satisfy the current sequence constraints to the end of the current sequence, forming a sequential expansion branch. If the connection of a sub-step causes the remaining sub-steps to no longer satisfy the sub-step sequence constraints, the sequential expansion branch is deleted. When all sub-steps have been connected and all sub-step positions satisfy the sub-step sequence constraints, the sub-step is arranged and written into the candidate sub-step sequence set. The system reads the stage response data and adjacent time-change data from the current state data, inputs each candidate sub-step into the sequence generation rule, and judges one by one whether the two adjacent sub-steps in each candidate sub-step sequence form a suppression relationship, an amplification relationship, or no effect on the dominant irreversible risk. If a suppression relationship is formed, a suppression flag is written; if an amplification relationship is formed, an amplification flag is written; otherwise, a neutral flag is written, thus forming a layer result corresponding to each candidate sub-step sequence. Specifically, the stage response data uses the state data after each sub-step is executed, and the adjacent time-change data uses the data difference between the current unified sampling time and the previous unified sampling time. The sequence generation rule pre-writes the relationship judgment conditions corresponding to the dominant irreversible risk type, and the relationship judgment conditions include at least suppression and amplification conditions. The edge computing unit reads the corresponding stage response data and adjacent time-change data of each pair of adjacent sub-steps in each candidate sub-step sequence and compares them with the relationship judgment conditions. When the suppression condition is met, a suppression flag is written; when the amplification condition is met, an amplification flag is written; when neither condition is met, a neutral flag is written, thus forming a layer result corresponding to each candidate sub-step sequence. The suppression, amplification, and neutral flags in the first-level result are read. Combined with the continuous time-period change data in the current state data, a progressive judgment is made on the sequential segment consisting of three consecutive sub-steps in each candidate sub-step sequence. If the suppression relationship formed by the previous sub-step is canceled out by the amplification relationship in the subsequent sub-step, a failure flag is written; if the suppression relationship formed by the previous sub-step continues in the subsequent sub-step, a retention flag is written; otherwise, a pending flag is written, forming the second-level result corresponding to each candidate sub-step sequence. Specifically, the continuous time-period change data uses the continuous change results of the same data item at multiple consecutive unified sampling times. The edge computing unit consists of three consecutive sub-steps forming a sequential segment. First, the markers of the first two sub-steps in the first-level result are read, followed by the markers of the last two sub-steps. If the first segment of the sequential segment is marked as a suppression marker and the second segment as an amplification marker, and the continuous time-period change data indicates that the change direction of the second segment is sufficient to offset the suppression effect of the first segment, a failure marker is written. If the first segment is marked as a suppression marker and the second segment is not marked as an amplification marker, and the continuous time-period change data indicates that the suppression effect of the first segment is still maintained in the second segment, a retention marker is written. In other cases, a pending marker is written, thus forming the second-level result corresponding to the sequence of each candidate sub-step. The system reads the execution conditions from the second-layer results and current state data, and checks the order of each candidate sub-step. If a sub-step with a failure flag exists and cannot be adjusted due to the order of sub-steps, an exclusion flag is written. If no exclusion flag exists but a sub-step with a pending flag exists, the sub-step is written to the postponement list. If neither an exclusion flag nor a pending flag exists, a retention flag is written, thus forming the filtering results corresponding to the order of each candidate sub-step. Specifically, the execution conditions include at least whether the sub-step has been executed, whether its position can be adjusted, and whether its current position must be maintained. The edge computing unit first checks whether a sub-step with a failure flag exists in the order of each candidate sub-step, and then determines whether the position of the sub-step can be adjusted under the order