Distributed wafer automatic cutting system and method

Abstract

The invention relates to a distributed wafer automatic cutting system and method, and belongs to the technical field of semiconductor chip packaging, and the system comprises a scribing machine which is used for receiving control data sent by an industrial personal computer, controlling a scribing mechanism to execute wafer scribing, and collecting sensing data of the scribing mechanism through a first data collection module; the splitting machine is used for receiving the control data issued by the industrial personal computer, controlling a chopper mechanism to perform wafer splitting, and acquiring sensing data of the chopper mechanism through a second data acquisition module; the industrial switch is used for distributed data exchange and transmission; an industrial personal computer; and the controller is used for receiving and storing the sensing data, storing formula parameters and a mapping rule base, issuing control data to the scribing machine, the splitting machine and the transmitter, and realizing automatic closed-loop wafer cutting of'perception-decision-execution '. The linkage processing technology is achieved through technology data sharing, the yield and the production efficiency are improved, the production cost is reduced, and the technology stability and efficiency are improved.

Description

Technical Field

[0001] This invention belongs to the field of semiconductor chip packaging technology. Specifically, this invention relates to a distributed automated wafer dicing system, method, and processor-readable storage medium. Background Technology

[0002] In semiconductor chip manufacturing, wafer dicing and singulation are two key steps in wafer cutting. Dicing involves using a laser or blade to create dicing lines on the wafer, preparing it for chip separation. Singulation is a process that applies stress along the dicing lines, using mechanical stress or lasers to completely separate the wafer into individual chips. It is an essential step before packaging, and the quality of singulation directly determines packaging yield and chip reliability.

[0003] Dicing and dicing together constitute the complete wafer dicing process and are key steps before packaging. Current technologies primarily use two different, independent pieces of equipment—dicing machines and dicing machines—to process wafers independently, which presents the following structural and technical drawbacks:

[0004] Process silos and information gaps: There is a lack of effective data interaction between the dicing machine and the dicing machine, and key quality information of the dicing process (such as cutting depth and blade status) cannot be transmitted to the dicing process in real time;

[0005] Highly dependent on human experience: The cleaving parameters rely heavily on the operator's subjective experience and manual adjustments, leading to unstable processes. The parameter settings set by different operators can vary by more than 20%, resulting in high personnel training costs.

[0006] Long process debugging cycle: When changing products or wafer batches, it takes several hours or a day to readjust the parameters, which seriously affects the production schedule;

[0007] Quality risks are difficult to control: Minor abnormalities in the dicing process cannot be detected and compensated for in time by the slicing process, which can easily lead to batch quality accidents and high costs.

[0008] In the existing technology, wafer dicing systems mainly refer to dicing machines. There is an urgent need in the field for a distributed automatic wafer dicing system and a distributed automatic wafer dicing method, which can realize the linkage processing of dicing machines and dicing machines through process data sharing, so as to improve yield and production efficiency and reduce production costs. Summary of the Invention

[0009] This invention patent proposes a distributed automatic wafer dicing system and a distributed automatic wafer dicing method. By sharing process data, it realizes a coordinated processing process, thereby improving yield and production efficiency, reducing production costs, and has high practical value.

[0010] In a first aspect, the present invention provides a distributed automated wafer dicing system, comprising:

[0011] The dicing machine is electrically connected to an industrial switch and is used to receive control data from an industrial computer via a servo drive system to control the dicing mechanism to perform wafer dicing. The first data acquisition module collects the sensor data of the dicing mechanism.

[0012] The wafer dicing machine is electrically connected to an industrial switch and is used to receive control data from an industrial computer via a servo drive system to control the dicing mechanism to perform wafer dicing. The second data acquisition module collects the sensor data of the dicing mechanism.

[0013] Industrial switches are used for distributed data exchange and transmission.

[0014] An industrial control computer is electrically connected to an industrial switch; it is used to receive and store the sensor data, store the recipe parameters and the mapping rule base Y; and it is used to send control data to the dicing machine, dicing machine and transmitter to realize automated closed-loop wafer dicing of "perception-decision-execution".

[0015] Furthermore, the transmitter, electrically connected to an industrial switch, is used to transfer wafers from the dicing mechanism of the dicing machine to the cleaving mechanism of the dicing machine, and to transfer the cleaved wafers to the wafer collection mechanism.

