Coaxial cable damage detection device and detection method for semiconductor manufacturing process

Abstract

The invention relates to a coaxial cable damage detection device and detection method for a semiconductor manufacturing process, a first detection device is arranged at the joint of a coaxial cable and a power supply, a second detection device is arranged at the joint of the coaxial cable and a matcher, and measurement data is transmitted to a control system for calculation and judgment. When the dissipation power measurement value exceeds the preset dissipation power set value, an alarm is triggered immediately to inform an operator to process, so that the monitoring of the operation state of the coaxial cable becomes possible, and the defect that the coaxial cable lacks an effective monitoring means is overcome; a dissipation power set value formula of the coaxial cable under a constant use condition is introduced, a system can accurately set a dissipation power set value conforming to the actual characteristics of the cable under a stable work condition, and a correction coefficient under the change of the use condition is introduced, so that the system can accurately set the dissipation power set value. The control system can dynamically adjust the dissipation power set value according to the change of the environment temperature, humidity and working frequency, and compensates the influence on the power loss of the cable due to the change of working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor manufacturing processes. Specifically, it relates to a coaxial cable damage detection device and a detection method for semiconductor manufacturing processes. Background Art

[0002] In the semiconductor manufacturing industry, the normal operation of machines such as PECVD (Plasma Enhanced Chemical Vapor Deposition), PVD (Physical Vapor Deposition), LPCVD (Low Pressure Chemical Vapor Deposition), and MOCVD (Metal Organic Chemical Vapor Deposition) depends on stable medium-frequency / radio-frequency power transmission. Currently, coaxial cables are generally used to connect medium-frequency / radio-frequency power supplies to the above-mentioned machines and their matching devices to achieve power transmission. However, under actual working conditions, coaxial cables face complex environmental factors such as temperature changes, mechanical vibrations, and electromagnetic interference, which may all cause damage to coaxial cables.

[0003] Once the coaxial cable is damaged, dangerous situations such as arcing, burning, and smoking will occur during power transmission. In the prior art, coaxial cables lack effective monitoring and detection means. When problems occur, they cannot actively give feedback. After damage occurs, there is no online detection method to timely discover and prevent the damage from expanding. It is not until serious dangers such as severe smoking and electric shock occur that an alarm will be triggered across the entire factory, forcing the factory to stop production for maintenance. This not only leads to production stagnation and huge economic losses, but also may affect the product delivery schedule due to the long time required to troubleshoot faults.

[0004] The above problems are worthy of solution. Summary of the Invention

[0005] In order to overcome the deficiencies of the existing technology, the present invention provides a coaxial cable damage detection device and a detection method for semiconductor manufacturing processes, which can detect the operating state of the power coaxial cable in real time and accurately, give an alarm in a timely manner when an abnormality occurs, avoid serious production accidents caused by coaxial cable problems, and ensure the continuity and stability of semiconductor production.

[0006] The technical solution of the present invention is as follows:

[0007] A coaxial cable damage detection device for semiconductor manufacturing processes, comprising:

[0008] A first detection device, which is arranged at the connection between the coaxial cable and the power supply and is used to measure the input end of the coaxial cable;

[0009] A second detection device, which is arranged at the connection between the coaxial cable and the matching device and is used to measure the output end of the coaxial cable;

[0010] A control system that can read data at the input end and output end of a coaxial cable, is used to set a dissipated power set value, and calculate a measured value of the dissipated power of the coaxial cable; when the measured value of the dissipated power is greater than the dissipated power set value, the control system triggers an alarm and executes a protection measure.

[0011] As a preferred technical solution of the present invention, the calculation formula for the measured value of the dissipated power of the coaxial cable is: P diss = P in - P out ;

[0012] Wherein, P in is the first power detection value at the input end of the coaxial cable; P out is the second power detection value at the output end of the coaxial cable.

[0013] As a preferred technical solution of the present invention, the control system is further used to synchronously detect the current, voltage, forward power and reverse power at the input end and output end of the coaxial cable; and compare the detection data at the input and output ends to determine whether there are transient abnormal phenomena such as arcing and intermittent; when a transient abnormality is detected, the control system triggers an alarm and executes a protection measure.

[0014] As a preferred technical solution of the present invention, the protection measures include:

[0015] Display the abnormal position and type through a human-machine interface;

[0016] Automatically disconnect the power output and start the standby power supply;

[0017] Generate an exception report, including the exception time, position, type and recommended handling solution.

[0018] As a preferred technical solution of the present invention, when the operating conditions remain constant and within the normal operating range of the coaxial cable, the dissipated power set value formulas for coaxial cables of different models and lengths are:

[0019] For a coaxial cable of model RG-58, the dissipated power set value P RG-58 = 2L + 3;

[0020] For a coaxial cable of model RG-11, the dissipated power set value P RG-11 = 3L + 9;

[0021] For a coaxial cable of model LMR-400, the dissipated power set value P LMR-400 = 2.5L + 8;

[0022] Wherein, L represents the length of the coaxial cable.