of sub-steps. When adjustment is not allowed, an exclusion flag is written to the corresponding candidate sub-step order. When adjustment is allowed and a sub-step with a pending flag exists, the sub-step is written to the postponement list. When neither an exclusion flag nor a pending flag exists, a retention flag is written to the corresponding candidate sub-step order, thus forming the filtering results corresponding to the order of each candidate sub-step. The filtering results are read, and the order of candidate sub-steps marked with retention marks is retained. For candidate sub-steps containing a delay list, the sub-steps in the delay list are moved to the order position after the corresponding release condition according to the sub-step order constraint. Then, the first-level and second-level judgments are re-executed. If an exclusion mark still exists after re-judgment, the corresponding candidate sub-step order is deleted; otherwise, the re-judged candidate sub-step order is written into the retention set. Specifically, the release condition is that all the preceding constraints of the sub-step are satisfied and the movement does not destroy the order constraints of the remaining sub-steps. The edge computing unit first directly writes the candidate sub-step order corresponding to the retention mark into the retention set. Then, it reads the sub-steps in the delay list one by one, moves the sub-step out of its original position and inserts it into the order position that meets the release condition to form the rearranged candidate sub-step order. After that, the first-level and second-level judgments are re-executed on the rearranged candidate sub-step order. If an exclusion mark still exists after re-judgment, the candidate sub-step order is deleted; if no exclusion mark exists after re-judgment, the rearranged candidate sub-step order is written into the retention set. The candidate sub-step sequence in the reserved set is read and sorted by the number of suppression markers (from high to low), the number of amplification markers (from low to high), and the number of sub-steps in the delayed list (from low to high). If the reserved set is not empty, the candidate sub-step sequence corresponding to the first position in the sorted set is read as the target sub-step sequence; otherwise, the latest stage response data is read and the sequence expansion, judgment, and filtering are re-executed. Specifically, the edge computing unit first counts the number of suppression markers, the number of amplification markers, and the number of sub-steps in the delayed list corresponding to each candidate sub-step sequence in the reserved set, and then sorts them in the order of priority: number of suppression markers, second priority: number of amplification markers, and third priority: number of sub-steps in the delayed list. When the reserved set is not empty, the candidate sub-step sequence corresponding to the first position in the sorted set is read as the target sub-step sequence and output to the subsequent instruction generation module for process control. When the reserved set is empty, it means that no candidate sub-step sequence that meets the filtering conditions has been formed in the current stage. At this time, the latest stage response data is read, and the sequence expansion, first-level judgment, second-level judgment, and filtering are re-executed until the target sub-step sequence is formed. Through the above processing, the path filtering module can sequentially expand, judge, filter, rearrange and sort the candidate sub-step sequence set according to the current dominant irreversible risk type on the edge computing side, and output the target sub-step sequence, so that the process control is adjusted around the current risk status, rather than fixedly using the original sub-step sequence. In practical applications: When the dominant irreversible risk type is identified as sidewall instability risk, the edge computing unit first forms a candidate sub-step sequence set based on the etching sub-step set corresponding to preprocessing, main etching, and over-etching, as well as the sub-step sequence constraints. Then, based on the stage response data, adjacent time change data, and continuous time change data in the current state data, it judges the suppression, amplification, and neutral relationships of each candidate sub-step sequence on the sidewall instability risk. Candidate sub-step sequences with failure markers that cannot be adjusted are directly excluded. Sub-step sequences with pending markers that can be adjusted are postponed and rearranged. Finally, the candidate sub-step sequence with the most suppression markers, the fewest amplification markers, and the fewest sub-steps in the postponed list is selected from the retained set as the target sub-step sequence corresponding to the current batch.