[0016] Wafer collection mechanism, including qualified wafer collectors and unqualified wafer collectors;

[0017] The first data acquisition module is used to acquire the sensing data of the dicing mechanism in real time, including: a high-precision circuit sensor and a first machine vision unit;

[0018] The high-precision circuit sensor is installed in the power supply circuit of the spindle motor of the dicing mechanism to collect the current data of the dicing mechanism and monitor the cutting load and blade wear status.

[0019] The first machine vision unit includes a CCD camera and a coaxial LED cold light source, and is set to face the wafer dicing operation table of the dicing mechanism to capture sensor data of wafer dicing marks.

[0020] The second data acquisition module is used to acquire sensor data of the chopping mechanism in real time, including: a second machine vision unit;

[0021] The second machine vision unit, including a CCD camera and a coaxial LED cold light source, is positioned directly opposite the cleaving mechanism operating table and is used to capture sensor data of wafer images.

[0022] Furthermore, the industrial control computer is equipped with a qualification detection module, which is used to identify whether the edges of the cleaved wafer are regular, neat, and undamaged based on the sensor data of the second data acquisition module. If so, it is judged as a qualified wafer and sent to the qualified wafer collector via the transmitter; if not, it is judged as a defective wafer and sent to the defective wafer collector via the transmitter. The wafer cleaving qualification rate data is statistically analyzed, and the optimal cleaving parameter Zy and compensation coefficient K of this batch are correlated and recorded with the cleaving yield data.

[0023] The industrial control computer is equipped with a database, including a time-series database and a relational database; the time-series database is used to store sensor data, and the relational database is used to store the recipe parameters and the mapping rule base Y.

[0024] The industrial control computer determines the optimal dicing parameters Zy based on the mapping rule base Y, the physical depth Wd of the wafer cleavage, and the blade health index HI, and then sends the optimal dicing parameters Zy to the dicing machine to perform wafer dicing.

[0025] Furthermore, the mapping rule base Y is used to query preset mapping parameters, which include:

[0026] Preset dicing depth Wd0, preset blade health index HI0, and preset dicing parameter Pi;

[0027] Wherein, the preset cleaving parameters Pi={P1, P2, P3}, P1 is the preset cleaving pressure, P2 is the preset cleaving speed, P3 is the preset support platform spacing, and i is the parameter number.

[0028] Furthermore, the physical depth Wd of the wafer grommets is calculated in real time based on the wafer grommets image sensing data from the first data acquisition module using an image edge detection algorithm. The specific algorithm for Wd is as follows:

[0029] Wd= Sd × Q (Formula 1);

[0030] Where Q = E / e; Sd is the pixel depth of the tool mark, obtained from the image sensing data of the machine vision module through the image edge detection algorithm; Q is the wafer tool mark image calibration coefficient, obtained through camera calibration; E is the actual physical size; and e is the pixel size.

[0031] Furthermore, the blade health index HI is obtained through real-time calculation based on the current data of the dicing mechanism. The specific algorithm for HI is as follows:

[0032] HI = |I_avg - I_new| / I_new (Formula 2);

[0033] Among them, I_avg is the average current value of each cutting cycle, and I_new is the initial reference current value after replacing the new blade.

[0034] Further, when HI > M1, it is determined that the blade is suddenly damaged, and the wafer dicing is immediately stopped and an alarm is given;

[0035] When HI ≤ M0, Pi is queried and determined in the mapping rule library Y, and Pi is set as the optimal chipping parameter Zy: Zy = Pi = {P1, P2, P3}; the optimal chipping parameter Zy is sent to the dicing machine and the wafer chipping is executed;

[0036] When M0 < HI ≤ M1, it is determined that the blade is significantly worn, and the new execution parameter Pi_new is obtained through the adaptive compensation algorithm, and Pi_new is set as the optimal chipping parameter Zy: Zy = Pi_new = {P1_new, P2_new, P3_new}; the optimal chipping parameter Zy is sent to the dicing machine and the wafer chipping is executed;

[0037] Among them, M1 is the damage warning value; M0 is the wear threshold; the wear threshold M0 and the damage warning value M1 are determined by the actual production process; P1_new is the new execution chipping pressure, P2_new is the new execution chipping speed, and P3_new is the new execution support table spacing.

[0038] Further, the new execution parameter Pi_new is determined by the following formula:

[0039] Pi_new = Pi × (1 + K × HI) (Formula 3).

