TWI690860B - Method of training artificial intelligence to estimate lifetime of storage device - Google Patents

Abstract

A method of training artificial intelligence to estimate a lifetime of a storage device is provided, which includes steps of: determining whether an operating parameter of the storage device executing a processing program on a plurality of bit values is greater than an operation threshold parameter value or not, if yes, decoding the bit values stored in the storage device by a decoder; determining whether the bit values stored in the storage device are successfully decoded by the decoder or not, if yes, classifying each of a plurality of memory units of the storage device in a strong correct region, a weak correct region, a strong error region or a weak error region; determining whether the number of the memory units falls within an allowable quantity range or not, if not, starting an artificial intelligence neural network system to use machine learning to estimate the lifetime of the storage device.

Description

Method for training artificial intelligence to estimate service life of storage device

The invention relates to a storage device, in particular to a method for training artificial intelligence to estimate the service life of a storage device.

Nowadays, memory applications are becoming more and more popular. In the course of use, some factors will cause internal damage to the memory due to factors such as the number of erases and writes, which will cause the error rate to rise, making non-volatile memory (non-volatile memory) reliable The degree of rapid decline has been reduced, so the reliability of non-volatile memory can be improved through reliability design technology, especially error correction technology, which can make the product more long-lived and stable.

In order to ensure the extension of the service life of the non-volatile memory, an error correction module is designed in the control circuit to correct the data read from the non-volatile memory to eliminate the errors caused by external factors in the non-volatile memory. Traditionally, the mainstream error correction coding is BCH (Bose-Chaudhuri-Hochquenghem) Code. The calculation speed of this coding is quite fast, and the correction capability will be stronger with more redundant bits. However, as the manufacturing technology of non-volatile memory becomes higher and higher, BCH encoding technology has been unable to provide sufficient correction capabilities, so it began to use the low density parity check code (LDPC) correction widely used in the field of communication Wrong technology, with powerful correction capabilities, has become a new trend in the storage field.

The technical problem to be solved by the present invention is to provide a method for training artificial intelligence to estimate the service life of a storage device in view of the deficiencies of the prior art, and a method for training artificial intelligence to estimate the service life of a storage device, which is suitable for storage devices. The storage device includes Multiple memory units. The method includes the following steps: (a) use the storage device to execute the data bit processing procedure on one or more bit values; (b) calculate the operating parameters of the storage device to execute the data bit processing procedure on the multiple bit values; (c ) Determine whether the operation parameter is less than the first operation critical parameter value, if yes, go to step (a), if not, go to the next step (d); (d) use the decoder to decode each bit value stored in the storage device; (e ) Determine whether the decoder successfully decodes the multiple bit values stored in the storage device, if so, perform step (a), if not, perform the next step (f); (f) define multiple storage status areas, multiple storage The status area includes a strong correct area, a weak correct area, a strong error area, and a weak error area; (g) According to the storage status of each memory unit, each memory unit is classified as a strong correct area, a weak correct area, a strong error area, or a weak error area ; (H) Calculate the number of memory cells classified into multiple memory cells in the storage status area; (i) Determine whether the number of memory cells is within the allowable range of the first number. If so, perform step (a). If not, execute the following One step (j); (j) Start the artificial intelligence neural network system, use machine learning to analyze whether the operating parameter is less than the second operating critical parameter value, and whether the decoder successfully decodes the multiple bit values stored in the storage device and remembers Whether the number of units is within the allowable range of the second quantity, if not, the service life of the storage device is estimated, and if so, step (a) is performed.

As described above, the method for training artificial intelligence provided by the present invention to estimate the service life of a storage device can be based on the operating parameter values of the storage device such as the number of reads and writes, the number of erasures and the busy time, and the storage device access data bits The error rate of the value (that is, the number of memory cells classified in a strong error area or other storage state), and the decoder's ability to correct the data bit value stored in the storage device, so as to effectively estimate the service life of the storage device.

In order to further understand the features and technical contents of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are for reference and explanation only, and are not intended to limit the present invention.