[0023] Further, when the operating conditions change, the control system dynamically adjusts the dissipated power set value according to the following formula:

[0024] P coax_set = P base ·f(T, H, F);

[0025] Wherein, P base is the dissipated power set value under stable operating conditions, T is the ambient temperature, H is the ambient humidity, F is the operating frequency; f(T, H, F) represents the operating condition adjustment coefficient;

[0026] The calculation formula of f(T, H, F) is: f(T, H, F) = 1 + α(T - T 0 ) + β(H - H 0 ) + γ(F - F 0 );

[0027] Wherein, T 0 is the reference temperature; H 0 is the reference humidity; F 0 is the reference frequency; α, β, and γ respectively represent the correction coefficients of temperature, humidity, and frequency.

[0028] As a preferred technical solution of the present invention, the correction coefficients α, β, and γ are determined in the following manner:

[0029] In the initial stage of device use, when there is a lack of historical data, according to the type and material characteristics of the coaxial cable, it is obtained by fitting experimental data;

[0030] After the device has been used for a period of time, with the accumulation of historical detection data, a machine learning algorithm is introduced, and through the machine learning algorithm, it is dynamically optimized based on historical detection data.

[0031] Further, in the initial stage of device use, when there is a lack of historical data, the specific steps for obtaining the correction coefficients α, β, and γ according to the type and material characteristics of the coaxial cable by fitting experimental data include:

[0032] Step A1. Set the reference temperature, reference humidity, and reference frequency. Under this reference operating condition, measure the power loss of coaxial cables of different lengths to obtain the dissipated power set value;

[0033] Step A2. Keep the humidity and frequency as reference values, change the temperature, measure the power loss of the coaxial cable at different temperatures, and record multiple groups of data at different temperatures;

[0034] Step A3. Keep the temperature and frequency as reference values, change the humidity, measure the power loss of the coaxial cable at different humidities, and record multiple groups of data at different humidities;

[0035] Step A4. Keep the temperature and humidity at reference values, change the frequency, measure the power loss of the coaxial cable at different frequencies, and record multiple groups of data at different frequencies;

[0036] Step A5. Calculate the correction coefficients α, β, and γ through data fitting.

[0037] Furthermore, dynamically optimizing the correction coefficients α, β, and γ based on historical detection data through a machine learning algorithm includes the following steps:

[0038] Step B1. Collect multiple data samples under normal operating conditions, including ambient temperature, ambient humidity, operating frequency, and corresponding measured values of actual dissipated power; perform data preprocessing and calculate the offset ΔT, ΔH, and ΔF for each sample.

[0039] Step B2. Construct a multiple linear regression model P coax_set =P base ·[1 + αΔT + βΔH + γΔF] to minimize the error between the predicted dissipated power set value and the measured value of actual dissipated power.

[0040] Step B3. Solve for the correction coefficients;

[0041] Step B4. Adjust the verification model.

[0042] Even further, obtaining the correction coefficients α, β, and γ through a machine learning algorithm also includes the steps:

[0043] Step B5. Regularly update the model to adapt to new changes in operating conditions;

[0044] Regularly collect new detection data, update the feature matrix X and the target variable y, and re-solve for α, β, and γ.

[0045] The present invention also provides a method for detecting damage to a coaxial cable used in a semiconductor manufacturing process, including the following steps:

[0046] Step 1. Measure the power at the input end and the output end of the coaxial cable;

[0047] Step 2. Calculate the measured value of the dissipated power of the coaxial cable;

[0048] Step 3. Determine the model and length of the coaxial cable;

[0049] Step 4. Determine whether the operating conditions are stable, including whether the ambient temperature, humidity, and operating frequency are within a stable range;

[0050] If the operating conditions are stable, then according to the model and length of the coaxial cable, determine the dissipated power set value P base ;

[0051] If the working condition is unstable, the dissipated power set value is dynamically adjusted according to the following formula:

[0052] P coax_set = P base ·f(T, H, F);

[0053] where f(T, H, F) = 1 + α(T - T 0 ) + β(H - H 0 ) + γ(F - F 0 ); T 0 is the reference temperature; H 0 is the reference humidity; F 0 is the reference frequency; α, β, and γ are correction coefficients;

[0054] Step 5: Determine whether the measured dissipated power is greater than the set value of the dissipated power;

[0055] If the measured dissipated power is greater than the set value of the dissipated power, an alarm is triggered and protective measures are executed.

[0056] The beneficial effects of the present invention according to the above solution are as follows:

[0057] 1. By setting the first detection device at the connection between the coaxial cable and the power supply to measure the input end, and setting the second detection device at the connection between the coaxial cable and the matcher to measure the output end, the present invention can obtain the relevant data of the input and output of the coaxial cable in real time. The data is transmitted to the control system for calculation and judgment. When the measured dissipated power exceeds the preset set value of the dissipated power, an alarm is immediately triggered to notify the operator to handle it, making it possible to monitor the operating state of the coaxial cable, solving the defect of the lack of effective monitoring means for coaxial cables in the existing semiconductor manufacturing process technology, and being able to detect abnormalities at the initial stage of cable damage (such as local damage and poor contact), avoiding the expansion of damage;