[0040] Among them, K is the compensation coefficient, and the value range is from 0.1 to 0.5. When the wafer is a gallium arsenide GaAs wafer, K is 0.3, and when the wafer is an indium phosphide InP wafer, K is 0.2, which is determined by material tests;

[0041] The value of Pi is queried and determined in the mapping rule library Y based on the physical depth Wd of the wafer knife mark and the blade health index HI value.

[0042] In the second aspect of the present invention, a distributed wafer automatic cutting method is proposed, which is applied to the foregoing distributed wafer automatic cutting system, and is characterized by including the following steps:

[0043] S1, based on the wafer material properties and process requirements, preset the mapping rule library Y and the wear threshold M0;

[0044] S2, send the wafer into the dicing machine, the dicing machine executes wafer dicing, and collect the sensing data of the wafer knife mark image and the current data of the dicing mechanism;

[0045] S3. Determine the optimal cleaving parameter Zy based on the mapping rule base Y, the physical depth Wd of the wafer cleavage, and the blade health index HI.

[0046] S4, send the wafer into the dicing machine, and send the optimal dicing parameters Zy to the dicing machine to perform wafer dicing;

[0047] S5, when the qualified rate of the cracked pieces is lower than the iteration threshold F for H consecutive batches, the mapping rule base Y is iterated.

[0048] In a third aspect, the present invention provides a processor-readable storage medium storing a computer program, wherein the processor executes the computer program to implement the distributed automatic wafer dicing method as described in the second aspect of the present invention.

[0049] The advantages of this invention compared to the prior art are:

[0050] The distributed automated wafer dicing system and method proposed in this invention establish a real-time data transmission and feedback mechanism between the dicing and dicing processes. Through process data sharing, it achieves coordinated processing, improving yield and production efficiency while reducing production costs. By eliminating information gaps, replacing manual decision-making, and enhancing process stability and efficiency through highly automated and intelligent dicing-dicing parameter matching, it achieves a qualitative leap from "experience-driven" to "data-driven" approaches and enables proactive quality control.

[0051] The system adopts a distributed design, which can make full use of the enterprise's existing dicing and splitting equipment for modification. Combined with the intelligent parameter matching method of rule base and real-time compensation algorithm, it can realize one industrial control computer to control multiple process equipment, realize real-time data closed-loop control across equipment, and reduce the enterprise's production costs.

[0052] The process debugging time has been reduced from an average of 2-4 hours to less than 5 minutes, a reduction of more than 95%; quality consistency has been significantly improved, human factors have been eliminated, and the CPK (process capability index) of the splitting process has been increased from 1.2 to more than 1.8; a quantitative assessment method for blade wear status based on current sensor data enables proactive maintenance, and the blade health index provides early warnings of when to replace blades, avoiding sudden quality accidents; a parameter mapping model method based on an experience database and a rule engine lowers the technical threshold and significantly reduces the technical experience requirements for operators, which is conducive to knowledge accumulation and large-scale production. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of a distributed automated wafer dicing system provided in an embodiment of the present invention.

[0054] Figure 2This is a schematic diagram of a distributed automatic wafer dicing method provided in an embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0056] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0057] Example

[0058] This invention discloses a distributed automatic wafer dicing system and a distributed automatic wafer dicing method, which realizes linkage processing through process data sharing to improve yield and production efficiency and reduce production costs.

[0059] In a first aspect, the present invention provides a distributed automated wafer dicing system, as shown in the appendix. Figure 1 As shown, it specifically includes:

[0060] The dicing machine is electrically connected to an industrial switch and is used to receive control data from an industrial computer via a servo drive system to control the dicing mechanism to perform wafer dicing. The first data acquisition module collects the sensor data of the dicing mechanism.

[0061] The wafer dicing machine is electrically connected to an industrial switch and is used to receive control data from an industrial computer via a servo drive system to control the dicing mechanism to perform wafer dicing. The second data acquisition module collects the sensor data of the dicing mechanism.

[0062] Industrial switches are used for distributed data exchange and transmission.

[0063] The transmitter, electrically connected to an industrial switch, is used to transfer wafers from the dicing mechanism of the dicing machine to the cleaving mechanism of the dicing machine, and to transfer the cleaved wafers to the wafer collection mechanism.

[0064] An industrial control computer, electrically connected to an industrial switch, is used to receive and store sensor data, recipe parameters, and mapping rules. It also sends control data to the dicing machine, dicing machine, and transmitter to achieve an automated closed-loop wafer dicing process of "sensing-decision-execution." The recipe parameters are the process parameters that control the dicing and dicing machines.