The following is a specific specific example to illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content provided in this specification. The present invention can be implemented or applied through other different specific embodiments. Various details in this specification can also be based on different viewpoints and applications, and various modifications and changes can be made without departing from the concept of the present invention. In addition, the drawings of the present invention are merely schematic illustrations, and are not drawn according to actual sizes, and are declared in advance. The following embodiments will further describe the related technical content of the present invention in detail, but the content provided is not intended to limit the protection scope of the present invention.

It should be understood that although terms such as “first”, “second”, and “third” may be used herein to describe various elements or signals, these elements or signals should not be limited by these terms. These terms are mainly used to distinguish one component from another component, or one signal from another signal. In addition, the term "or" as used herein may include any one or a combination of more than one of the associated listed items, depending on the actual situation.

[First embodiment]

Please refer to FIG. 1, which is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a first embodiment of the present invention. As shown in FIG. 1, the method for training artificial intelligence to estimate the service life of a storage device in this embodiment includes the following steps S101 to S119, which are suitable for storage devices such as solid state drives (Solid state drive), etc. The storage device includes multiple memories The unit is, for example, a memory cell.

In step S101, the storage device executes a data bit processing procedure on one or more bit values. For example, the storage device further includes a storage control unit configured to write one or more data bit values or a bit stream composed of multiple bit values to the memory unit, that is, to control the memory unit to access one or more Data bit values. Further, the storage control unit of the storage device can read the data bit value stored in the memory unit and control the memory unit to erase the stored data bit value.

In step S103, the operation parameters of the storage device executing the data bit processing procedure on one or more bit values include the number of times the data bit processing procedure is executed, the operation time of each execution of the data bit processing procedure, or a combination thereof.

In step S105, it is determined whether the operation parameter of the storage device is less than the first operation threshold parameter value. For example, it is determined whether the number of times the storage device executes the data bit processing procedure is less than the threshold of the first number of operations. Additionally or alternatively, it is determined whether the operation time of the storage device each time the data bit processing program is executed is less than the first operation time threshold.

If it is determined that the operation parameter of the storage device is less than the first operation threshold parameter value, for example, when the operation times of the storage device is less than the first operation threshold value and the time taken for each execution of the procedure is less than the first operation time threshold, the storage device is determined It still has good performance. Therefore, step S101 can be repeatedly executed, and the storage device can execute the data bit processing procedure again, for example, the memory unit can access the new data bit value, or can erase other data bit values stored in the memory unit.

Conversely, if it is determined that the operation parameter of the storage device is equal to or greater than the first operation threshold parameter value, for example, the operation count of the storage device is equal to or greater than the first operation threshold value and the time taken for each execution of the procedure is equal to or greater than the first When the operating time is critical, it is preliminarily estimated that the storage device may deteriorate due to the number of uses or other environmental factors, which will lead to the deterioration of the operating performance of the storage device, which will lead to the extension of the operating time, so that the service life of the storage device can be preliminarily estimated or even preliminarily estimated. The storage device is about to be damaged or damaged and cannot be used. In order to estimate the service life of the storage device more accurately, the next step S107 may be further executed.

In step S107, the decoder is used to decode one or more bit values stored in the memory unit of the storage device. Specifically, the decoder can simultaneously or sequentially decode multiple single bit values stored in the same or different memory units, and multiple bit values in the same bit stream. For example, the decoder may decode the multiple bit values sequentially according to the order in which the storage device stores the multiple bit values. Alternatively, if multiple memory cells are arranged in an array in the storage device, the order of decoding multiple bit values sequentially may be determined according to the row/column positions of the multiple memory cells arranged in the array.

In step S109, it is determined whether the decoder successfully decoded the bit value. In this embodiment, it is assumed that the decoder does not have the ability to invert the bit value stored in the memory unit of the storage device, that is, the decoder does not invert the bit value logic 1 stored in the memory unit to 0 or the bit value logic 0. Flip into logic 1. Under this condition, if the decoder fails to decode the bit value stored in the storage device, it is determined that the storage device may misjudge the bit value when accessing the data, for example, the bit value logic 0 is misjudged as logic 1 or logic 1 Misjudgment is logic 0, and the wrong bit value is stored. If this happens, step S101 can be repeated to use the storage device to re-access the data bit value.

