Blade damage detection method, device, dicing machine and storage medium

Abstract

The present application provides a blade damage detection method, device, dicing machine and storage medium, which solves the problem of low accuracy of blade detection in related technologies. The dicing machine of the present application converts the signal of the corresponding luminous flux from the time domain to the frequency domain, and then extracts the signal characteristics that can better represent the luminous flux under each working mode. Compared with the processing of the time domain signal, the dicing machine can effectively eliminate the influence of irregular changes in the luminous flux caused by interference factors through processing in the frequency domain, and can combine the corresponding wear detection and breakage detection neural network models, and can simultaneously realize the functional detection of tool wear and tool breakage, thereby improving the robustness and diversity of blade status detection, and has strong anti-interference ability and high recognition accuracy, which helps to reduce false alarms or missed alarms, thereby improving production efficiency.

Description

Technical Field

[0001] The present application relates to the technical field of dicing machines, and in particular to a blade damage detection method, device, dicing machine and storage medium. Background Art

[0002] The dicing machine is an important execution equipment of the post-packaging process of integrated circuits. Its processing capacity determines the quality and quantity of the chips. The working principle of the dicing machine is to achieve the dicing and separation of materials under the strong grinding action of the blade installed on the motor spindle. When working, the blade rotates at high speed driven by the motor to perform the dicing action. It is conceivable that the blade is prone to damage or wear due to long-term work during the production process. At this time, if the blade continues to dicing, it will affect the dicing quality and even damage the cut part.

[0003] In this regard, the industry requires that an alarm signal be issued when the blade is damaged or reaches a certain amount of wear during the cutting process, so as to stop cutting in time. In the related technology, the traditional blade status recognition solution is achieved by real-time monitoring of the absolute amount of light flux, but the accuracy of the conclusion of blade wear or damage obtained by this solution is low, and it is very easy to have false alarms or missed alarms, resulting in low production efficiency and even economic losses. Summary of the invention

[0004] The present application provides a blade damage detection method, device, dicing machine and storage medium, which solves the problem of low accuracy of blade detection in related technologies. The present application can accurately identify the status of the blade, has strong anti-interference ability and high recognition accuracy.

[0005] In a first aspect, the present application provides a blade damage detection method, which is applied to a dicing machine, wherein the dicing machine includes an optical transceiver component and a blade, wherein the optical transceiver component is used to send and receive optical signals, and to perform photoelectric conversion on the received optical signals to obtain a luminous flux signal, wherein the blade damage detection method includes:

[0006] Acquire a luminous flux signal, and perform data processing on the acquired luminous flux signal to obtain frequency domain characteristic information corresponding to the luminous flux signal;

[0007] Based on the first training model, feature recognition is performed on the frequency domain feature information to determine the current operating mode corresponding to the frequency domain feature information based on the signal features of the frequency domain feature information. The first training model is a neural network model that can identify different operating modes by training with frequency domain feature information corresponding to a plurality of preset operating modes as a training set;

[0008] According to the preset output strategy, the luminous flux value under the current working mode is determined. The output strategy generates the luminous flux value in different ways corresponding to different working modes;

[0009] When the working mode is the target mode, the blade is subjected to wear detection based on the luminous flux value corresponding to the luminous flux signal; and / or,

[0010] The blade is detected for damage based on the second training model to determine the corresponding detection result. The second training model is a neural network model that can identify whether the blade state is a damaged state by training the corresponding frequency domain feature information of blades in different blade states under multiple preset working modes as a training set.

[0011] In a second aspect, the present application further provides a blade damage detection device, the dicing machine includes an optical transceiver component and a blade, the optical transceiver component is used to send and receive optical signals, and obtain a luminous flux signal after photoelectric conversion of the received optical signal, the blade damage detection device includes:

[0012] A signal processing module is configured to obtain a luminous flux signal and perform data processing on the obtained luminous flux signal to obtain frequency domain characteristic information corresponding to the luminous flux signal;

[0013] The operating condition recognition module is configured to perform feature recognition on the frequency domain feature information based on the first training model, and determine the current operating condition mode corresponding to the frequency domain feature information based on the signal characteristics of the frequency domain feature information, wherein the first training model is a neural network model that can recognize different operating conditions and is obtained by training with the frequency domain feature information corresponding to a plurality of preset operating conditions as a training set;

[0014] The luminous flux output module is configured to determine the luminous flux value under the current working mode according to a preset output strategy, and the output strategy generates the luminous flux value in different ways corresponding to different working modes;

[0015] a damage detection module configured to detect wear of the blade based on the luminous flux value corresponding to the luminous flux signal when the working mode is the target mode; and / or,

[0016] The blade is detected for damage based on the second training model to determine the corresponding detection result. The second training model is a neural network model that can identify whether the blade state is a damaged state by training the corresponding frequency domain feature information of blades in different blade states under multiple preset working modes as a training set.

[0017] In a third aspect, the present application further provides a dicing machine, the dicing machine comprising:

[0018] one or more processors;

[0019] a storage device for storing one or more programs,

[0020] When one or more programs are executed by one or more processors, the one or more processors implement the blade damage detection method of the present application.

[0021] In a fourth aspect, the present application further provides a storage medium storing computer executable instructions, which are used to execute the blade damage detection method of the present application when executed by a processor.

[0022] The dicing machine of the present application converts the signal corresponding to the luminous flux from the time domain to the frequency domain, and then extracts the signal characteristics that can better represent the luminous flux under each working mode. Compared with the processing of the time domain signal, the dicing machine can effectively eliminate the influence of irregular changes in the luminous flux caused by interference factors through processing in the frequency domain, and can be combined with the corresponding neural network model for wear detection and breakage detection, and can simultaneously realize the functional detection of tool wear and tool breakage, thereby improving the robustness and diversity of blade status detection, and it has strong anti-interference ability and high recognition accuracy, which helps to reduce false alarms or missed alarms, thereby improving production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic diagram of the steps of a blade damage detection method provided in one embodiment of the present application.