[0058] 2. Introducing the formula for the set value of the dissipated power of the coaxial cable under constant operating conditions, the system can accurately set the set value of the dissipated power that conforms to the actual characteristics of the cable under stable operating conditions; calculating the set value of the dissipated power according to the formula for different coaxial cable models provides a more practical standard for accurately judging whether the coaxial cable is working properly, greatly improving the accuracy and pertinence of detection, reducing misjudgment situations caused by unreasonable set values of the dissipated power, and ensuring the stability of power transmission in the semiconductor manufacturing process;

[0059] 3. By introducing the correction coefficient under the change of working conditions, the control system can dynamically adjust the dissipated power setting value according to the changes in ambient temperature, humidity and working frequency, which can effectively compensate for the impact of the working condition change on the cable power loss, and timely adjust the dissipated power setting value to avoid misjudging the power loss caused by normal working condition changes as cable damage, which significantly improves the adaptability and reliability of the detection device under complex working conditions;

[0060] 4. The present invention adopts different algorithms to determine the correction coefficient according to different situations in the use stage of the device; specifically, in the initial use of the device, when there is a lack of historical data, the correction coefficient is obtained by fitting the experimental data according to the model and material properties of the coaxial cable, which provides the system with an initial correction basis based on the basic physical properties of the cable, so that the system can initially adapt to different working conditions when it is put into use, and realize preliminary effective monitoring of the status of the coaxial cable; and after the device has been used for a period of time and historical detection data has been accumulated, a machine learning algorithm is introduced to dynamically optimize the correction coefficient based on these data. The machine learning algorithm can explore the potential rules in the data, better adapt to complex and changeable actual working conditions, and continuously improve the accuracy and adaptability of the correction coefficient over time, thereby further improving the intelligence level and detection accuracy of the detection device. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a structural schematic diagram of the present invention;

[0062] Figure 2 The figure is a flow chart of the method of the present invention.

[0063] In the figure,

[0064] 1. Power supply; 2. Matching device; 3. Coaxial cable; 4. First detection device; 5. Second detection device; 6. Control system. DETAILED DESCRIPTION

[0065] In order to better understand the purpose, technical scheme and technical effect of the present invention, the present invention is further explained in conjunction with the accompanying drawings and embodiments. It should be noted that similar numbers and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, it is stated that the embodiments described below are only used to explain the present invention and are not used to limit the present invention.

[0066] It should be noted that when an element is referred to as being "fixed on" or "disposed on" another element, it may be directly on the other element or there may also be a central element, and when an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time.

[0067] The indicated orientation or positional relationship is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of this application is customarily placed during use, or the orientation or positional relationship commonly understood by those skilled in the art, or the orientation or positional relationship in which the product of this application is customarily placed during use. It is only for the convenience of describing this application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to this application.

[0068] The terms "first" and "second" are only used for the purpose of convenient description, and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of technical features.

[0069] As Figure 1 and Figure 2 As shown, a coaxial cable damage detection device for semiconductor manufacturing processes includes:

[0070] A first detection device 4, disposed at the connection between the coaxial cable 3 and the power supply 1 and used to measure the input end of the coaxial cable;

[0071] A second detection device 5, disposed at the connection between the coaxial cable 3 and the matcher 2 and used to measure the output end of the coaxial cable;

[0072] A control system 6, which can read the data of the input end and the output end of the coaxial cable, is used to set the dissipated power set value, and calculate the measured value of the dissipated power of the coaxial cable; when the measured value of the dissipated power is greater than the dissipated power set value, the control system triggers an alarm and executes protection measures. The protection measures include: displaying the abnormal position and type through a human-machine interface; automatically disconnecting the power output and starting the backup power supply; generating an abnormal report, including the abnormal time, position, type and recommended treatment plan.

[0073] In the present invention, the device has a human-machine interface, and can display the abnormal position and type through the human-machine interface. In the system, each coaxial cable will be assigned a unique identifier (such as a number or name), and corresponds to the physical position. The abnormal type refers to the specific nature of the fault, such as too high a measured value of dissipated power, and there are problems such as arcing, intermittency, and impedance mismatch in the coaxial cable. When an abnormality is detected, the control system will determine the identifier of the faulty cable according to the detection data, and display its position in a graphical or text form on the human-machine interface. In a specific embodiment, there is a coaxial cable numbered "Cable-01" in the system, which is connected between the power supply A and the matcher B. When a fault occurs in this cable and it is detected that the measured value of the dissipated power exceeds the threshold, the human-machine interface will display "Cable-01 fault, position: power supply A → matcher B; type: too high a measured value of dissipated power".

[0074] In the present invention, a backup power supply is arranged between the power supply of the semiconductor manufacturing machine tool and the matcher, enabling the coaxial cable damage detection device to automatically disconnect the power output and start the backup power supply. The backup power supply refers to a backup power supply system that can immediately take over the work of the main power supply when the main power supply fails or malfunctions. The backup power supply can be a UPS (uninterruptible power supply), a backup generator, or other forms of emergency power supply. When an abnormality of the coaxial cable is detected, the control system will immediately disconnect the output of the main power supply and at the same time start the backup power supply to provide continuous power supply for the semiconductor device. Through this technical solution, equipment downtime caused by cable faults can be avoided, the continuity of the production process can be ensured, and time can be gained for maintenance personnel to troubleshoot and repair faults without affecting production; the reliability and fault tolerance of the system can be improved, and production interruption caused by single-point faults can be avoided.