Conversely, if the decoder successfully decodes the bit value stored in the storage device, it is determined that the storage device can correctly interpret the bit value and store the correct bit value when accessing the data bit value, and then preliminarily infer that the storage device still has Good performance. Therefore, step S111 is next performed.

In step S111, a plurality of storage state areas are defined. For example, the plurality of storage state areas may include more areas such as strong correct areas, weak correct areas, strong error areas, and weak error areas.

In step S113, according to the storage state of the memory unit, the classified memory unit belongs to a strong correct area, a weak correct area, a strong error area, or a weak error area. For example, when the correct rate of the judgment bit value of the memory unit is higher than a correct rate threshold, the memory unit is classified in the strong correct area. Conversely, when it is judged that the correct rate of the memory cell's interpretation bit value is lower than the correct rate threshold, the memory cell is weakly corrected. In addition, when it is judged that the error rate of the read bit value of the memory unit is higher than the error rate threshold, the memory unit is classified in the strong error area. Conversely, when it is judged that the error rate of the read bit value of the memory unit is lower than the error rate threshold, the memory unit is classified in the weak error area.

In step S115, the number of memory cells classified in any or each of the storage state areas such as the strong correct area, the weak correct area, the strong error area, and the weak error area is calculated.

In step S117, it is determined whether the number of memory cells classified in any one or each of the storage state areas such as the strong correct area, the weak correct area, the strong error area, and the weak error area is within the first number allowable range. If the number of memory units classified in any or each storage state area according to the storage state of the memory unit is within the preset first number allowable range, it is determined that the storage device can continue to be used. Conversely, when it is determined that the number of memory cells classified in the storage state area exceeds the preset first number allowable range, it is preliminarily estimated that the storage device may be unusable.

In step S119, an artificial intelligence neural network system is started, and the relevant values of the storage device are utilized by machine learning to more accurately analyze the service life of the storage device. The operation of the artificial intelligence neural network system is described in more detail in the second embodiment below.

[Second Embodiment]

Please refer to FIG. 2, which is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a second embodiment of the present invention. As shown in FIG. 2, the method for training artificial intelligence to estimate the service life of a storage device in this embodiment includes the following steps S201 to S215, which are applicable to a storage device, and the storage device includes multiple memory units. It should be understood that the steps of the multiple embodiments described herein can be implemented in combination according to actual needs.

After performing step S119 or step S203 to activate the artificial intelligence neural network system, the following steps S205 to S215 can be performed. That is to say, the storage device is tested for the first time in steps S103-S117, and the storage device is tested for the second time in steps S205-S215. By detecting the storage device twice, the storage can be accurately estimated The service life of the device.

Alternatively, as shown in FIG. 2, after performing the data bit processing procedure on the bit value in step S101 or S201, the storage device may omit steps S103~S117 and directly start the artificial intelligence neural network system to execute steps S205~S215 Estimate the service life of the storage device. It should be understood that the steps described in this document, such as steps S207, S209, and S213, may be executed in a swapped order or performed simultaneously according to actual needs.

In step S201, a data bit processing procedure is performed on the bit value using the storage device, such as writing the bit value into the memory unit, reading the bit value stored in the memory unit, and erasing the bit value stored in the memory unit, etc. operation.

In step S203, an artificial intelligence neural network system is activated.

In step S205, the artificial intelligence neural network system is used to analyze the second operation critical parameter value for different types of storage devices with different service lives using machine learning.

In step S207, machine learning is used to determine whether the operating parameter of the storage device is less than the second operating critical parameter value. In this embodiment, the second operation critical parameter value is less than the first operation critical parameter value, but the invention is not limited to this.

For example, the operation critical parameter value includes the operation time for accessing the data bit value. For example, if the storage device wants to access data with a bit amount within a specific bit amount range, the first operation critical parameter value is set to access The time is 0.5ms, and the second operation critical parameter value is set as the access time 0.2ms. First, in step S105, the time when the artificial intelligence neural network system uses machine learning to analyze the storage device to access the data bit value, for example, 0.3ms is less than the first operation critical parameter value, for example, 0.5ms. Next, in this step S203, the time when the artificial intelligence neural network system uses the machine learning analysis storage device to access the data bit value, for example, 0.3ms is less than the second operation critical parameter value, for example, 0.2ms.