[0024] Figure 2 A schematic diagram of the steps for calculating the luminous flux value under the fifth operating mode provided in an embodiment of the present application.

[0025] Figure 3 A schematic structural diagram of a blade damage detection device provided in one embodiment of the present application.

[0026] Figure 4 A schematic diagram of the structure of a dicing machine provided in one embodiment of the present application. DETAILED DESCRIPTION

[0027] The embodiments of the present application are further described in detail below in conjunction with the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are only used to explain the embodiments of the present application, rather than to limit the embodiments of the present application. It should also be noted that, for ease of description, only the parts related to the embodiments of the present application rather than all structures are shown in the accompanying drawings, and those skilled in the art should be able to think of it after reading the specification of this application that as long as the technical features do not contradict each other, any combination of the technical features can constitute an optional implementation method.

[0028] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable when appropriate, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally a class, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally represents that the objects associated before and after are in an "or" relationship. In the description of the present application, "multiple" means two or more, and "several" means one or more.

[0029] The dicing machine is an important execution equipment for the post-packaging process of integrated circuits. Its processing capacity determines the quality and quantity of chips. The dicing machine blade is a tool on the dicing machine, which is used to cut various wafers, semiconductor packaging substrates, ceramics, glass and other hard and brittle materials. That is, the dicing machine realizes the dicing and separation of materials through the strong grinding action of the blade installed on the motor spindle. In this regard, during the production process, the blade is prone to breakage or wear due to long-term work. At this time, if the blade continues to cut, it will affect the cutting quality and even damage the cut part.

[0030] Therefore, it is necessary to identify the blade to determine whether it is damaged or worn. The traditional identification scheme is achieved by real-time monitoring of the absolute amount of luminous flux. The basis for judging wear is: as the cutting progresses, the blade gradually wears in the radial direction, causing the luminous flux to slowly increase over a long period of time. When the luminous flux reaches the set threshold, it is considered that the blade has insufficient life and the cutting process cannot continue. The basis for judging damage is: when the blade is instantly damaged, the crack on the blade will let in more light, causing the monitored luminous flux to increase instantly, while the undamaged area will continue to block the light, and the corresponding luminous flux value will remain at a small value.

[0031] On a dicing machine, usually during the dicing process, the dicing machine moves the blade into the material through the motor spindle, so that the rotating blade can achieve the dicing and separation of the material under the strong grinding action, and the dicing machine will spray water on the blade through the water nozzle on it to reduce the cutting temperature and friction heat, thereby extending the service life of the tool. At the same time, the water spray can also clean the debris on the surface of the tool and the surface of the cut object to prevent clogging of the blade, thereby improving the dicing efficiency.

[0032] It can be seen that the actual working conditions of the dicing machine are complex. For example, the cutting water flow, the chips generated by the processed material, the stability of the optical fiber sensor itself, the placement of the sensor, the vibration interference of the electromechanical equipment, the opening and closing of the water flow, the start and stop interference of the cutter wheel shaft and other factors will cause irregular changes in the light flux. Therefore, the traditional recognition scheme monitors the absolute value change of the light flux in the time domain to detect wear or damage, but the detection accuracy is low, which can easily lead to the failure of blade status detection, resulting in false alarms or missed alarms, resulting in low production efficiency and even economic losses.

[0033] In this regard, the present application provides a blade damage detection method, which is applied to a dicing machine to detect the wear and / or damage of the blade. It can be imagined that the dicing machine includes an optical transceiver component, a blade, a motor and a water nozzle, wherein the main shaft of the motor is connected to the blade and drives the blade to rotate to achieve the dicing of the processed material; and the water nozzle is located on one side of the blade to spray water to the blade during the dicing process; and the optical transceiver component is used to send and receive optical signals. It can be imagined that the optical transceiver component includes an optical transmitter and an optical receiver, and the optical transmitter and the optical receiver are respectively located on both sides of the blade so that the light beam emitted by the optical transmitter can be perpendicular to the blade surface and received by the optical receiver. It should be noted that the optical transmitter can be a laser transmitter or a fiber optic transmitter, and correspondingly, the optical receiver is a laser receiver or a fiber optic receiver. In this regard, in practical applications, the installation positions of the optical transmitter and the optical receiver are offset so that the emitted light beam is partially blocked by the blade, thereby facilitating the detection of the blade state through the light flux.

[0034] In addition, the optical transceiver assembly also includes a photoelectric conversion unit, such as a photoelectric conversion circuit, a photoelectric conversion device, etc. Exemplarily, the photoelectric conversion device is used as an example to illustrate that the analog quantity such as the received optical signal is photoelectrically converted to obtain a corresponding analog signal, and then the light flux signal as a digital signal is usually obtained based on the AD sampling principle. For example, after the optical signal is received at one end of the optical receiver, the signal is amplified and filtered by the amplification and filtering circuit set on the device, and then the ADC chip of the device is used to convert the signal into a discrete digital signal, that is, the above-mentioned light flux signal, so as to facilitate further signal processing. The circuit structure of the amplification and filtering circuit can refer to the relevant technology, and no further elaboration is made here.

[0035] Of course, it is conceivable that the dicing machine also includes a processor and a memory. Among them, the processor and the memory can be connected via a bus or other means. In this regard, the blade damage detection method in the embodiment of the present application applied by the dicing machine is executed by a processor, the number of processors is one or more, and the processor may include a central processing unit (CPU), a microcontroller unit (MCU) and / or a field programmable gate array (FPGA) and other processing units.

[0036] As a computer-readable storage medium, the memory can be used to store software programs, computer executable programs and modules, such as the program instructions / modules corresponding to the blade damage detection method in the embodiment of the present application. The memory can store computer programs required for at least one function, data created according to the use of the dicing machine, etc. At present, the memory stores at least a relevant computer program for implementing the blade damage detection method, so that the processor can implement the blade damage detection method based on the computer program stored in the memory.