[0065] The wafer collection mechanism includes qualified wafer collectors and unqualified wafer collectors.

[0066] The first data acquisition module is used to acquire sensor data of the dicing mechanism in real time, such as process parameters and status parameters; it includes: a high-precision circuit sensor and a first machine vision unit.

[0067] A high-precision circuit sensor is installed in the spindle motor power supply circuit of the dicing mechanism to collect current data of the dicing mechanism and monitor the cutting load and blade wear status.

[0068] For example, the high-precision circuit sensor has a range of 0-10A, a resolution of 1mA, and a sampling frequency of 1kHz.

[0069] The first machine vision unit, including a CCD camera and a coaxial LED cold light source, is positioned opposite the wafer dicing stage of the dicing mechanism to capture sensor data of wafer dicing marks.

[0070] For example, the CCD camera is a 5-megapixel monochrome camera.

[0071] The second data acquisition module is used to acquire sensor data of the splitting mechanism in real time, such as process parameters and status parameters; it includes: a second machine vision unit.

[0072] The second machine vision unit, including a CCD camera and a coaxial LED cold light source, is positioned directly opposite the cleaving mechanism's operating table to capture sensor data of wafer images.

[0073] For example, the CCD camera is a 5-megapixel monochrome camera.

[0074] The industrial control computer is equipped with a qualification detection module, which identifies whether the edges of the cleaved wafer are regular, neat, and undamaged based on the sensor data from the second data acquisition module. If so, it is judged as a qualified wafer and sent to the qualified wafer collector via a transmitter; otherwise, it is judged as a defective wafer and sent to the defective wafer collector via a transmitter. The module also statistically analyzes the wafer cleaving yield data (identifying indicators such as wafer edge integrity and chip breakage rate) and correlates and records the optimal cleaving parameters Zy and K (compensation coefficients) for this batch with the cleaving yield data.

[0075] The industrial control computer is equipped with databases, including time-series databases and relational databases; the time-series database (such as InfluxDB) is used to store sensor data, and the relational database (such as MySQL) is used to store recipe parameters and mapping rule base Y.

[0076] The industrial control computer determines the optimal dicing parameters Zy based on the mapping rule base Y, the physical depth Wd of the wafer cleavage, and the blade health index HI, and then sends the optimal dicing parameters Zy to the dicing machine to perform wafer dicing.

[0077] Furthermore, the mapping rule base Y is used to query preset mapping parameters, which include: preset dicing depth Wd0, preset blade health index HI0, and preset dicing parameters Pi. Wherein, the preset dicing parameters Pi = {P1, P2, P3}, where P1 is the preset dicing pressure, P2 is the preset dicing speed, P3 is the preset support table spacing, and i is the parameter number.

[0078] For example, taking a wear threshold M0 of 0.15 as an example, the mapping rule base Y is set as follows:

[0079] Table 1. Mapping rule base Y when wear threshold M0 is 0.15

[0080]

[0081] Among them, P1 ranges from 17 to 25, and the unit is g; P2 ranges from 11 to 15, and the unit is mm / s; P3 ranges from 350 to 372, and the unit is μm.

[0082] Furthermore, the physical depth Wd of the wafer grommets is calculated in real time based on the wafer grommets image sensing data of the first machine vision unit using an image edge detection algorithm. The specific algorithm for Wd is as follows:

[0083] Wd= Sd × Q (Formula 1).

[0084] Where Q = E / e; the physical depth Wd of the wafer grommets is in μm; Sd is the pixel depth of the grommets, in pixels, obtained from image sensing data based on the machine vision module through an image edge detection algorithm (such as the Canny algorithm); Q is the wafer grommets image calibration coefficient, obtained through camera calibration; E is the actual physical size, in μm; and e is the pixel size, in pixels.

[0085] Ideally, camera calibration should be performed under the same optical conditions to avoid the impact of lens distortion on accuracy.

[0086] For example, if 1 pixel corresponds to an actual physical size of 5 μm, then the calibration coefficient Q is 5.

[0087] Further, the blade health index HI is obtained by real-time calculation based on the current data of the scribing mechanism. The specific algorithm for the blade health index HI is as follows:

[0088] HI = |I_avg - I_new| / I_new (Formula 2).