In practice, different operation critical parameter values can be set according to different types of storage devices. In addition or alternatively, the maximum number of bits that can be accessed by the memory unit of the storage device, such as single-level cell (Single-Level Cell, SLC), can access 1 bit value, multi-level memory cell (Multi -Level Cell, MLC) can access 2 bit values, the fourth-order memory cell Quad-Level Cell (QLC) can access 4 bit values and the amount of data bits actually accessed to set the corresponding Operational critical parameter value.

If the artificial intelligence neural network system is used to analyze the operation parameters of the storage device using machine learning to be less than the second operation critical parameter value, for example, the operation times of the storage device are less than the second operation threshold value and the time taken for each execution of the procedure is less than At the second operating time threshold, it is determined that the storage device still has good performance and can be used continuously. Therefore, step S201 can be executed again.

Conversely, if the artificial intelligence neural network system is used to analyze the operating parameters of the storage device by using machine learning to be greater than or equal to the second operating critical parameter value, for example, the operating times of the storage device is greater than or equal to the second operating critical value and each time When the time taken to execute the procedure is greater than or equal to the second operating time threshold, it is preliminarily inferred that the storage device may no longer be used continuously or the number of re-operable times is not much. Further, to more accurately estimate the service life of the storage device, step S209 is then executed.

In step S209, an artificial intelligence neural network system is used to determine whether the decoder successfully decodes one or more bit values stored in the memory unit of the storage device using machine learning.

If the artificial intelligence neural network system is used to determine whether the decoder fails to decode the bit value stored in the memory unit using machine learning, step S201 is repeatedly executed. On the contrary, if the artificial intelligence neural network system is used to determine that the decoder can successfully decode the bit value stored in the memory unit through one or more decoding procedures, step S211 is executed.

In step S211, machine learning is used to analyze the second allowable range of storage devices for different types and having different service lives. The second quantity allowable range is smaller than the first quantity allowable range, but the invention is not limited thereto.

In step S213, use machine learning to determine whether the number of memory cells classified in any one or each of the storage status areas such as strong correct area, weak correct area, strong error area, and weak error area is within the second number allowable range .

In step S215, the artificial intelligence neural network system is started, and the machine learning is used to determine whether the storage device is currently usable according to the judgment results of steps S201~S213, and it is inferred whether it is out of use due to old or damaged, and whether it needs repair Or directly replace the new storage device. Moreover, the artificial intelligence neural network system is activated, and machine learning is used to execute steps S201-S213 again to more accurately estimate whether the storage device's service life is about to expire, for example, whether the service life time is less than the service life critical time, In order to roughly estimate whether the use device can still be used for at least half a year or more.

[Third Embodiment]

Please refer to FIG. 3, which is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a third embodiment of the present invention. As shown in FIG. 3, the method for training artificial intelligence to estimate the service life of a storage device in this embodiment includes the following steps S301 to S317, which are applicable to a storage device. The storage device includes multiple memory units. Steps S301 to S317 of this embodiment are further examples of steps S103 and S105 of the first embodiment and steps S205 and S207 of the second embodiment. It should be understood that the steps of the multiple embodiments described herein can be implemented in combination according to actual needs.

In step S301, the storage control module of the storage device is used to write the data bit value to the memory unit. For example, the same memory unit of the storage device can write only a single data bit value, or sequentially read and write multiple data bit values, or write a data bit stream composed of multiple data bit values. The data bit value is, for example, logic 0 or logic 1. It should be understood that the present invention is not limited to the number of bit values read by the memory unit.

In step S303, the number of reading and writing of the storage device is counted.

In step S305, it is compared whether the currently accumulated reading and writing times of the storage device is less than a preset threshold of reading and writing times. If the current accumulated reading and writing times of the comparison storage device is less than the preset threshold of reading and writing times, it is determined that the storage device can be used again to perform the operation of step S301. Conversely, if the current accumulated number of reads and writes of the compared storage device is greater than or equal to the preset threshold of read and write times, then further steps are performed to estimate the storage device more accurately through various operating parameters of the storage device Whether it is still usable.