[0037] Based on this, Figure 1 A schematic diagram of the steps of a blade damage detection method provided in an embodiment of the present application. When the blade damage detection method is executed by the above-mentioned dicing machine, the method is specifically executed by the processor thereon. Figure 1 The blade damage detection method includes steps S110 to S140:

[0038] Step S110: Acquire a luminous flux signal, and perform data processing on the acquired luminous flux signal to obtain frequency domain characteristic information corresponding to the luminous flux signal.

[0039] The dicing machine can obtain a corresponding luminous flux signal after receiving and transmitting the light signal through the optical transceiver component thereon and performing photoelectric conversion. It can be imagined that the optical receiver in the optical transceiver component can obtain an analog voltage signal corresponding to the light intensity of the light signal after receiving the light signal, and then obtain the luminous flux signal as a digital signal through sampling, holding, quantization, etc., and then perform data processing.

[0040] It can be imagined that the acquired luminous flux signal is a time domain signal. In order to prevent the luminous flux change in the time domain from being excessively affected by external factors and resulting in low accuracy of the detection result, the luminous flux signal is processed to convert the luminous flux signal from the time domain to the frequency domain, thereby obtaining the frequency domain characteristic information of the corresponding luminous flux signal.

[0041] In one embodiment, the frequency domain characteristic information is recorded in a spectrum diagram, such as a linear amplitude spectrum, a logarithmic amplitude spectrum, etc., to represent the relationship between the voltage amplitude and the frequency. It can be imagined that the luminous flux is represented by a voltage value. Therefore, the data processing of the luminous flux signal is based on the processing of the digital signal, and the corresponding steps include step S111-step S112:

[0042] Step S111 : Based on a preset denoising processing algorithm, perform signal denoising on the luminous flux signal to determine a denoised luminous flux signal.

[0043] Step S112: convert the denoised luminous flux signal from the time domain to the frequency domain, and record the relationship between the frequency and amplitude of the luminous flux signal in the frequency domain to generate a spectrum diagram corresponding to the frequency domain characteristic information of the luminous flux signal.

[0044] After the light flux signal is acquired, the signal is denoised, such as by using a preset denoising algorithm, such as Gaussian low-pass filtering, statistical filtering, Savitzky-Golay filtering algorithm, etc., to filter out the noise interference in the light flux signal. After the signal denoising is completed, the light flux signal is converted from the time domain to the frequency domain, and the relationship between the frequency and amplitude of the light flux signal in the frequency domain is recorded, that is, the relationship between the light flux and the frequency is recorded.

[0045] It is conceivable that the luminous flux signal can be converted from the time domain to the frequency domain through signal processing techniques such as Fourier transform and wavelet transform. It can be understood that the Fourier transform uses a sine wave of infinite width as the basis function, which pays more attention to the frequency domain characteristics of the signal; the wavelet transform uses a wavelet function of finite width as the basis function, which can provide the frequency domain signal of the signal and also provide the localization information of the signal in the time domain. In this regard, the dicing machine in the embodiment focuses on the characteristics of the luminous flux signal in the frequency domain. Either Fourier transform or wavelet transform can be set to obtain the frequency domain characteristic information of the luminous flux signal in the frequency domain. Furthermore, the frequency domain characteristic information corresponding to the luminous flux signal is recorded through a spectrum diagram to facilitate the subsequent identification of the frequency domain characteristic information.

[0046] Therefore, the noise interference in the signal acquisition and conversion process is removed through denoising, and the signal characteristics of the luminous flux signal can be more obvious by converting the signal into the frequency domain. Compared with the processing of time domain signals, the dicing machine can effectively eliminate the influence of irregular changes in luminous flux caused by interference factors through processing in the frequency domain, thereby better detecting the blade status of the blade, which helps to improve the accuracy of the detection results of the blade status.

[0047] Step S120: Based on the first training model, feature recognition is performed on the frequency domain feature information to determine the current operating mode corresponding to the frequency domain feature information based on the signal characteristics of the frequency domain feature information.

[0048] The first training model is a neural network model that can identify different working modes by training with frequency domain feature information corresponding to a plurality of preset working modes as a training set. It can be imagined that by obtaining a luminous flux signal under different working modes of the dicing machine and converting the luminous flux signal into a frequency domain signal, the corresponding frequency domain feature information is extracted and the extracted frequency domain feature information is used as a training set to train the neural network model, so that the obtained first training model can identify which working mode the current working mode is according to the input frequency domain feature information.

[0049] Then, for the first training model after the training is completed, the frequency domain feature information corresponding to the acquired luminous flux signal is used as the input parameter of the first training model. The first training model performs feature recognition on the frequency domain feature information to determine the current operating mode corresponding to the frequency domain feature information according to the corresponding signal characteristics, that is, to determine the current operating mode of the dicing machine.

[0050] Step S130: Determine the luminous flux value under the current working mode according to a preset output strategy, wherein the output strategy generates the luminous flux value in different ways corresponding to different working modes.

[0051] The working condition module of the dicing machine includes a first working condition mode associated with turning on the motor, a second working condition mode associated with turning off the motor, a third working condition mode associated with turning on the water flow nozzle, a fourth working condition mode associated with turning off the water flow nozzle, and a fifth working condition mode associated with turning on the motor and the water flow nozzle. Accordingly, corresponding to different working conditions, the dicing machine is provided with corresponding output strategies to generate luminous flux values ​​in different ways for different working conditions.

[0052] In one embodiment, based on different ways of generating luminous flux values, the sawing machine is provided with a first mode, a second mode and a third mode, wherein the first mode includes a first operating mode and a second operating mode, the second mode includes a third operating mode and a fourth operating mode, and the third mode includes a fifth operating mode.

[0053] Therefore, when the current working mode is the first type of mode, the processor of the dicing machine calculates the peak value of the acquired luminous flux signal and uses it as the luminous flux value. It can be imagined that the processor can calculate the signal peak value at the sampling moment according to the luminous flux signal corresponding to each sampling moment, and use it as the luminous flux value. In addition, in some embodiments, the generated luminous flux value can also be that the processor converts the signal peak value at the sampling moment into a light intensity value according to the numerical correspondence between voltage and light intensity in the photoelectric conversion process, and uses the light intensity value as the luminous flux value. Optionally, after determining the luminous flux value, the dicing machine can also display the current luminous flux value to the user through a display device or a host computer, such as transmitting the luminous flux value to a display device or a host computer through a wireless or wired transmission method for display by the display device and the host computer.