[0020] The instruction generation module is used to input the current status data and the target sub-step sequence into the parameter generation rules, calculate the process parameters corresponding to each sub-step in the target sub-step sequence, and form control instructions corresponding to the target sub-step sequence. In this specific embodiment, the function of the instruction generation module is to convert the current state data and the target sub-step sequence into control instructions that the etching equipment can directly execute, so that the target sub-step sequence output by the path filtering module can be further applied to the corresponding process parameters and execution content. In specific implementation, firstly, the state data and step position corresponding to each sub-step are extracted from the current state data according to the target sub-step sequence to form parameter data; then, the corresponding process parameters are calculated for each sub-step according to the parameter generation rules to form a parameter sequence; then, the identifier of each sub-step is combined with the corresponding process parameters to form the control content corresponding to each sub-step; finally, the control content corresponding to each sub-step is continuously written according to the target sub-step sequence to form control instructions corresponding to the target sub-step sequence. The implementation process includes the following steps: The system reads the current status data and the target sub-step sequence, extracts the status data and step position corresponding to each sub-step according to the target sub-step sequence, and forms parameter data. Specifically, the current status data is the status data of the current batch after being output by the data processing module and filtered by risk identification and path selection. The target sub-step sequence is the target sub-step sequence output by the path selection module. The instruction generation module reads the status data corresponding to each sub-step one by one according to the target sub-step sequence, and simultaneously reads the step position of the sub-step in the target sub-step sequence. The status data and step position of the same sub-step are written into the same parameter record to form parameter data. The status data here includes at least the existing results in the stage response data, adjacent time change data, or continuous time change data corresponding to the sub-step. The step position is used to indicate the order of the sub-step in the current target sub-step sequence to ensure that the subsequent process parameter calculation is consistent with the sequence position. The parameter data is input into the parameter generation rules, and the process parameters corresponding to each sub-step are calculated according to the target sub-step order to form a parameter sequence. Specifically, the parameter generation rules adopt the calculation relationships determined and written into the rule table during the process verification stage. The rule table records the process parameter values ​​corresponding to different sub-steps under different state data. The instruction generation module reads the parameter data one by one according to the target sub-step order, calls the corresponding parameter generation rules for each sub-step, and solves the process parameters corresponding to that sub-step. The process parameters include at least one or more of the following: power, gas flow rate, chamber pressure, execution time, or temperature control value corresponding to the execution of the sub-step. When a sub-step corresponds to multiple process parameters, they are written in the order of the parameter fields of that sub-step. When there are invalid markers in the parameter data, the corresponding process parameters are obtained according to the alternative values ​​in the parameter generation rules. After all sub-steps are calculated, the process parameters corresponding to each sub-step are arranged continuously according to the target sub-step order to form a parameter sequence. The parameter sequence and the target sub-step sequence are read. Each sub-step identifier is combined with its corresponding process parameter according to the target sub-step sequence to form the control content for each sub-step. Specifically, the instruction generation module reads the process parameters in the parameter sequence item by item according to the target sub-step sequence, and reads the sub-step identifier corresponding to the same sequential position. The sub-step identifier and the process parameter at that sequential position are written into the same control item to form the control content for that sub-step. Each control content includes at least a sub-step identifier, a sequential position, and process parameters. The sequential position indicates the execution order of the control content within the overall control instruction, and the process parameters indicate the specific control values ​​when the sub-step is executed. After this processing, the control content of each sub-step can correspond item by item with the target sub-step sequence. The control content corresponding to each sub-step is continuously written in the order of the target sub-steps to form control instructions corresponding to the order of the target sub-steps. Specifically, the instruction generation module reads the control content corresponding to each sub-step from front to back in the order of the target sub-steps and writes it continuously according to the data format of the instructions received by the etching equipment. After the control content corresponding to the previous sub-step is written, the control content corresponding to the next sub-step is written until all control content is written, forming control instructions corresponding to the order of the target sub-steps. The order of each sub-step in this control instruction is consistent with the order of the target sub-steps, and the process parameters of each sub-step are consistent with the calculation results in the parameter sequence, thereby ensuring that the subsequent process control module can directly drive the etching equipment to execute the etching process of the current batch according to the control instructions. Through the above processing, the instruction generation module further converts the current status data and the target sub-step sequence into control instructions that correspond one-to-one with the target sub-step sequence, so that the subsequent process control module can directly call the process parameters of the corresponding sub-step to perform etching, and ensure that the correspondence between the sub-step sequence and the process parameters is not misaligned. In practical applications: When the path filtering module outputs the target sub-step sequence corresponding to the current batch of a certain LED chip, the instruction generation module first extracts the status data and step positions corresponding to the preprocessing sub-step, main etching sub-step, and over-etching sub-step according to the target sub-step sequence. Then, according to the parameter generation rules, it calculates the power, gas flow rate, cavity pressure, and execution time corresponding to each sub-step to form a parameter sequence. After that, the identifier of each sub-step is combined with the corresponding process parameters to form control content, and written continuously according to the target sub-step sequence to form control instructions that can be directly sent to the etching equipment for subsequent process control modules to call.