[0089] Where, I_avg is the average current value of each cutting cycle, and I_new is the initial reference current value after replacing the new blade.

[0090] When HI > M1, it is determined that the blade is suddenly damaged, and the wafer scribing is immediately stopped and an alarm is given;

[0091] When HI ≤ M0, Pi is queried and determined in the mapping rule library Y, and Pi is set as the optimal chipping parameter Zy: Zy = Pi = {P1, P2, P3}; the optimal chipping parameter Zy is sent to the chipping machine and the wafer chipping is executed;

[0092] When M0 < HI ≤ M1, it is determined that the blade is significantly worn, and the new execution parameter Pi_new is obtained through the adaptive compensation algorithm, and Pi_new is set as the optimal chipping parameter Zy: Zy = Pi_new = {P1_new, P2_new, P3_new}; the optimal chipping parameter Zy is sent to the chipping machine and the wafer chipping is executed.

[0093] Where, M1 is the damage warning value; M0 is the wear threshold, and the value range is 0.15 to 0.5; the wear threshold M0 and the damage warning value M1 are determined by the actual production process; P1_new is the new execution chipping pressure, P2_new is the new execution chipping speed, and P3_new is the new execution support table spacing; the new execution parameter Pi_new is determined by the following formula:

[0094] Pi_new = Pi × (1 + K × HI) (Formula 3).

[0095] Where, K is the compensation coefficient, and the value range is 0.1 to 0.5. For example, K is 0.3 for gallium arsenide GaAs wafers and K is 0.2 for indium phosphide InP wafers, which is determined by material tests.

[0096] The value of Pi is queried and determined in the mapping rule library Y based on the physical depth Wd of the wafer knife marks and the value of the blade health index HI.

[0097] After compensation, the value of Pi_new must ensure that it is within the safe operation range of the cutting machine.

[0098] Preferably, the industrial control computer sends the new execution parameter value Pi_new into the designated register of the chipper PLC via the Modbus TCP protocol. After receiving the new parameter, the chipper PLC will immediately apply it in the next working cycle.

[0099] For example, consider the batch parameter adjustment process of a certain batch of gallium arsenide (GaAs) wafers with a thickness of 150 μm:

[0100] The calculated dicing depth Wd is 45 μm, and the blade health index HI = 0.12. Therefore, the 4th row (i.e., dicing depth 40-50 μm, HI 0.1-0.15) is selected from the mapping rule base Y in Table 1. The preset dicing pressure P1 = 20 g, preset dicing speed P2 = 13 mm / s, and preset support stage spacing P3 = 365 μm are obtained from the table. Wafer dicing is then performed using Pi as the optimal dicing parameter Zy.

[0101] When the blade health index HI deteriorates and exceeds the wear threshold M0=0.15, the adaptive compensation algorithm is activated. For GaAs wafer material K=0.3, the values ​​of P1_new, P2_new and P3_new are calculated based on Formula 3, and the new execution cleaving pressure P1_new=20×(1+0.3×0.12)=20.72g≈21g is obtained. The new execution cleaving speed is maintained at 13mm / s, and the new execution support stage spacing is finely adjusted to 367μm.

[0102] Pi_new was sent to the dicing machine as the optimal dicing parameter Zy. The dicing machine completed the parameter update and performed wafer dicing within 10ms. The yield of this batch of wafer dicing reached 99.8%, which is 1.2% higher than that of manual adjustment.

[0103] Furthermore, the industrial control computer automatically calculates the wafer cleaving yield data (identifying indicators such as wafer edge integrity and chip breakage rate) through the data from the second machine vision unit, and records the optimal cleaving parameters Zy and compensation coefficient K of this batch in association with the wafer cleaving yield data.

[0104] Furthermore, when the cleavage pass rate data corresponding to a certain set of parameters is lower than the iteration threshold F for three consecutive batches, the mapping rule base Y is iterated:

[0105] 1) Extract characteristic parameters such as the physical depth Wd of the wafer tool marks, the blade health index HI value, and the optimal cleaving parameter Zy of this batch of wafers;

[0106] 2) Compare the deviation with the historical best parameter set;

[0107] 3) Automatically adjust the preset mapping parameters of the corresponding interval in the mapping rule library Y (e.g., fine-tune the reference pressure of HI0.1-0.15 and the cutting depth 40-50μm interval from 20g to 21g).