[0054] It can be understood that when the motor is turned on or off to make the blade rotate or stop, and the relative position of the blade and the optical transceiver assembly remains unchanged, the effect of the blade rotating or not on the luminous flux is small and can be ignored. For this reason, the peak value of the luminous flux signal is selected as the luminous flux value.

[0055] The dicing machine records the value of the luminous flux signal at each sampling moment. When the current working mode is the second type of mode, the processor selects the signal value of the luminous flux signal at the previous sampling moment as the luminous flux value. It can be imagined that the signal value of the luminous flux signal at each sampling moment can be the signal peak value at the sampling moment. It can be understood that for the opening or closing of the water flow nozzle, the instantaneous change brought about by this situation has a greater impact on the luminous flux; while at the previous sampling moment, the water flow nozzle is in a stable state of not closing or opening, and its impact on the luminous flux is relatively small. In this regard, using the value at the previous sampling moment as the luminous flux value can better reflect the luminous flux.

[0056] When the current working mode is the third type of mode, the dicing machine collects the values ​​of the luminous flux signal at multiple sampling moments through the set sampling window. The number of collected values ​​is determined according to the window span of the sampling window. It can be imagined that the sampling window is virtual, which is used to limit the collection range in the time series during the digital signal processing. For example, the sampling window includes four sampling moments from moment T(n) to moment T(n+4), and accordingly, the window span of the sampling window is 4. Based on this, the processor obtains the first window data and the second window data in the luminous flux signal through the sampling window, that is, according to the signal values ​​corresponding to the luminous flux signal obtained by the sampling window at multiple sampling moments, and then obtains the first luminous flux corresponding to the first window data and the second luminous flux corresponding to the second window data to determine the luminous flux value.

[0057] Step S140: when the current operating mode is the target mode, perform wear detection on the blade based on the luminous flux value corresponding to the luminous flux signal; and / or perform damage detection on the blade based on the second training model to determine the corresponding detection result.

[0058] Optionally, in one embodiment, the fifth operating mode is used as the target mode. If the first training model recognizes that the current operating mode is the fifth operating mode, the dicing machine can perform wear detection and / or breakage detection according to a preset detection strategy. For example, the detection strategy set is to perform wear detection and breakage detection. Accordingly, the dicing machine performs wear detection and breakage detection respectively, and then feeds back whether the blade is worn and whether it is broken in the detection result. For another example, if the detection strategy set is to perform wear detection, the dicing machine performs wear detection. For another example, if the detection strategy set is to perform breakage detection, the dicing machine performs breakage detection.

[0059] Among them, for wear detection, the dicing machine performs wear detection on the blade based on the luminous flux value, that is, the wear amount of the blade is calculated by the luminous flux value, so that when the wear amount exceeds the corresponding threshold value, the blade state of the blade is determined to be a worn state. For damage detection, the dicing machine identifies the acquired luminous flux signal through a preset second training model to determine whether the blade is damaged, wherein the second training model is a neural network model that can identify whether the blade state is a damaged state by training with the frequency domain feature information corresponding to blades in different blade states under multiple preset working modes as a training set. It can be imagined that by obtaining the corresponding luminous flux signal under different working modes of the dicing machine carrying a normal blade, and then converting the luminous flux signal into a frequency domain signal, the corresponding frequency domain feature information is extracted as the first training set; and obtaining the corresponding luminous flux signal under different working modes of the dicing machine carrying a damaged blade, and then converting the luminous flux signal into a frequency domain signal, the corresponding frequency domain feature information is extracted as the second training set. The neural network model is trained using the training set as training data, so that the obtained second training model can identify whether the current blade is damaged according to the input frequency domain feature information.

[0060] Furthermore, after completing the above-mentioned tests, the blade status is recorded in the final test results. If a wear test is performed and the blade status is determined to be a worn state, the blade status is recorded as a worn state in the test results; if a breakage test is performed and the blade status is determined to be a broken state, the blade status is recorded as a broken state in the test results.

[0061] From the above scheme, it can be seen that the dicing machine converts the signal corresponding to the luminous flux from the time domain to the frequency domain, and then extracts the signal characteristics that can better represent the luminous flux under each working mode, and can be combined with the corresponding neural network model of wear detection and breakage detection, which can realize the functional detection of tool wear and tool breakage at the same time, and improve the robustness and diversity of blade status detection. It also has strong anti-interference ability and high recognition accuracy, which helps to reduce false alarms or missed alarms, thereby improving production efficiency.

[0062] Figure 2 A schematic diagram of the steps for calculating the luminous flux value under the fifth working mode provided by an embodiment of the present application. In one embodiment, for the first window data and the second window data, the processor obtains them through sampling windows respectively. Of course, the sampling moments included in the first window data and the second window data are not completely consistent. For the luminous flux value under the fifth working mode, the specific steps are as follows:

[0063] Step S210, respectively obtain a first mean value and a second mean value, and use the first mean value as the first luminous flux and the second mean value as the second luminous flux.

[0064] Step S220: Determine the difference between the first luminous flux and the second luminous flux to determine the luminous flux difference.

[0065] Step S230: Determine the sum of the luminous flux difference and the luminous flux value at the previous moment, and use it as the luminous flux value in the third mode.

[0066] It can be imagined that the first mean is the average value of the signal values ​​corresponding to multiple sampling moments in the first window data, and after selecting the corresponding sampling moment through the sampling window, the signal value corresponding to each sampling moment is obtained to calculate the corresponding average value. Similarly, the second mean is the average value of the signal values ​​corresponding to multiple sampling moments in the second window data, and after selecting the corresponding sampling moment through the sampling window, the signal value corresponding to each sampling moment is obtained to calculate the corresponding average value.