[0021] The process control module is used to send control commands to the etching equipment, control the etching equipment to perform etching according to the target sub-step sequence and process parameters, and collect stage response data after each sub-step is executed; In this specific embodiment, the process control module's function is to implement the control commands output by the command generation module into the actual execution process of the etching equipment, and to collect the corresponding stage response data after each sub-step is executed, so that the subsequent cycle update module can continue to adjust the subsequent etching process based on the actual execution results. In specific implementation, the control commands are first read and the target sub-step sequence and the corresponding process parameters of each sub-step in the control commands are written into the etching equipment to form a sequence of steps to be executed; then, the process parameters corresponding to each sub-step are called one by one according to the target sub-step sequence to control the etching equipment to execute the corresponding etching operation, forming the execution result corresponding to each sub-step; then, the execution result corresponding to each sub-step and the equipment data returned by the etching equipment are read, and the corresponding equipment data after the execution of each sub-step is extracted according to the target sub-step sequence to form stage response data. The implementation process includes the following steps: The process control module reads control instructions and writes the target sub-step sequence and corresponding process parameters of each sub-step into the etching equipment to form a sequence of steps to be executed. Specifically, the control instructions are output by the instruction generation module and include at least the target sub-step sequence, the identifier of each sub-step, and the corresponding process parameters. The process control module first performs format verification on the control instructions, checking whether the target sub-step sequence is complete, whether each sub-step identifier corresponds one-to-one with the sub-steps in the current recipe of the etching equipment, and whether the corresponding process parameter fields are complete. After the format verification is passed, the process control module writes the sub-step identifier, sequence position, and process parameters from the control instructions into the execution buffer of the etching equipment item by item, and forms a sequence of steps to be executed according to the target sub-step sequence. If there are missing fields in the control instructions or the sub-step sequence does not match the current recipe, the writing stops and an exception record is output, and the subsequent execution process is not entered. The process control module reads the sequence of steps to be executed, calls the corresponding process parameters for each sub-step in the order of the target sub-steps, and controls the etching equipment to perform the corresponding etching operations, forming the execution results for each sub-step. Specifically, starting from the first step in the sequence of steps to be executed, the process control module reads the corresponding sub-step identifier and process parameters one by one in the order of the target sub-steps, and sends the process parameters to the corresponding execution interface of the etching equipment. After receiving the process parameters corresponding to the current sub-step, the etching equipment calls the corresponding power control unit, gas control unit, pressure control unit, and time control unit to execute the etching operation corresponding to the current sub-step. After the current sub-step is completed, the etching equipment returns the execution end signal and execution status corresponding to the sub-step, and the process control module writes the execution end signal and execution status into the execution result corresponding to the sub-step. If the current sub-step does not return the execution end signal, the current sub-step remains in the execution state and the next sub-step is not called until the current sub-step is completed or the equipment abnormality is triggered. The process control module reads the execution results of each sub-step and the equipment data returned by the etching equipment, and extracts the corresponding equipment data after the execution of each sub-step in the order of the target sub-steps to form the stage response data after the execution of each sub-step. Specifically, after the execution of each sub-step, the process control module reads the execution result corresponding to that sub-step and the equipment data returned by the etching equipment during and at the end of the execution of that sub-step. The equipment data includes at least one or more of the following after the execution of the sub-step: power feedback value, gas flow feedback value, cavity pressure feedback value, temperature feedback value, bias feedback value, execution time, and step completion status. The process control module extracts the corresponding equipment data after the execution of each sub-step one by one in the order of the target sub-steps, and writes the equipment data, along with the corresponding sub-step identifier and sequence position, into the same stage record to form the stage response data after the execution of each sub-step. When the equipment data corresponding to a certain sub-step is missing, the missing marker is retained and written into the stage response data for processing by the subsequent cycle update module. Through the above processing, the process control module completes the conversion of control commands into actions to be performed by the etching equipment, and generates stage response data that is consistent with the sequence of the target sub-steps after each sub-step is executed, so that subsequent cyclic updates can continue based on actual execution results rather than based on preset values. In practical applications: When the instruction generation module outputs control instructions for a specific batch of LED chips, including pre-processing sub-steps, main etching sub-steps, and over-etching sub-steps, the process control module first writes the identifier, sequence position, and corresponding power, gas flow rate, cavity pressure, and execution time of each sub-step into the execution buffer of the etching equipment, forming a sequence of steps to be executed. Then, it calls the corresponding process parameters one by one according to the target sub-step sequence to control the etching equipment to perform pre-processing, main etching, and over-etching operations. After each sub-step is completed, it reads the execution status and power feedback value, flow rate feedback value, cavity pressure feedback value, and temperature feedback value returned by the equipment, ultimately forming stage response data that corresponds one-to-one with the target sub-step sequence, which is then used by the cyclic update module.