[0108] 4) The updated mapping rule base Y will take effect immediately in the next batch, forming a continuous iterative closed loop of "data collection → anomaly triggering → rule optimization → effect verification".

[0109] Optionally, F is 99% to 99.9%; preferably, F = 99.5%. H is 1 to 5; preferably, H = 3.

[0110] Optionally, the dicing machine, slicing machine, and transmitter can be one or more units, and distributed deployment processing can be achieved through industrial switches.

[0111] A second aspect of the present invention provides a distributed automated wafer dicing method, as shown in the appendix. Figure 2 As shown, it includes:

[0112] S1, based on the properties of wafer materials and process requirements, preset the mapping rule library Y and wear threshold M0;

[0113] S2, the wafer is fed into the dicing machine, the dicing machine performs wafer dicing, and the sensor data of the wafer dicing mark image and the current data of the dicing mechanism are collected;

[0114] S3. Determine the optimal cleaving parameter Zy based on the mapping rule base Y, the physical depth Wd of the wafer cleavage, and the blade health index HI.

[0115] S4, send the wafer into the dicing machine, and send the optimal dicing parameters Zy to the dicing machine to perform wafer dicing;

[0116] S5, when the qualified rate of the cracked pieces is lower than the iteration threshold F for H consecutive batches, the mapping rule base Y is iterated.

[0117] Based on the same design concept as the first aspect of the present invention, the specific technical details of the method are also based on the same concept as the first aspect of the present invention, and will not be repeated here.

[0118] A third aspect of the present invention provides a processor-readable storage medium storing a computer program, wherein the processor executes the computer program to implement the distributed automated wafer dicing method as described in the second aspect of the present invention. Based on the same design concept as the first aspect of the present invention, further details are omitted here.

[0119] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A distributed automated wafer dicing system, characterized in that, include: The dicing machine is electrically connected to an industrial switch and is used to receive control data from an industrial computer via a servo drive system to control the dicing mechanism to perform wafer dicing. The first data acquisition module collects the sensor data of the dicing mechanism. The wafer dicing machine is electrically connected to an industrial switch and is used to receive control data from an industrial computer via a servo drive system to control the dicing mechanism to perform wafer dicing. The second data acquisition module collects the sensor data of the dicing mechanism. Industrial switches are used for distributed data exchange and transmission. An industrial control computer is electrically connected to an industrial switch; it is used to receive and store the sensor data, store recipe parameters and mapping rule base Y; and it is used to send control data to the dicing machine, dicing machine and transmitter to realize automated closed-loop wafer dicing of perception-decision-execution.

2. The distributed automated wafer dicing system according to claim 1, characterized in that, Also includes: The transmitter, electrically connected to an industrial switch, is used to transfer wafers from the dicing mechanism of the dicing machine to the cleaving mechanism of the dicing machine, and to transfer the cleaved wafers to the wafer collection mechanism. Wafer collection mechanism, including qualified wafer collectors and unqualified wafer collectors; The first data acquisition module is used to acquire the sensing data of the dicing mechanism in real time, including: a high-precision circuit sensor and a first machine vision unit; The high-precision circuit sensor is installed in the power supply circuit of the spindle motor of the dicing mechanism to collect the current data of the dicing mechanism and monitor the cutting load and blade wear status. The first machine vision unit includes a CCD camera and a coaxial LED cold light source, and is set to face the wafer dicing operation table of the dicing mechanism to capture sensor data of wafer dicing marks. The second data acquisition module is used to acquire sensor data of the chopping mechanism in real time, including: a second machine vision unit; The second machine vision unit, including a CCD camera and a coaxial LED cold light source, is positioned directly opposite the cleaving mechanism operating table and is used to capture sensor data of wafer images.

3. The distributed automated wafer dicing system according to claim 2, characterized in that: The industrial control computer is equipped with a qualification detection module, which is used to identify whether the edges of the pyrolyzed wafer are regular, neat and undamaged based on the sensor data of the second data acquisition module. If so, it is judged as a qualified wafer and sent to the qualified wafer collector via the transmitter. If not, it is judged as an unqualified wafer and sent to the unqualified wafer collector via the transmitter. Statistical analysis of wafer cleaving yield data was conducted, and the optimal cleaving parameter Zy and compensation coefficient K for this batch were correlated and recorded with the cleaving yield data. The industrial control computer is equipped with a database, including a time-series database and a relational database; the time-series database is used to store sensor data, and the relational database is used to store the recipe parameters and the mapping rule base Y. The industrial control computer determines the optimal dicing parameters Zy based on the mapping rule base Y, the physical depth Wd of the wafer cleavage, and the blade health index HI, and then sends the optimal dicing parameters Zy to the dicing machine to perform wafer dicing.