[0067] In this regard, the first mean value is used as the first luminous flux, the second mean value is used as the second luminous flux, and accordingly, the difference between the first luminous flux and the second luminous flux is calculated, and the difference is used as the luminous flux difference. Further, the luminous flux value in the third mode is determined to be the sum of the luminous flux value at the previous moment and the luminous flux difference.

[0068] Exemplarily, the window span of the sampling window is 4, and accordingly, the first window data and the second window data have the same amount of data, such as the first window data includes the signal value from time T(n) to time T(n-4), and the second window data includes the signal value from time T(n-1) to time T(n-5), that is, the first window data is shifted backward by one sampling time in time sequence relative to the second window data. Accordingly, the luminous flux value corresponding to the fifth working mode is:

[0069]

[0070] Among them, Q T is the signal value at time T(n), Q T-1 is the signal value at time T(n-1), L j is the first mean, L i It should be noted that the window span can also be 5, 6, 7, etc., which can be set according to application needs.

[0071] Optionally, in one embodiment, the calculation of the first mean value may be to add up the signal values ​​at each sampling moment according to the corresponding weighted coefficients and then take the average value, that is, to perform weighted average calculation. Similarly, the second mean value is also calculated by weighted average. It should be noted that, in some embodiments, the first luminous flux corresponding to the first window data and the second luminous flux corresponding to the second window data may also be calculated according to a preset proportional coefficient, that is, each sampling moment corresponds to a proportional coefficient, and the proportional coefficients corresponding to each sampling moment are different but the sum is 1, and then the corresponding luminous flux is determined by the product of the proportional coefficient and the signal value at each sampling moment.

[0072] Therefore, this scheme can use different luminous flux update methods according to different working modes, avoiding the cumbersome shortcomings of the traditional luminous flux calculation method, and can more easily generate the corresponding luminous flux. The luminous flux values ​​obtained corresponding to different working modes are more representative, and thus can more accurately represent the luminous flux of the current working mode.

[0073] In one embodiment, the dicing machine detects the wear of the blade after determining that the working mode is the target mode, such as the target mode is the fifth working mode, that is, the working mode associated with turning on the motor and the water nozzle. Moreover, in this working mode, the dicing machine is provided with a mode threshold, which is associated with the luminous flux of the blade in a normal state, so as to calculate the corresponding wear amount based on the mode threshold, and the specific steps are as follows: Step S141-Step S142:

[0074] Step S141 : determining the wear amount of the blade based on the difference between the preset mode threshold and the light flux value.

[0075] Step S142: When the wear amount is greater than or equal to a preset wear threshold, determine that the blade state of the blade is a worn state.

[0076] It can be understood that the difference between the mode threshold and the light flux value is calculated, and then the proportion of the difference in the mode threshold is used to determine the wear amount of the blade, such as the mode threshold n and the light flux value x. Accordingly, the wear amount w of the blade is:

[0077]

[0078] The wear amount is then compared with the preset wear threshold. When the wear amount is greater than or equal to the preset wear threshold, the dicing machine can determine that the blade state of the blade is a worn state. Exemplarily, the preset wear threshold is 60%, and the mode threshold is 100. If the currently determined luminous flux value is 60, the corresponding wear amount is 40%. The wear amount is less than 60%, and the blade state is not a worn state. Optionally, the dicing machine can set corresponding alarm controls according to the detection results. For example, when the detection result shows that the blade is in a worn state, a corresponding alarm signal is output to inform the user that the current blade is worn and needs to be replaced.

[0079] In this regard, after determining the luminous flux value of the target mode, the dicing machine can improve the accuracy of judging the blade wear by calculating the corresponding wear amount and then judging whether the blade is worn, without having to frequently stop the machine to check the blade wear amount.

[0080] In some embodiments, the dicing machine detects blade damage based on a second training model set thereon. The second training model is a neural network model that can identify whether the blade state is a damaged state by training with frequency domain feature information corresponding to blades in different blade states under multiple preset working modes as a training set. In this regard, by inputting the frequency domain feature information corresponding to the currently acquired luminous flux signal, the dicing machine can determine whether the current blade is damaged through the second training model. The corresponding detection process specifically includes steps S143-S144:

[0081] Step S143: Based on the second training model, feature recognition is performed on the frequency domain feature information to determine whether the signal features of the frequency domain feature information meet the signal features corresponding to the damage state.

[0082] Step S144: When the signal characteristics of the frequency domain characteristic information meet the signal characteristics corresponding to the damaged state, it is determined that the blade state of the blade is a damaged state.

[0083] It can be understood that after determining the frequency domain feature information corresponding to the current light flux signal, the frequency domain feature information is used as the input parameter of the second training model, and then the second training model is used to perform feature recognition on the frequency domain feature information, so as to determine whether the blade is damaged, that is, to determine whether the signal characteristics of the frequency domain feature information meet the signal characteristics corresponding to the damaged state. It can be imagined that the second training model obtained after training can identify the signal characteristics corresponding to the damaged state, and then after identifying the signal characteristics of the input parameters, it is determined whether the signal characteristics of the input parameters meet the signal characteristics corresponding to the damaged state, so as to achieve the effect of detecting whether the blade is damaged.

[0084] Furthermore, when the signal characteristics of the frequency domain feature information meet the signal characteristics corresponding to the damaged state, the dicing machine can determine that the blade state of the blade is a damaged state, that is, the current blade is damaged. It can be imagined that in the case of a damaged blade, the blade is rotating in the fifth operating mode, and accordingly, the acquired luminous flux signal will show an abnormal increase in periodic signal values. In this regard, the second training model can identify this feature and determine that the current blade state is a damaged state. In this regard, the dicing machine determines whether the blade is damaged based on the neural network model, which helps to improve the accuracy of the blade state judgment, and there is no need to frequently stop the machine to check whether the blade is damaged.

[0085] In one embodiment, in order to further avoid the occurrence of misidentification, the dicing machine is provided with a sensitivity value, that is, the damage detection is performed repeatedly through the second training model, so that when the number of times the damaged state is identified meets the sensitivity value, that is, the number is greater than or equal to the sensitivity value, it is determined that the current blade state is a damaged state.