[0075] In the present invention, the calculation formula for the measured dissipated power value of the coaxial cable is: P diss = P in - P out ; where P in is the first power detection value at the input end of the coaxial cable; P out is the second power detection value at the output end of the coaxial cable. Through the above formula, the system quickly calculates the measured dissipated power value of the coaxial cable, obtains the energy loss of the cable itself, clearly characterizes the loss degree of the coaxial cable, and avoids the error of relying only on single-port detection; it can quickly identify cable abnormalities, reduce the system operation burden, and improve real-time performance.

[0076] In the present invention, when the operating conditions (such as ambient temperature, humidity, operating frequency) remain constant and are within the normal operating range of the coaxial cable, the formula for the set dissipated power value of coaxial cables of different models and different lengths is:

[0077] For a coaxial cable of model RG-58, the set dissipated power value P RG-58 = 2L + 3;

[0078] For a coaxial cable of model RG-11, the set dissipated power value P RG-11 = 3L + 9;

[0079] For a coaxial cable of model LMR-400, the set dissipated power value P LMR-400 = 2.5L + 8;

[0080] where L represents the length of the coaxial cable.

[0081] The present invention realizes for the first time a specific formula for setting the dissipated power value based on the cable model and length, which can adapt to different cables, take into account the inherent loss characteristics of different cables, and through multiple experiments, summarizes the cable length range applicable to the formula, avoiding the misjudgment rate caused by the inapplicability of the formula, and ensuring that the threshold setting is scientific and reasonable. Users only need to input the cable model and length to automatically generate a reference value, simplifying the operation process and improving the usability of the device.

[0082] In an actual semiconductor manufacturing workshop, the ambient temperature and humidity will change with seasons and production activities, and the working frequency may also change due to process adjustments. For example, when the temperature rises, the resistance of the cable may increase, resulting in an increase in power loss. Therefore, on the basis of the static dissipated power setting value, the present invention further introduces a dynamic adjustment mechanism to cope with the changes in working conditions (temperature, humidity, working frequency) in the actual production environment.

[0083] When the working conditions change, the control system dynamically adjusts the dissipated power setting value according to the following formula:

[0084] P coax_set =P base ·f(T, H, F);

[0085] Wherein, P base is the dissipated power setting value under stable working conditions (such as P RG-58 , P RG-11 , P LMR-400 in the foregoing scheme), T is the ambient temperature, H is the ambient humidity, F is the working frequency; f(T, H, F) represents the working condition adjustment coefficient;

[0086] The calculation formula of f(T, H, F) is: f(T, H, F) = 1 + α(T - T 0 ) + β(H - H 0 ) + γ(F - F 0 );

[0087] Wherein, T 0 is the reference temperature; H 0 is the reference humidity; F 0 is the reference frequency; α, β, and γ respectively represent the correction coefficients of temperature, humidity, and frequency.

[0088] It can be seen that by dynamically adjusting the dissipated power setting value, the present invention makes up for the deficiencies of the static formula under complex working conditions, enables the system to adapt to different working conditions in real time, always maintains accurate monitoring of the working state of the coaxial cable, further improves the reliability and stability of the detection device, and reduces the detection error caused by the change of working conditions.

[0089] In the present invention, the correction coefficients α, β, and γ are determined by experimental data fitting and machine learning algorithms according to different stages of device use. In the early stage of device use, when historical data is lacking, the correction coefficients are obtained by fitting experimental data according to the model and material properties of the coaxial cable, which provides the system with an initial correction basis based on the basic characteristics of the cable, so that the system can make preliminary adjustments to the dissipated power setting value under different working conditions at the initial stage of operation, and has basic working condition adaptability.

[0090] After the device has been used for a period of time, as historical detection data accumulates, machine learning algorithms are introduced to dynamically optimize based on historical detection data. Machine learning algorithms can mine potential patterns in data, for example, learning the real power loss variation patterns of coaxial cables under different temperature, humidity and frequency combinations, and then more accurately adjust the dissipated power setting value, reduce false alarms and missed alarms, improve the intelligence level and detection accuracy of the detection device, and ensure that during long-term use, the system can continuously and stably monitor and protect coaxial cables effectively.

[0091] In the present invention, according to the model and material properties of the coaxial cable, the specific steps of obtaining the correction coefficients α, β, and γ by fitting the experimental data include:

[0092] Step A1. Set a reference temperature, a reference humidity, and a reference frequency. Under the reference working conditions, measure the power loss of coaxial cables of different lengths to obtain a dissipated power setting value;

[0093] Step A2. Keep the humidity and frequency as reference values, change the temperature, measure the power loss of the coaxial cable at different temperatures, and record multiple sets of data at different temperatures;

[0094] Step A3. Keep the temperature and frequency as reference values, change the humidity, measure the power loss of the coaxial cable under different humidity, and record multiple sets of data under different humidity;

[0095] Step A4. Keep the temperature and humidity as reference values, change the frequency, measure the power loss of the coaxial cable at different frequencies, and record multiple sets of data at different frequencies;

[0096] Step A5: Calculate the correction coefficients α, β, and γ by data fitting.