In step S307, the storage control module of the storage device is used to erase one or more data bit values stored in the one or more memory units.

In step S309, the number of erasures of the data bit value stored in the storage control module of the storage device to erase the memory unit is counted.

In step S311, it is compared whether the current accumulated erasure times of the storage device is less than a preset erasure count threshold. If the current accumulated erasure count of the comparison storage device is less than the preset erasure count threshold, it is determined that the storage device can be used again to perform the operation of step S301. Conversely, if the current accumulated erasing count of the storage device is not less than the preset erasure count threshold, the subsequent step S313 must be executed to further determine the state of the storage device.

In step S313, the busy time of the storage device is counted. For example, counting the time that the storage device executes the data bit processing procedure on the data bit value includes the time for the storage control module to receive and transmit the data bit value to the memory unit, and the storage control module controls the memory unit to access the data bit The time of the meta value and the time of erasing the bit value stored in the memory unit.

In step S315, compare whether the busy time of the storage device is less than the busy time threshold. If the busy time of the comparison storage device is less than the busy time threshold, it is judged that the accumulated operating time of the storage device is not long, and it should still have good operating performance and step S301 can be repeated. Conversely, if the busy time of the comparison storage device is not less than the busy time threshold, it is determined that the storage device can still be used continuously to perform step S317.

In step S317, it is estimated whether the service life of the storage device will expire. In detail, if the current cumulative number of reads and writes in the comparison storage device is not less than the preset threshold of read and write times in step S305, and the current cumulative number of erasures in the comparison storage device in step S311 is not less than the preset erase Except for the critical number of times, and when the latest busy time of the storage device compared in step S315 is not less than the busy time critical value, it can be preliminarily estimated that the usable time of the storage device may not be long, that is, the useful life is about to expire.

[Fourth embodiment]

Please refer to FIG. 4, which is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a fourth embodiment of the present invention. As shown in FIG. 4, the method for training artificial intelligence to estimate the service life of the storage device in this embodiment includes the following steps S401 to S409, which are applicable to the storage device, and the storage device includes multiple memory units. Steps S401 to S403 of this embodiment are directed to steps S115 to S117 and step S407 of the first embodiment are further examples of step S213 of the first embodiment. It should be understood that the steps of the multiple embodiments described herein can be implemented in combination according to actual needs.

In step S401, according to the storage state of the data bit values of the plurality of memory units of the storage device, for example, according to the accuracy and error rate of the memory unit to judge the bit value, the plurality of memory units of the classification storage device belong to the same or different storage states Areas, such as strong correct areas, weak correct areas, strong error areas, weak error areas, and more. Next, the number of memory cells classified in the strong error area is calculated, that is, the number of memory cells whose error rate of the miscalculated bit value is higher than the error rate threshold is calculated.

In step S403, it is determined whether the number of memory cells classified in the strong error area is within the allowable range of the first strong error number, that is, whether it is less than the threshold value of the first strong error number. If the number of memory cells classified in the strong error area is within the allowable range of the first strong error number, that is, less than the threshold value of the first strong error number, it is determined that the storage device has a certain ability to correctly interpret the bit value, and Continue to use the storage device.

After the storage device continues to be used, the storage device's storage status, such as the ability to read, may be reduced due to the increased number of storage device usage, increased use time, environmental factors, etc. Classified into different storage status areas. For example, a memory unit that was originally classified into a weak error zone is reclassified into a strong error zone after a period of use. Therefore, if it is determined that the number of memory cells classified in the strong error area is within the allowable range of the first strong error number, step S401 may be executed again. If it is determined that the number of memory cells classified in the strong error area exceeds the allowable range of the first strong error number, that is, if the determination is greater than the threshold value of the first strong error number, the next step S405 is executed.

In step S405, the artificial intelligence neural network system is activated.

In step S407, the artificial intelligence neural network system is used to analyze whether the number of memory cells classified in the strong error region exceeds the allowable range of the second strong error number, that is, whether it is smaller than the threshold value of the second strong error number. The allowable range of the second strongest error quantity may be smaller than the allowable range of the first strongest error quantity, that is, the second strongest error quantity threshold may be smaller than the first strongest error quantity threshold.