[0086] Among them, the sensitivity value is related to the hardness parameter of the processed material, the cutting thickness and the properties of the tool. It can be understood that for the hardness parameter of the processed material, the harder the processed material is, the greater the risk of blade breakage, and the smaller the sensitivity value is accordingly. For example, when cutting hard and brittle materials such as wafers, the blade is very easy to break, so the sensitivity value can be set to 3 times; while when cutting easy-to-cut materials such as PCB (Printed Circuit Board), the blade is not easy to break, so the sensitivity value can be set to 6 times. For the cutting thickness, the greater the cutting thickness, the greater the risk of blade breakage, and the smaller the set sensitivity value. For example, the cutting thickness of the two batches of PCBs is different, such as the latter cutting thickness is thicker, and the corresponding sensitivity value is set to 5 times, or even smaller. As for the properties of the tool, the blades used by the dicing machine are divided into soft blades and hard blades. For example, a circular thin grinding wheel is called a soft blade, and a grinding wheel that combines a grinding wheel with an aluminum alloy tool holder is called a hard blade. For a soft blade, the risk of breakage is higher than that of a hard blade. For example, after cutting the same PCB twice, when a hard blade is used for cutting, the corresponding sensitivity value may be 6 times, while when a soft blade is used for cutting, the corresponding sensitivity value should be less than 6 times, such as set to 4 times.

[0087] In this regard, if it is determined for the first time that the signal characteristics of the frequency domain feature information meet the signal characteristics corresponding to the damaged state, the dicing machine performs feature recognition on the frequency domain feature information for multiple times in a row to re-determine the detection result. It can be imagined that after the second training model first determines that the frequency domain feature information is information that meets the signal characteristics corresponding to the damaged state, the frequency domain feature information is used as the target information, and then in subsequent continuous detections, if it is determined again that the target information meets the requirements, the number of recognitions is increased by 1 to update the number of recognitions. It should be noted that determining that the signal characteristics of the frequency domain feature information meet the signal characteristics corresponding to the damaged state is regarded as a successful recognition, and the statistics of the number of recognitions may include the first successful recognition, or may not include the first successful recognition, which can be set according to actual application requirements. Optionally, in one embodiment, the number of continuous detections is set with a corresponding threshold, and then in a limited number of continuous detections, the number of successful recognitions is accumulated, and finally it is determined whether the number of recognitions is greater than or equal to the sensitivity value. Also optionally, in one embodiment, a corresponding threshold is set for the number of continuous detections, and continuous detection is required until successful recognition is achieved before the corresponding number of times is counted. If the recognition is interrupted during the continuous detection process, the number of recognitions will no longer be counted. For example, the current number of recognitions is 2, and if the next recognition is successful, the number of recognitions will increase by 1, that is, the latest number of recognitions is 3; if the next recognition is unsuccessful, the continuous detection is terminated, and the final number of recognitions is 2.

[0088] When the number of recognitions meets the sensitivity value, it is finally determined that the signal characteristics of the frequency domain feature information meet the signal characteristics corresponding to the damaged state, that is, when the number of recognitions is greater than or equal to the sensitivity value, the current blade state is determined to be a damaged state. It is conceivable that, in one embodiment, after the continuous detection is completed, the number of recognitions is compared with the sensitivity value to determine whether the blade is damaged. Optionally, in another embodiment, it is also possible to compare the number of recognitions with the sensitivity value each time it is updated during the continuous detection process to more quickly identify whether the blade is damaged.

[0089] In this regard, the sensitivity value of blade breakage recognition can be set independently according to the material being processed, cutting thickness and tool properties, which helps to reduce the occurrence of false detection and missed detection, making blade breakage detection more accurate.

[0090] Figure 3 This is a schematic diagram of the structure of a blade damage detection device provided in an embodiment of the present application. The device can be applied to the above-mentioned dicing machine and is used to execute the blade damage detection method provided in the above-mentioned embodiment, and has the corresponding functional modules and beneficial effects of the execution method. As shown in the figure, the device includes a signal processing module 301, a working condition identification module 302, a light flux output module 303 and a damage detection module 304.

[0091] The signal processing module 301 is configured to obtain a light flux signal and perform data processing on the obtained light flux signal to obtain frequency domain characteristic information corresponding to the light flux signal;

[0092] The operating condition identification module 302 is configured to perform feature identification on the frequency domain feature information based on the first training model, and determine the current operating condition mode corresponding to the frequency domain feature information based on the signal characteristics of the frequency domain feature information, wherein the first training model is a neural network model that can identify different operating conditions by training with the frequency domain feature information corresponding to the preset multiple operating conditions as a training set;

[0093] The luminous flux output module 303 is configured to determine the luminous flux value under the current working mode according to a preset output strategy, and the output strategy generates the luminous flux value in different ways corresponding to different working modes;

[0094] The damage detection module 304 is configured to detect the wear of the blade based on the luminous flux value corresponding to the luminous flux signal when the working mode is the target mode; and / or,

[0095] The blade is detected for damage based on the second training model to determine the corresponding detection result. The second training model is a neural network model that can identify whether the blade state is a damaged state by training the corresponding frequency domain feature information of blades in different blade states under multiple preset working modes as a training set.

[0096] On the basis of the above embodiment, in the case of recording the frequency domain feature information in a spectrum diagram, the signal processing module 301 is specifically configured as follows:

[0097] Based on a preset denoising processing algorithm, the luminous flux signal is denoised to determine a denoised luminous flux signal;

[0098] The denoised luminous flux signal is converted from the time domain to the frequency domain, and the relationship between the frequency and amplitude of the luminous flux signal in the frequency domain is recorded to generate a spectrum diagram corresponding to the frequency domain characteristic information of the luminous flux signal.