[0022] The cyclic update module is used to merge the stage response data with the current status data to form updated status data. Based on the updated status data, it re-identifies the dominant irreversible risk type and re-selects the target sub-step sequence until the etching control result corresponding to the current batch is formed. In this specific implementation, the cyclic update module's function is to write back the stage response data output by the process control module to the current state data, and based on the updated state data, re-identify the dominant irreversible risk type and re-select the target sub-step sequence, so that the subsequent etching process can continuously adjust according to the actual execution results, rather than fixedly following the previous control path. In specific implementation, firstly, the stage response data is written into the corresponding sub-step position in the current state data according to the target sub-step sequence to form updated state data; then, the updated state data is input into the irreversible risk identification rule to recalculate the risk value corresponding to each irreversible risk and re-identify the dominant irreversible risk type; then, the updated state data, the re-identified dominant irreversible risk type, the etching sub-step set, and the sub-step sequence restriction input order generation rule are used to re-select the target sub-step sequence; finally, based on the sub-step execution status corresponding to the updated target sub-step sequence, it is determined whether to output the basis for continuing etching or output the etching control result corresponding to the current batch. The implementation process includes the following steps: The process reads stage response data and current status data, writes the stage response data into the corresponding sub-step positions in the current status data according to the target sub-step order, and replaces the original status data at the corresponding positions with the stage response data to form updated status data. Specifically, the stage response data uses data collected by the process control module after each sub-step is executed, and the current status data uses the status data used in the previous round of risk identification and path screening. The loop update module first reads the stage response data corresponding to each sub-step in the target sub-step order, then searches the current status data for data intervals with the same sub-step identifier and the same order position, and replaces the original status data in the data interval with the corresponding stage response data. If a sub-step has not been executed in the current round, the original status data of that sub-step in the current status data is retained unchanged. If the stage response data corresponding to a sub-step has a missing or invalid flag, it is written into the corresponding sub-step position along with the flag. After all sub-steps are processed, updated status data is formed, so that the updated status data simultaneously includes the actual response results of executed sub-steps and the original status data of unexecuted sub-steps. The updated status data is input into the irreversible risk identification rules to recalculate the risk value corresponding to each irreversible risk. The dominant irreversible risk type is then re-identified based on the risk value sorting results. Specifically, the cyclic update module calls the same irreversible risk identification rules as the risk identification module to re-extract, calculate, and summarize each risk data in the updated status data. The risk value is still summarized based on the current value, the difference between adjacent time points, and the sub-risk values ​​corresponding to the continuous change duration. Since the data corresponding to the executed sub-steps has been replaced by the stage response data, the recalculated risk value reflects the actual risk status of the current batch after this round of execution. After completing the calculation of the risk value corresponding to each irreversible risk, the risk values ​​are sorted from high to low, and the risk type corresponding to the first position in the sort is read as the re-identified dominant irreversible risk type. The updated target sub-step order is generated by updating the state data, re-identifying the dominant irreversible risk type, the etch sub-step set, and the input order generation rules for sub-step sequence constraints. Specifically, the loop update module calls the same sequence generation rules as the path filtering module. First, it re-expands the candidate sub-step order based on the etch sub-step set and sub-step sequence constraints. Then, it combines the actual stage response results reflected in the updated state data to re-perform a first-level judgment, a second-level judgment, filtering, delayed rearrangement, and sorting for each candidate sub-step order. If a candidate sub-step order still meets the filtering conditions under the updated state data, it is retained. If the original target sub-step order no longer meets the filtering conditions under the updated state data, it is replaced with a new candidate sub-step order according to the re-filtering results. After sorting, the candidate sub-step order corresponding to the first position in the sort is read to form the updated target sub-step order. The update module reads the execution status of each sub-step corresponding to the updated target sub-step sequence. If there are any unexecuted sub-steps, the updated target sub-step sequence is output as the basis for continuing etching. If there are no unexecuted