4. The distributed automated wafer dicing system according to claim 3, characterized in that: The mapping rule base Y is used to query preset mapping parameters, which include: A preset scribing cutting depth Wd0, a preset blade health index HI0, and a preset chip breaking parameter Pi; Among them, the preset chip breaking parameter Pi = {P1, P2, P3}, P1 is the preset chip breaking pressure, P2 is the preset chip breaking speed, P3 is the preset distance between support platforms, and i is the parameter serial number.

5. A distributed wafer automatic cutting system according to claim 3, characterized in that: The physical depth Wd of the wafer knife mark is obtained by real-time calculation through an image edge detection algorithm based on the wafer knife mark image sensing data of the first data acquisition module. The specific algorithm of Wd is: Wd = Sd × Q (Formula 1); Among them, Q = E / e; Sd is the pixel depth of the knife mark, obtained through an image edge detection algorithm based on the image sensing data of the machine vision module; Q is the wafer knife mark image calibration coefficient, obtained through camera calibration; E is the actual physical size; e is the pixel size.

6. A distributed wafer automatic cutting system according to claim 3, characterized in that: The blade health index HI is obtained by real-time calculation based on the current data of the scribing mechanism. The specific algorithm of HI is: HI = |I_avg - I_new| / I_new (Formula 2); Among them, I_avg is the average current value of each cutting cycle, and I_new is the initial reference current value after replacing the new blade.

7. A distributed wafer automatic cutting system according to claim 6, characterized in that: When HI > M1, it is determined that the blade is suddenly damaged, and the wafer scribing is immediately stopped and an alarm is given; When HI ≤ M0, query and determine Pi in the mapping rule library Y, and set Pi as the optimal chip breaking parameter Zy: Zy = Pi = {P1, P2, P3}; send the optimal chip breaking parameter Zy to the chip breaker and perform wafer chip breaking; When M0 < HI ≤ M1, it is determined that the blade has been significantly worn, and the new execution parameter Pi_new is obtained through an adaptive compensation algorithm, and Pi_new is set as the optimal chip breaking parameter Zy: Zy = Pi_new = {P1_new, P2_new, P3_new}; send the optimal chip breaking parameter Zy to the chip breaker and perform wafer chip breaking; Among them, M0 is the wear threshold; M1 is the damage warning value; the wear threshold M0 and the damage warning value M1 are determined by the actual production process; P1_new is the new execution chip breaking pressure, P2_new is the new execution chip breaking speed, and P3_new is the new execution distance between support platforms.

8. A distributed wafer automatic cutting system according to claim 7, characterized in that: The new execution parameter Pi_new is determined by the following formula: Pi_new = Pi × (1 + K × HI) (Formula 3). Among them, K is the compensation coefficient, and the value range is 0.1 to 0.

5. When the wafer is a gallium arsenide GaAs wafer, K is 0.3, and when the wafer is an indium phosphide InP wafer, K is 0.2, which is determined by material tests; The value of Pi is determined by querying in the mapping rule library Y based on the physical depth Wd of the wafer knife mark and the value of the blade health index HI.

9. A distributed automated wafer dicing method, applied to the distributed automated wafer dicing system according to any one of claims 1 to 8, characterized in that, Including the following steps: S1, based on the properties of wafer materials and process requirements, preset the mapping rule library Y and wear threshold M0; S2, the wafer is fed into the dicing machine, the dicing machine performs wafer dicing, and the sensor data of the wafer dicing mark image and the current data of the dicing mechanism are collected; S3. Determine the optimal cleaving parameter Zy based on the mapping rule base Y, the physical depth Wd of the wafer cleavage, and the blade health index HI. S4, send the wafer into the dicing machine, and send the optimal dicing parameters Zy to the dicing machine to perform wafer dicing; S5, when the qualified rate of the cracked pieces is lower than the iteration threshold F for H consecutive batches, the mapping rule base Y is iterated.

10. A processor-readable storage medium, characterized in that, The processor-readable storage medium stores a computer program, and when the processor executes the computer program, it implements the distributed automatic wafer dicing method as described in claim 9.

Images