[0099] On the basis of the above embodiment, the dicing machine is further provided with a motor and a water flow nozzle, the main shaft of the motor is connected to the blade and is used to drive the blade to rotate, and the water flow nozzle is used to spray water to the blade;

[0100] The luminous flux output module 303 is specifically configured as follows:

[0101] When the current working mode is a first type of mode, the peak value of the acquired luminous flux signal is used as the luminous flux value, and the first type of mode includes a first working mode associated with turning on the motor and a second working mode associated with turning off the motor;

[0102] When the current working mode is the second type of mode, the signal value of the acquired luminous flux signal at the last sampling moment in the time domain is used as the luminous flux value, and the second type of mode includes a third working mode associated with opening the water flow nozzle and a fourth working mode associated with closing the water flow nozzle;

[0103] When the current operating mode is the third type mode, based on the first window data and the second window data obtained according to the sampling window, the first luminous flux corresponding to the first window data and the second luminous flux corresponding to the second window data are obtained to determine the luminous flux value. The third type mode includes a fifth operating mode associated with turning on the motor and the water nozzle, and the first window data and the second window data respectively include signal values ​​corresponding to the luminous flux signal obtained according to the sampling window at multiple sampling times.

[0104] Based on the above embodiment, the luminous flux output module 303 is specifically configured as follows:

[0105] Obtain a first mean and a second mean respectively, and use the first mean as the first luminous flux and the second mean as the second luminous flux, wherein the first mean is the average value of signal values ​​corresponding to multiple sampling moments in the first window data, and the second mean is the average value of signal values ​​corresponding to multiple sampling moments in the second window data;

[0106] Determine a difference between the first luminous flux and the second luminous flux to determine a luminous flux difference;

[0107] The sum of the luminous flux difference and the luminous flux value at the previous moment is determined and used as the luminous flux value in the third mode.

[0108] Based on the above embodiment, the damage detection module 304 is specifically configured as follows:

[0109] Determining the amount of blade wear based on a difference between a preset mode threshold and a light flux value;

[0110] When the wear amount is greater than or equal to the preset wear threshold, it is determined that the blade state of the blade is a worn state.

[0111] Based on the above embodiment, the damage detection module 304 is specifically configured as follows:

[0112] Based on the second training model, feature recognition is performed on the frequency domain feature information to determine whether the signal feature of the frequency domain feature information satisfies the signal feature corresponding to the damage state;

[0113] When the signal characteristics of the frequency domain characteristic information meet the signal characteristics corresponding to the damaged state, it is determined that the blade state of the blade is a damaged state.

[0114] Based on the above embodiment, the dicing machine is provided with a sensitivity value, which is associated with the hardness parameter of the processed material, the cutting thickness and the tool properties. The damage detection module 304 is specifically configured as follows:

[0115] When it is determined for the first time that the signal characteristics of the frequency domain characteristic information meet the signal characteristics corresponding to the damage state, the frequency domain characteristic information is repeatedly identified, and the second training model is updated to identify the frequency domain characteristic information as the number of times of identification of the target information, and the signal characteristics of the target information meet the signal characteristics corresponding to the damage state;

[0116] When the number of identification times meets the sensitivity value, it is finally determined that the signal characteristics of the frequency domain characteristic information meet the signal characteristics corresponding to the damage state.

[0117] It is worth noting that in the embodiment of the above-mentioned device, the modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the modules are only for the convenience of distinguishing each other, and are not used to limit the protection scope of the embodiments of the present application.

[0118] Figure 4The schematic diagram of the structure of a dicing machine provided in one embodiment of the present application is used to execute the blade damage detection method provided in the above embodiment, and has the functional modules and beneficial effects corresponding to the execution method. As shown in the figure, the dicing machine also includes a processor 401, a memory 402, an input device 403 and an output device 404. The number of processors 401 can be one or more, and one processor 401 is taken as an example in the figure; the processor 401, the memory 402, the input device 403 and the output device 404 can be connected through a bus or other means, and the figure takes the connection through a bus as an example. The memory 402, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as the program instructions / modules corresponding to the blade damage detection method in the embodiment of the present application. The processor 401 executes the corresponding various functional applications and data processing by running the software programs, instructions and modules stored in the memory 402, that is, realizing the above-mentioned blade damage detection method.

[0119] The memory 402 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data recorded or created during use, etc. In addition, the memory 402 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 402 may further include a memory remotely arranged relative to the processor 401, and these remotely arranged memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0120] The input device 403 can be used to input corresponding digital or character information to the processor 401, and to generate key signal input related to the user settings and function control of the device; the output device 404 can be used to send or display key signal output related to the user settings and function control of the device.

[0121] An embodiment of the present application further provides a storage medium storing computer executable instructions, which, when executed by a processor, are used to perform relevant operations in the blade damage detection method provided in any embodiment of the present application.

[0122] Computer-readable storage media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0123] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0124] Note that the above are only preferred embodiments of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A blade damage detection method, characterized in that: Applied to a dicing machine, the dicing machine includes an optical transceiver component and a blade, the optical transceiver component is used to send and receive optical signals, and obtain a luminous flux signal after photoelectric conversion of the received optical signal, the blade damage detection method includes: Acquiring the luminous flux signal, and performing data processing on the acquired luminous flux signal to obtain frequency domain characteristic information corresponding to the luminous flux signal; Based on a first training model, feature recognition is performed on the frequency domain feature information to determine the current operating mode corresponding to the frequency domain feature information based on the signal characteristics of the frequency domain feature information, wherein the first training model is a neural network model that can identify different operating modes and is obtained by training with frequency domain feature information corresponding to a plurality of preset operating modes as a training set; Determine the luminous flux value under the current working mode according to a preset output strategy, wherein the output strategy generates the luminous flux value in different ways corresponding to different working modes; When the current operating mode is the target mode, the blade is detected for wear based on the luminous flux value corresponding to the luminous flux signal; and the blade is detected for damage based on the second training model to determine the corresponding detection result, wherein when the number of times the second training model determines that the blade state is a damaged state meets a preset sensitivity value, the blade state is finally determined to be a damaged state, and the second training model is a neural network model that is trained using frequency domain feature information corresponding to blades with different blade states under multiple preset operating modes as a training set and can be used to identify whether the blade state is a damaged state.