sub-steps, the etching control result corresponding to the current batch is output. Specifically, the loop update module reads the execution status of each sub-step one by one according to the updated target sub-step sequence. When at least one sub-step is not executed, it means that the etching process of the current batch is not yet complete. At this time, the updated target sub-step sequence is output to the instruction generation module and the process control module as the basis for continuing etching in the next round. When all sub-steps in the updated target sub-step sequence have been executed, it means that the etching process of the current batch is complete. At this time, the etching control result corresponding to the current batch is output. The etching control result here includes at least the final target sub-step sequence corresponding to the current batch, the execution status of each sub-step, and the stage response data corresponding to each sub-step, which is used to characterize the final execution status of the etching control of this batch. Through the above processing, the cyclic update module can continuously write the stage response data back to the current status data, and re-identify the dominant irreversible risk type and re-select the target sub-step sequence based on the updated actual status, so that the subsequent control path is adjusted synchronously with the actual execution results until the etching control result corresponding to the current batch is formed. In practical applications: After a batch of LED chips completes the preprocessing sub-step and the main etching sub-step, the cyclic update module first replaces the stage response data corresponding to these two sub-steps with the corresponding positions in the current state data to form updated state data; then, it recalculates the risk values ​​corresponding to sidewall instability risk, residual accumulation risk, and endpoint margin compression risk, and re-identifies the dominant irreversible risk type of the current stage; then, based on the updated state data and the re-identified dominant irreversible risk type, it re-screens the remaining unexecuted over-etching sub-steps and their related sequences to form the updated target sub-step sequence; if there are still unexecuted sub-steps in the updated target sub-step sequence, the sequence is output as the basis for continuing etching; if all sub-steps have been executed, the etching control result corresponding to this batch is output.

[0023] Working Principle: This solution, under an edge computing architecture, organizes online data during the LED chip etching process and identifies the most critical irreversible risk requiring priority handling based on the organized current state data. This drives subsequent process control and multi-process parameter collaborative optimization. After identifying the critical irreversible risk, the system does not continue to execute according to a fixed process sequence. Instead, it combines the set of etching sub-steps and the constraints on the order of sub-steps to reselect a more suitable target sub-step sequence for the current state. Subsequently, based on the target sub-step sequence and the current state data, it performs linked calculations on multiple process parameters such as power, gas flow rate, cavity pressure, and execution time, generating control commands corresponding to the target sub-step sequence and sending them to the etching equipment for execution. The stage response data obtained after the etching equipment executes is written back to the current state data. The edge computing side then re-identifies the critical irreversible risk, re-selects the target sub-step sequence, and re-optimizes multiple process parameters based on the updated actual state until all sub-steps are completed, outputting the etching control result corresponding to the current batch. In other words, this solution integrates risk identification, sequence adjustment, and multi-process parameter collaborative optimization into the same process control chain during the LED chip etching process and executes them continuously. For example, in a continuous mass production scenario for LED chips, after the main etching of a chip, the edge computing unit reads the stage response data returned by the equipment and finds that the risk of sidewall instability is higher than the risk of residual accumulation and the risk of end-point margin compression in the current state. The process control then no longer mechanically follows the original sub-step sequence, but prioritizes selecting the target sub-step sequence that suppresses the risk of sidewall instability. Multiple process parameters, such as power, gas flow rate, cavity pressure, and execution time, are then adjusted collaboratively around this target sub-step sequence. New control commands are then issued to the etching equipment for continued execution. If subsequent execution results indicate that the risk focus shifts to the risk of end-point margin compression, the edge computing side will again adjust the target sub-step sequence based on the updated state data and simultaneously recalculate multiple process parameters corresponding to this sequence, ensuring that subsequent process control continues to revolve around the new dominant irreversible risk. In practical applications, this solution can leverage edge computing to achieve collaborative optimization of multiple process parameters and dynamic process control during the LED chip etching process, ensuring that the subsequent execution path and its corresponding parameters for each chip remain consistent with the actual processing state.