[0097] Taking the RG-58 coaxial cable as an example, in order to obtain the influence of temperature, humidity and frequency on the power loss of the coaxial cable, the ambient temperature, humidity and operating frequency will be changed respectively, and the power loss of the coaxial cable under different working conditions will be measured to obtain experimental data.

[0098] The experimental equipment includes: a temperature control box for controlling the temperature of the experimental environment; a humidity generator for adjusting the humidity of the experimental environment; a signal generator for generating signals of different frequencies.

[0099] As described in the aforementioned experimental steps, specifically:

[0100] Step A1. Set the reference temperature T 0 = 25 °C, the reference humidity H 0 = 50%, the reference frequency F 0 = 100 MHz. Under this reference operating condition, measure the power loss of RG-58 coaxial cables of different lengths to obtain the basic dissipation power setting value P base .

[0101] Step A2. Keep the humidity and frequency at the reference values and change the temperature T to obtain several groups of temperature experimental data:

[0102] T 1 = 20 °C, P 11 = 4.5 W;

[0103] T 2 = 25 °C, P 12 = 5.0 W;

[0104] T 3 = 30 °C, P 13 = 5.5 W;

[0105] T 4 = 35 °C, P 14 = 6.0 W.

[0106] Step A3. Keep the temperature and frequency at the reference values and change the humidity H to obtain several groups of humidity experimental data:

[0107] H 1 = 40%, P 21 = 4.8 W;

[0108] H 2 = 50%, P 22 = 5.0 W;

[0109] H 3 = 60%, P 23 = 5.2 W;

[0110] H 4 = 70%, P 24 = 5.4 W.

[0111] Step A4. Keep the temperature and humidity at the reference values and change the frequency F to obtain several groups of frequency experimental data:

[0112] F 1= 80 MHz, P 31 = 4.6 W;

[0113] F 2 = 100 MHz, P 32 = 5.0 W;

[0114] F 3 = 120 MHz, P 33 = 5.4 W;

[0115] F 4 = 140 MHz, P 34 = 5.8 W.

[0116] Step A5. Calculate the temperature correction coefficient α;

[0117] According to the formula P coax_set = P base ·f(T, H, F) = P base ·(1 + α(T - T 0 ))), when the humidity and frequency are reference values, use the temperature experiment data for fitting. Taking T 0 = 25 °C, P base = 5.0 W as an example, select two groups of data T 1 = 20 °C, P 11 = 4.5 W and T 3 = 30 °C, P 13 = 5.5 W. For T 1 = 20 °C, P 11 = 4.5 W, P 11 = P base (1 + α(T - T 0 ))), that is, 4.5 = 5.0×(1 + α(20 - 25)), and we can get α = 0.02. Use the other group of data T 3 = 30 °C, P 13 = 5.5 W for verification:

[0118] P 13 = P base (1 + α(T - T 0 )) = 5.0×(1 + 0.02(30 - 25)) = 5.5 W, which is consistent with the measured value.

[0119] Calculate the temperature correction coefficient β;

[0120] According to the formula P coax_set = P base ·f(T, H, F) = P base ·(1 + β(H - H 0 ))), when the temperature and frequency are reference values, use the humidity experiment data for fitting. Taking H 0= 50%, P base = 5.0 W as an example, select two sets of data H 1 = 40%, P 21 = 4.8 W and H 3 = 60%, P 23 = 5.2 W. For H 1 = 40%, P 21 = 4.8 W, P 21 = P base (1 + β(H - H 0 ))), that is, 4.8 = 5.0×(1 + β(40 - 50)), we can get β = 0.004. Use another set of data H 3 = 60%, P 23 = 5.2 W for verification:

[0121] P 23 = P base ·(1 + β(H - H 0 )) = 5.0×(1 + 0.004(60 - 50)) = 5.2 W, which is consistent with the measured value.

[0122] Calculate the temperature correction coefficient γ;

[0123] According to the formula P coax_set = P base ·f(T, H, F) = P base ·(1 + γ(F - F 0 ))), when the temperature and humidity are reference values, use the frequency experimental data for fitting. Take F 0 = 100 MHz, P base = 5.0 W as an example, select two sets of data F 1 = 80 MHz, P 31 = 4.6 W and F 3 = 120 MHz, P 33 = 5.4 W. For F 1 = 80 MHz, P 31 = 4.6 W, P 31 = P base ·(1 + γ(F - F 0 ))), that is, 4.6 = 5.0×(1 + γ(80 - 100)), we can get γ = 0.004. Use another set of data F 3 = 120 MHz, P 33 = 5.4 W for verification:

[0124] P 33 = P base ·(1 + γ(F - F 0 )) = 5.0×(1 + 0.004(120 - 100)) = 5.4 W, which is consistent with the measured value.

[0125] Through the above experiments and data fitting, the correction factors of the coaxial cable of model RG-58 under different working conditions are obtained: α = 0.02, β = 0.004, γ = 0.004.

[0126] This step provides a scientific and repeatable method for determining the correction factors, ensuring that the set value of the dissipated power can be reasonably adjusted according to the working conditions based on the characteristics of the cable itself and experimental data in the absence of historical data. The correction factors obtained in this way enable the system to have a preliminary adaptability to different working conditions at the initial stage of device use, laying a foundation for accurately judging the working state of the coaxial cable subsequently, improving the accuracy and reliability of detection during the initial operation of the system, and ensuring the stable operation of the equipment in the initial stage of semiconductor production.