If the artificial intelligence neural network system is used to analyze the number of memory units classified in the strong error area using machine learning analysis is within the allowable range of the second strong error number, that is, when the judgment is less than the threshold value of the second strong error number, after a period of time Step S401 can be executed again. Conversely, if the number of memory cells classified in the strong error area exceeds the allowable range of the second strong error number, that is, it is judged to be greater than the threshold value of the second strong error number, step S409 is executed.

In step S409, the artificial intelligence neural network system is used to estimate the service life of the storage device based on the number of memory cells in the strong error area and the difference from the threshold of the second strong error number using machine learning.

[Fifth Embodiment]

Please refer to FIG. 5, which is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a fifth embodiment of the present invention. As shown in FIG. 5, the method for training artificial intelligence to estimate the service life of a storage device in this embodiment includes the following steps S501 to S513, which are applicable to a storage device. The storage device includes multiple memory units. It should be understood that the steps of the multiple embodiments described herein can be implemented in combination according to actual needs.

Before starting the artificial intelligence neural network system, using other systems such as the system built in the storage control module of the storage device to initially perform steps S501 to S511 of this embodiment. Then, after the artificial intelligence neural network system is activated, steps S501 to S511 of this embodiment are executed again, wherein different allowable ranges of the second strongest error quantity and the second strongest error critical ratio have been replaced with the first strongest error quantity respectively Allowable range and critical ratio of the first strongest error. In another embodiment, after the artificial intelligence neural network system is started, steps S501 to S511 of this embodiment are executed.

In step S501, a plurality of memory cells of the storage device are used to perform a data bit processing procedure on a plurality of bit values.

In step S503, the number of strong errors of the memory cells classified in the strong error area is calculated.

In step S505, it is determined whether the number of strong errors is within the allowable range of the number of first strong errors. If it is determined that the number of strong errors is within the allowable range of the first strong errors, step S501 may be executed again. Conversely, if it is determined that the number of strong errors exceeds the allowable range of the first strong errors, step S507 is executed.

In step S507, the number of weak error memory cells classified into the memory cells in the weak error area is calculated.

， 其中，SER代表強錯誤比例， SE表示強錯誤區的記憶單元數量，WE表示弱錯誤區的記憶單元數量。 In step S509, the ratio of strong errors in the number of multiple memory units classified in the strong error area to the number of multiple memory units classified in the strong error area and the weak error area is calculated. The Strong Error Ratio (SER) is expressed as the following formula: SER=

, Where SER represents the proportion of strong errors, SE represents the number of memory units in the strong error area, and WE represents the number of memory units in the weak error area.

In step S511, it is determined whether the strong error ratio is smaller than the first strong error critical ratio. If it is determined that the strong error ratio is less than the first strong error critical ratio, step S501 may be executed again. Conversely, if the strong error rate is equal to or greater than the first strong error critical rate, step S513 is executed.

In step S513, an artificial intelligence neural network system is activated to estimate the service life of the storage device and more accurately estimate the service life of each memory unit of the storage device.

It should be understood that, in practice, step S505 may be omitted to perform a comparison operation between the number of memory cells classified in the strong error area and the allowable range of the number of first strong errors, while only performing step S511 to compare the ratio of strong errors and less than the first strong errors Critical proportion of operations.

[Sixth Embodiment]

Please refer to FIG. 6, which is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a sixth embodiment of the present invention. As shown in FIG. 6, the method for training artificial intelligence to estimate the service life of a storage device in this embodiment includes the following steps S601 to S609, which are applicable to a storage device, and the storage device includes a plurality of memory units. It should be understood that the steps of the multiple embodiments described herein can be implemented in combination according to actual needs.

In step S601, according to the storage state of the data bit values of the plurality of memory units of the storage device, for example, according to the accuracy and error rate of the memory unit to determine the bit value, the plurality of memory units of the classification storage device belong to the same or different storage states Areas include, for example, strong correct areas, weak correct areas, strong error areas, weak error areas, and more. Next, the number of strongly correct memory cells classified in the strongly correct area is calculated.