[0024] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An LED chip etching system with multi-process parameter collaborative optimization, including a data processing module, a risk identification module, a path selection module, an instruction generation module, a process control module, and a cyclic update module; Its features are: The data processing module is used to read the chip identifier, etching sub-step set, sub-step sequence restrictions and online data collected by the etching equipment for the current batch, and perform unified time alignment processing on the online data to form the current status data; The risk identification module is used to input the current status data into the identification rules corresponding to each irreversible risk, solve the risk value corresponding to each irreversible risk, and identify the dominant irreversible risk type according to the risk value sorting results. The path filtering module is used to generate rules for the dominant irreversible risk type, the set of etching sub-steps, and the input order of sub-steps, construct a set of candidate sub-step sequences that meet the execution conditions, and filter the target sub-step sequence from the set of candidate sub-step sequences. Read the dominant irreversible risk type, the set of etching sub-steps, and the sub-step sequence constraints. Expand the set of etching sub-steps sequentially according to the sub-step sequence constraints to form a candidate sequence set. Read the stage response data and adjacent time change data from the current state data, input each candidate sequence into the sequence generation rule, and determine one by one whether the two adjacent sub-steps in each candidate sequence form a suppression relationship, an amplification relationship, or no effect on the dominant irreversible risk. If a suppression relationship is formed, write a suppression flag; if an amplification relationship is formed, write an amplification flag; otherwise, write a neutral flag to form a layer result corresponding to each candidate sequence. Read the suppression, amplification, and neutral flags from the first-level results, and combine them with the continuous time-period change data in the current state data to progressively judge the sequence segment composed of three consecutive sub-steps in each candidate sequence. If the suppression relationship formed by the previous sub-step is canceled by the amplification relationship in the next sub-step, write the failure flag; if the suppression relationship formed by the previous sub-step continues to be maintained in the next sub-step, write the retention flag; otherwise, write the pending flag to form the second-level results corresponding to each candidate sequence. Read the step execution conditions from the second-level results and current status data, and judge each candidate order one by one. If there is a sub-step corresponding to the failure mark and the sub-step is restricted by the order of sub-steps and cannot be adjusted, write the exclusion mark. If there is no exclusion mark but there is a sub-step corresponding to the pending mark, write the sub-step to the postpone list. If there is no exclusion mark and no pending mark, write the retention mark to form the filtering results corresponding to each candidate order. The instruction generation module is used to input the current status data and the target sub-step sequence into the parameter generation rules, calculate the process parameters corresponding to each sub-step in the target sub-step sequence, and form control instructions corresponding to the target sub-step sequence. The process control module is used to send control commands to the etching equipment, control the etching equipment to perform etching according to the target sub-step sequence and process parameters, and collect stage response data after each sub-step is executed; The cyclic update module is used to merge the stage response data with the current status data to form updated status data. Based on the updated status data, the dominant irreversible risk type is re-identified and the target sub-step sequence is re-selected until the etching control result corresponding to the current batch is formed.

2. The LED chip etching system with multi-process parameter synergistic optimization according to claim 1, characterized in that: The data processing module includes: Read the chip identifier, etching sub-step set, sub-step sequence restrictions, and online data collected by the etching equipment for the current batch; and aggregate the online data according to the current batch and chip identifier to form the raw data. Extract the collection time corresponding to each data item in the raw data, and perform time mapping on data items with different collection times according to a unified sampling time to form aligned data; The missing data items in the aligned data are padded, and the data items that are out of the valid range are marked as invalid, thus forming regularized data. Based on the set of etching sub-steps and the constraints on the order of sub-steps, the regularized data is divided into the corresponding sub-step intervals and organized in chronological order to form the current state data.