[0127] Dynamically optimizing the correction factors α, β, γ based on historical detection data through machine learning algorithms includes the following steps:

[0128] Step B1. Collect multiple data samples under normal working conditions, including ambient temperature, ambient humidity, working frequency, and corresponding measured values of actual dissipated power; perform data preprocessing and calculate the offset ΔT, ΔH, ΔF of each sample.

[0129] Step B2. Construct a multiple linear regression model P coax_set = P base · [1 + αΔT + βΔH + γΔF] to minimize the error between the predicted set value of the dissipated power and the measured value of the actual dissipated power.

[0130] Step B3. Solve the correction factors.

[0131] Step B4. Adjust the verification model.

[0132] Still taking the coaxial cable of model RG-58 as an example for illustration. Specifically, perform Step B1, and the detection data of the coaxial cable under normal working conditions are as follows in the table:

[0133]

[0134]

[0135] Define the reference working condition: reference temperature T 0 = 25 °C, reference humidity H 0 = 50%, reference frequency F 0 = 100 MHz. For each sample, calculate the working condition adjustment coefficient f(T, H, F) = 1 + αΔT + βΔH + γΔF, where ΔT = T - T 0 ΔH = H - H 0 ΔF = F - F 0. Calculate the operating condition offset: ΔT = T - T 0 , ΔH = H - H 0 , ΔF = F - F 0 , and the results are as follows:

[0136]

[0137] According to the measured dissipation power value P diss_actual under normal conditions, the adjusted dissipation power set value P coax_set can cover these normal values, that is, P coax_set ≥ P diss_actual . According to the normal operating state of the cable, when there is no damage, the measured dissipation power value should be within a certain reasonable range, and when there is damage, the measured dissipation power value will increase abnormally. Therefore, the goal of machine learning may be to learn the variation law of the measured dissipation power value with T, H, and F under normal operating conditions, so as to dynamically adjust P coax_set , making it able to adapt to these changes and avoid false alarms. For example, when the temperature rises, the normal loss of the coaxial cable may also increase, so P coax_set needs to be increased accordingly to avoid misjudging the normal loss as abnormal at high temperatures. Therefore, through the machine learning model, the effects of temperature, humidity, and frequency on the normal loss can be learned, so as to adjust the coefficients α, β, and γ, making P coax_set = P base · f(T, H, F) can accurately reflect the maximum allowable loss under the current operating conditions.

[0138] Execute step B2, the dynamic dissipation power set value adjustment formula: P coax_set = P base · [1 + αΔT + βΔH + γΔF];

[0139] Among them, P base is the dissipation power set value of the coaxial cable under the reference operating conditions. For example, for the coaxial cable of model RG - 58 mentioned above, P RG-58 = 2L + 3. Let L = 10 meters, then P RG-58 = 23W.

[0140] Define a loss function, mean square error (MSE), to make the predicted P coax_set as close as possible to the actually measured P diss_actual , and the formula is:

[0141] P diss_actual represents the measured dissipation power value under the actual normal operating conditions.

[0142] Execute step B3, the problem is transformed into minimizing the MSE, and the calculation process is as follows:

[0143] First, transform it into a multiple linear regression:

[0144]

[0145] where ε is the error term.

[0146] Then, define the target variable The feature matrix X = [ΔT, ΔH, ΔF].

[0147] Perform data transformation:

[0148]

[0149] Then, for the feature matrix X and the target variable y:

[0150]

[0151] Then, solve for the coefficients α, β, and γ by the least squares method:

[0152] θ = (X T X) -1 X T y;

[0153] where θ = [α, β, γ] T ;

[0154] Manually calculating the inverse of a 3x3 matrix is cumbersome. It can be solved using a numerical calculation tool such as numpy in Python software. Substituting gives:

[0155]

[0156] Execute step B4. When the environmental temperature T = 30°C, humidity H = 60%, and frequency F = 150 MHz, from the previous steps, we have:

[0157] ΔT = +5, ΔH = +10, ΔF = +50;

[0158] Then, the adjustment coefficient: f(T, H, F) = 1 + (-0.002)×5 + (-0.001)×10 + 0.0005×50 = 1.005;

[0159] Therefore, the adjustment of the dissipated power setting value: P coax_set = P base ×1.005 = 23×1.005 = 23.115 W.

[0160] By learning the variation law of the measured value of the dissipated power with the environmental temperature, humidity, and operating frequency under normal operating conditions, continuously optimize the correction coefficient so that the set value of the dissipated power P coax_setIt can more accurately reflect the maximum allowable loss under the current working conditions. In actual production, with the fluctuations of environmental factors, the machine learning model can adjust the correction coefficient in real time, thereby adjusting the set value of the dissipated power, effectively avoiding false alarms caused by changes in working conditions, and further improving the adaptability and detection accuracy of the detection device to complex working conditions.