2. The blade damage detection method according to claim 1, characterized in that: In the case where the frequency domain characteristic information is recorded in a spectrum diagram, the step of obtaining the luminous flux signal and performing data processing on the obtained luminous flux signal to obtain the frequency domain characteristic information corresponding to the luminous flux signal includes: Based on a preset denoising processing algorithm, the luminous flux signal is subjected to signal denoising to determine a denoised luminous flux signal; The denoised luminous flux signal is converted from the time domain to the frequency domain, and the relationship between the frequency and amplitude of the luminous flux signal in the frequency domain is recorded to generate a spectrum diagram corresponding to the frequency domain characteristic information of the luminous flux signal.

3. The blade damage detection method according to claim 1, characterized in that: The dicing machine is also provided with a motor and a water flow nozzle, the main shaft of the motor is connected to the blade and is used to drive the blade to rotate, and the water flow nozzle is used to spray water to the blade; The luminous flux value under the current working mode is determined according to a preset output strategy, and the output strategy generates the luminous flux value in different ways corresponding to different working modes, including: When the current operating mode is a first type of mode, the peak value of the acquired luminous flux signal is used as the luminous flux value, and the first type of mode includes a first operating mode associated with turning on the motor and a second operating mode associated with turning off the motor; When the current working mode is the second type of mode, the signal value of the luminous flux signal at the last sampling moment in the time domain is used as the luminous flux value, and the second type of mode includes a third working mode associated with turning on the water flow nozzle and a fourth working mode associated with turning off the water flow nozzle; When the current operating mode is a third-type mode, based on the first window data and the second window data obtained according to the sampling window, the first luminous flux corresponding to the first window data and the second luminous flux corresponding to the second window data are obtained to determine the luminous flux value. The third-type mode includes a fifth operating mode associated with turning on the motor and the water nozzle, and the first window data and the second window data respectively include signal values ​​corresponding to the luminous flux signal obtained according to the sampling window at multiple sampling times.

4. The blade damage detection method according to claim 3, characterized in that: When the current operating mode is the third type mode, based on the first window data and the second window data acquired according to the sampling window, acquiring a first luminous flux corresponding to the first window data and a second luminous flux corresponding to the second window data to determine a luminous flux value, includes: Obtain a first mean and a second mean respectively, and use the first mean as the first luminous flux, and use the second mean as the second luminous flux, wherein the first mean is an average value of signal values ​​corresponding to multiple sampling moments in the first window data, and the second mean is an average value of signal values ​​corresponding to multiple sampling moments in the second window data; determining a difference between the first luminous flux and the second luminous flux to determine a luminous flux difference; The sum of the luminous flux difference and the luminous flux value at the previous moment is determined and used as the luminous flux value in the third mode.

5. The blade damage detection method according to claim 1, characterized in that: When the current working mode is the target mode, performing wear detection on the blade based on the luminous flux value corresponding to the luminous flux signal to determine the corresponding detection result includes: Determining the wear amount of the blade based on the difference between a preset mode threshold and the light flux value; When the wear amount is greater than or equal to a preset wear threshold, it is determined that the blade state of the blade is a worn state.

6. The blade damage detection method according to claim 1, characterized in that: When the working mode is the target mode, performing damage detection on the blade based on the second training model to determine a corresponding detection result includes: Based on the second training model, feature recognition is performed on the frequency domain feature information to determine whether the signal feature of the frequency domain feature information satisfies the signal feature corresponding to the damage state; When the signal characteristics of the frequency domain characteristic information meet the signal characteristics corresponding to the damaged state, it is determined that the blade state of the blade is a damaged state.

7. The blade damage detection method according to claim 6, characterized in that: The sensitivity value is related to the hardness parameter of the material being processed, the cutting thickness and the properties of the tool; The performing feature recognition on the frequency domain feature information based on the second training model to determine whether the signal feature of the frequency domain feature information satisfies the signal feature corresponding to the damage state includes: When it is determined for the first time that the signal characteristics of the frequency domain characteristic information meet the signal characteristics corresponding to the damage state, the frequency domain characteristic information is repeatedly identified, and the second training model is updated to identify the frequency domain characteristic information as the number of times of target information, and the signal characteristics of the target information meet the signal characteristics corresponding to the damage state; When the number of identifications meets the sensitivity value, it is finally determined that the signal characteristics of the frequency domain characteristic information meet the signal characteristics corresponding to the damage state.

8. A blade damage detection device, characterized in that: Applied to a dicing machine, the dicing machine includes an optical transceiver component and a blade, the optical transceiver component is used to send and receive optical signals, and obtain a luminous flux signal after photoelectric conversion of the received optical signal, and the blade damage detection device includes: a signal processing module, configured to acquire the luminous flux signal, and perform data processing on the acquired luminous flux signal to acquire frequency domain characteristic information corresponding to the luminous flux signal; A working condition identification module is configured to perform feature identification on the frequency domain feature information based on a first training model, so as to determine a current working condition mode corresponding to the frequency domain feature information based on signal characteristics of the frequency domain feature information, wherein the first training model is a neural network model that can identify different working conditions and is obtained by training with frequency domain feature information corresponding to a plurality of preset working conditions as a training set; A luminous flux output module, configured to determine the luminous flux value under the current working mode according to a preset output strategy, wherein the output strategy generates the luminous flux value in different ways corresponding to different working modes; a damage detection module, configured to perform wear detection on the blade based on the luminous flux value corresponding to the luminous flux signal when the operating mode is the target mode; and The blade is detected for damage based on the second training model to determine the corresponding detection result, wherein when the number of times the second training model determines that the blade state is a damaged state satisfies a preset sensitivity value, the blade state is finally determined to be a damaged state, and the second training model is a neural network model that is trained using frequency domain feature information corresponding to blades with different blade states under multiple preset working modes as a training set and can be used to identify whether the blade state is a damaged state.

9. A dicing machine, characterized in that: The dicing machine comprises: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the blade damage detection method according to any one of claims 1 to 7.

10. A storage medium storing computer executable instructions, characterized in that: The computer executable instructions are used to execute the blade damage detection method according to any one of claims 1 to 7 when executed by a processor.

Images