3. The LED chip etching system with multi-process parameter synergistic optimization according to claim 2, characterized in that: The risk identification module includes: Read the current status data, extract the corresponding risk data from the current status data according to the data items corresponding to each irreversible risk, and form the judgment data corresponding to each irreversible risk; Input the judgment data corresponding to each irreversible risk into the corresponding identification rules, calculate the current value, the difference between adjacent time points and the duration of continuous change in the judgment data, and form the sub-risk value corresponding to each irreversible risk. The individual risk values ​​corresponding to each irreversible risk are summarized and calculated to form the total risk value corresponding to each irreversible risk, and the risk is ranked from high to low according to the total risk value. Read the first risk type in the risk ranking results to form the dominant irreversible risk type corresponding to the current state.

4. The LED chip etching system with multi-process parameter synergistic optimization according to claim 3, characterized in that: The path filtering module also includes: Read the filtering results, retain the candidate order written with the retention mark, and for the candidate order containing the delay list, move the sub-steps in the delay list to the order position after the corresponding release condition according to the order of sub-steps, and then re-execute the first-level judgment and the second-level judgment. If there is still an exclusion mark after re-judgment, delete the corresponding candidate order; otherwise, write the re-judged candidate order into the retention set. Read the candidate order in the hold set, sort them by the number of suppression markers from high to low, the number of amplification markers from low to high, and the number of sub-steps in the delay list from low to high. If the hold set is not empty, read the candidate order corresponding to the first sorted position as the target sub-step order. Otherwise, read the latest stage response data and re-execute the order expansion, judgment, and filtering.

5. The LED chip etching system with multi-process parameter synergistic optimization according to claim 4, characterized in that: The instruction generation module includes: Read the current status data and the target sub-step sequence, extract the status data and step position corresponding to each sub-step according to the target sub-step sequence, and form parameter data; Input the parameter data into the parameter generation rules, calculate the process parameters corresponding to each sub-step according to the target sub-step sequence, and form a parameter sequence. Read the parameter sequence and the target sub-step order, and combine the sub-step identifier with the corresponding process parameter according to the target sub-step order to form the control content corresponding to each sub-step; The control content corresponding to each sub-step is written continuously in the order of the target sub-steps to form control instructions corresponding to the order of the target sub-steps.

6. The LED chip etching system with multi-process parameter synergistic optimization according to claim 5, characterized in that: The process control module includes: Read the control instructions, write the target sub-step sequence and the corresponding process parameters of each sub-step into the etching equipment to form a sequence of steps to be executed; Read the sequence of steps to be executed, call the process parameters corresponding to each sub-step one by one in the order of the target sub-steps, control the etching equipment to perform the corresponding etching operations, and form the execution results corresponding to each sub-step. Read the execution results of each sub-step and the device data returned by the etching equipment. Extract the corresponding device data after the execution of each sub-step in the order of the target sub-steps to form the stage response data after the execution of each sub-step.

7. The LED chip etching system with multi-process parameter synergistic optimization according to claim 6, characterized in that: The cyclic update module includes: Read the stage response data and current status data, write the stage response data into the corresponding sub-step position in the current status data according to the target sub-step order, and replace the original status data at the corresponding position with the stage response data to form updated status data; The updated status data is input into the irreversible risk identification rules, the risk value corresponding to each irreversible risk is recalculated, and the dominant irreversible risk type is re-identified based on the results of sorting by risk value.

8. The LED chip etching system with multi-process parameter synergistic optimization according to claim 7, characterized in that: The cyclic update module includes: The updated target sub-step order is generated by updating the status data, re-identifying the dominant irreversible risk type, the set of etching sub-steps, and the input order generation rules that restrict the order of sub-steps. Read the execution status of the sub-steps corresponding to the updated target sub-step order. If there are unexecuted sub-steps, output the updated target sub-step order as the basis for continuing etching; if there are no unexecuted sub-steps, output the etching control result corresponding to the current batch.

Images