[0161] In a preferred embodiment, the steps of obtaining the correction coefficients α, β, and γ through the machine learning algorithm further include:

[0162] Step B5. Regularly update the model to adapt to new changes in working conditions;

[0163] Regularly collect new detection data, update the feature matrix X and the target variable y, and re-solve α, β, and γ.

[0164] This supplementary step regularly collects new detection data, updates the feature matrix X and the target variable y, and re-solves α, β, and γ. In the actual semiconductor manufacturing environment, the working conditions are constantly changing, which may introduce new interference factors, or the performance of the coaxial cable may change due to equipment aging and other reasons. By regularly updating the model, the system can timely capture these changes, adjust the correction coefficient, ensure that the set value of the dissipated power always conforms to the actual situation, enhance the self-adaptability and long-term stability of the system, and ensure that the semiconductor production can always reliably detect the damage of the coaxial cable under changing working conditions, avoid production accidents caused by failure to adapt to changes in working conditions in a timely manner, and ensure the continuity and stability of the production process.

[0165] In the present invention, the control system is also used to synchronously detect the current, voltage, forward power, and reverse power at the input end and the output end of the coaxial cable; and compare the detection data at the input and output ends to determine whether there are transient abnormal phenomena such as arcing and intermittency. Its working principle: In the normal working state, the current, voltage, forward power, and reverse power of the coaxial cable are all within a relatively stable range and have a specific relationship. Once an arcing phenomenon occurs, a large current change and power fluctuation will be generated instantaneously, the forward power may suddenly drop, and the reverse power will increase abnormally; while the intermittent situation will cause intermittent interruption or unstable changes in the current and voltage. By capturing these abnormal changes, the system can timely and accurately identify the transient abnormality.

[0166] When detecting ignition, since ignition is usually accompanied by instantaneous peaks of current and voltage, the control system monitors the transient fluctuations of current and voltage in real time through the high-sampling-rate current and voltage detection functions of the first detection device and the second detection device. When detecting intermittency, the continuity of current and voltage is monitored at the input end and the output end respectively through the first detection device and the second detection device, and at the same time, the changes in forward power and reverse power are monitored. By comparing the power changes at the input end and the output end, it is judged whether there is an intermittency phenomenon. When detecting impedance mismatch, the control system monitors the forward power and reverse power in real time through the first detection device and the second detection device, and can also calculate the voltage standing wave ratio (VSWR), and can judge the impedance matching state: when the reverse power or VSWR exceeds the corresponding set threshold, it is determined as impedance mismatch.

[0167] When a transient anomaly is detected, the control system triggers an alarm and executes protection measures, and the protection measures include:

[0168] Display the anomaly location and type through the human-machine interface;

[0169] Automatically disconnect the power output and start the standby power supply;

[0170] Generate an anomaly report, including the anomaly time, location, type and recommended handling solutions.

[0171] As Figure 2 shown, the present invention also provides a coaxial cable damage detection method for semiconductor manufacturing processes, including the following steps:

[0172] Step 1, measure the power at the input end and the output end of the coaxial cable;

[0173] Step 2, calculate the measured value of the dissipated power of the coaxial cable;

[0174] Step 3, determine the model and length of the coaxial cable;

[0175] Step 4, determine whether the operating conditions are stable, including whether the ambient temperature, humidity and operating frequency are within the stable range;

[0176] If the operating conditions are stable, then according to the model and length of the coaxial cable, determine the set value P of the dissipated power through a preset static formula base ;

[0177] If the operating conditions are unstable, then dynamically adjust the set value of the dissipated power according to the following formula:

[0178] P coax_set = P base ·f(T, H, F);

[0179] wherein, f(T, H, F) = 1 + α(T - T 0) + β(H - H 0 ) + γ(F - F 0 ); T 0 is the reference temperature; H 0 is the reference humidity; F 0 is the reference frequency; α, β, and γ are correction factors;

[0180] Step 5, determine whether the measured dissipated power is greater than the set dissipated power;

[0181] If the measured dissipated power is greater than the set dissipated power, trigger an alarm and execute protection measures.

[0182] Among them, the correction factors α, β, and γ are determined as follows: In the initial stage of device use, when there is a lack of historical data, they are obtained by fitting experimental data according to the model and material characteristics of the coaxial cable; after the device has been used for a period of time, with the accumulation of historical detection data, a machine learning algorithm is introduced, and through the machine learning algorithm, it is dynamically optimized based on the historical detection data. The specific determination process is as described above and will not be elaborated here.

[0183] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0184] The above embodiments only represent several implementation manners of the present invention. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the invention patent should be subject to the appended claims.

Claims

1. A coaxial cable damage detection device for semiconductor process, characterized in that: include: A first detection device is provided at the connection between the coaxial cable and the power supply and is used to measure the input end of the coaxial cable; A second detection device is provided at the connection between the coaxial cable and the matching device and is used to measure the output end of the coaxial cable; A control system is used to set a dissipation power setting value and calculate a dissipation power measurement value of the coaxial cable; when the dissipation power measurement value is greater than the dissipation power setting value, the control system triggers an alarm and executes protective measures.

2. The coaxial cable damage detection device for semiconductor process according to claim 1, characterized in that: The calculation formula for the measured power dissipation of coaxial cable is: P diss =P in -P out ; Among them, P in is the first power detection value at the coaxial cable input end; P out is the second power detection value at the output end of the coaxial cable.

3. The coaxial cable damage detection device for semiconductor process according to claim 1, characterized in that: The control system is also used to synchronously detect the current, voltage, forward power and reverse power of the coaxial cable input end and the coaxial cable output end; and compare the detection data of the input and output ends to determine whether there are transient abnormal phenomena such as sparking and intermittent; when a transient abnormality is detected, the control system triggers an alarm and executes protection measures.

4. The coaxial cable damage detection device for semiconductor process according to claim 1 or 3, characterized in that: The protection measures include: Display the abnormal location and type through the human-machine interface; Automatically disconnect the power output and start the backup power supply; Generate an exception report, including the exception time, location, type and recommended treatment plan.

5. The coaxial cable damage detection device for semiconductor process according to claim 1, characterized in that: When the operating conditions remain constant and within the normal operating range of the coaxial cable, the formulas for setting the power dissipation of coaxial cables of different models and lengths are: The power dissipation setting value P of the RG-58 coaxial cable RG-58 =2L+3; The power dissipation setting value P of the RG-11 coaxial cable RG-11 =3L+9; The power dissipation setting value P of the coaxial cable model LMR-400 LMR-400 =2.5L+8; Where L is the length of the coaxial cable.

6. The coaxial cable damage detection device for semiconductor process according to claim 5, characterized in that: When the operating conditions change, the control system dynamically adjusts the dissipated power setting value according to the following formula: P coax_set =P base ·f(T,H,F); Among them, P base is the power dissipation setting value under stable operating conditions, T is the ambient temperature, H is the ambient humidity, and F is the operating frequency; f(T,H,F) represents the operating condition adjustment coefficient; The calculation formula of f(T,H,F) is: f(T,H,F)=1+α(T-T0)+β(H-H0)+γ(F-F0); Among them, T0 is the reference temperature; H0 is the reference humidity; F0 is the reference frequency; α, β, and γ represent the correction coefficients of temperature, humidity, and frequency respectively.

7. The coaxial cable damage detection device for semiconductor process according to claim 6, characterized in that: The correction coefficients α, β, and γ are determined in the following way: In the early stage of device use, when historical data is lacking, it is obtained by fitting experimental data based on the model and material properties of the coaxial cable; After the device has been used for a period of time and with the accumulation of historical detection data, a machine learning algorithm is introduced to perform dynamic optimization based on the historical detection data.

8. The coaxial cable damage detection device for semiconductor process according to claim 7, characterized in that: In the early stage of using the device, when there is a lack of historical data, the specific steps for fitting the correction coefficients α, β, and γ by experimental data according to the model and material properties of the coaxial cable include: Step A1. Set a reference temperature, a reference humidity, and a reference frequency. Under the reference working conditions, measure the power loss of coaxial cables of different lengths to obtain a dissipated power setting value; Step A2. Keep the humidity and frequency as reference values, change the temperature, measure the power loss of the coaxial cable at different temperatures, and record multiple sets of data at different temperatures; Step A3. Keep the temperature and frequency as reference values, change the humidity, measure the power loss of the coaxial cable under different humidity, and record multiple sets of data under different humidity; Step A4. Keep the temperature and humidity as reference values, change the frequency, measure the power loss of the coaxial cable at different frequencies, and record multiple sets of data at different frequencies; Step A5: Calculate the correction coefficients α, β, and γ by data fitting.

9. The coaxial cable damage detection device for semiconductor process according to claim 7, characterized in that: Through machine learning algorithms, the dynamic optimization of correction coefficients α, β, and γ based on historical detection data includes the following steps: Step B1. Collect multiple data samples under normal working conditions, including ambient temperature, ambient humidity, operating frequency and corresponding actual dissipated power measurement values; preprocess the data and calculate the offset ΔT, ΔH, ΔF of each sample; Step B2. Construct multiple linear regression model P coax_set =P base [1+αΔT+βΔH+γΔF], which minimizes the error between the predicted power dissipation setting and the actual power dissipation measurement. Step B3. Calculate the correction coefficient; Step B4. Adjust the verification model.

10. A coaxial cable damage detection method for semiconductor process, characterized in that: The following steps are involved: Step 1, measuring the power at the input end and the power at the output end of the coaxial cable; Step 2, calculating the measured value of the dissipated power of the coaxial cable; Step 3: Determine the model and length of the coaxial cable; Step 4: Determine whether the operating conditions are stable, including whether the ambient temperature, humidity and operating frequency are within a stable range; If the working condition is stable, the dissipated power setting value P is determined by the preset static formula according to the model and length of the coaxial cable. base ; If the operating conditions are unstable, the power dissipation setting value is dynamically adjusted according to the following formula: P coax_set =P base ·f(T,H,F); Where, f(T,H,F)=1+α(T-T0)+β(H-H0)+γ(F-F0); T0 is the reference temperature; H0 is the reference humidity; F0 is the reference frequency; α, β, γ are correction coefficients; Step 5: determine whether the measured value of the dissipated power is greater than the set value of the dissipated power; If the measured power dissipation value is greater than the set power dissipation value, an alarm is triggered and protective measures are executed.

Images