In step S603, it is determined whether the strong correct number of the memory cells classified in the strong correct area is greater than the first strong correct number threshold, for example, 8 memory cells. If it is judged that the strong correct number of memory units classified in the strong correct area, such as 10 memory units, is greater than the first strong correct number threshold, such as 8 memory units, it is judged that the storage device can continue to operate, and the storage device can be operated again when the storage device is subsequently operated Go to step S601. Conversely, if it is determined that the number of memory units classified in the strong correct area, such as 6 memory units, is less than or equal to the first strong correct number threshold, such as 8 memory units, step S605 is executed.

In step S605, the artificial intelligence neural network system is activated.

In step S607, use machine learning to analyze whether the number of strong correct numbers of the memory cells classified in the strong correct area is greater than the threshold value of the second strong correct number. For example, the second strongest correct quantity threshold value, for example, 5 memory units is less than the first strongest correct quantity threshold value, for example, 8 memory units.

If it is judged that the strong correct number of memory units classified in the strong correct area, for example, 6 memory units is greater than the second strong correct number threshold, for example, 5 memory units, the storage device is judged to be able to operate continuously, and can be re-executed when the storage device is subsequently operated Go to step S601. Conversely, if it is determined that the number of strong correct units of memory units classified in the strong correct area, for example, 4 memory units is less than or equal to the second strong correct number threshold, such as 5 memory units, step S609 is executed.

In step S609, it is estimated that the service life of the storage device is about to expire.

[Seventh Embodiment]

Please refer to FIG. 7, which is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a seventh embodiment of the present invention. As shown in FIG. 7, the method for training artificial intelligence to estimate the service life of a storage device in this embodiment includes the following steps S701 to S713, which are applicable to a storage device, and the storage device includes multiple memory units. It should be understood that the steps of the multiple embodiments described herein can be implemented in combination according to actual needs.

In step S701, a plurality of memory cells of the storage device are used to perform a data bit processing procedure on a plurality of bit values.

In step S703, the number of strongly correct memory cells classified in the strongly correct area is calculated.

In step S705, it is determined whether the number of strong correct numbers of the memory cells classified in the strong correct area is greater than the threshold value of the first strong correct number. If it is determined that the number of strong correct numbers of the memory cells classified in the strong correct area is greater than the threshold value of the first strong correct number, step S701 may be repeatedly executed. On the contrary, if it is determined that the number of strong correct numbers of the memory cells classified in the strong correct area is less than or equal to the threshold value of the first strong correct number, step S707 is executed.

In step S707, the number of weakly correct memory cells classified in the weakly correct area is calculated.

其中，SCR代表強正確比例，SC表示強正確區的記憶單元數量，WC表示弱正確區的記憶單元數量。 In step S709, the ratio of the number of memory units classified in the strong correct area to the number of memory units classified in the sum of the strong correct area and the weak correct area is calculated. Strong correct ratio (SCR) is expressed by the following calculation formula: SCR=

Among them, SCR stands for the strong correct ratio, SC stands for the number of memory units in the strong correct area, and WC stands for the number of memory units in the weak correct area.

In step S711, it is determined whether the strong correct ratio is greater than the first strong correct critical ratio. If it is determined that the strong correct ratio is greater than the first strong correct critical ratio, step S701 may be repeatedly executed. On the contrary, if it is determined whether the strong correct ratio is not greater than the first strong correct critical ratio, step S713 is executed.

In step S713, the artificial intelligence neural network system is activated to estimate the service life of the storage device based on the above determination result. Even, the artificial intelligence neural network system can be activated to perform the above steps S701 to S711 again, the threshold value of the first strong correct quantity and the threshold of the first strong correct quantity can be replaced with the second strong correct quantity threshold and the second Strong correct critical ratio.

As described above, the sixth embodiment calculates the strong error ratio, and the seventh embodiment calculates the strong correct ratio. Alternatively, in yet another embodiment, the ratio of strong correct errors of the number of memory units classified in the strong error area to the number of memory units classified in the strong correct area is calculated. Next, it is determined whether the proportion of strong correct errors is less than the critical ratio of strong correct errors, or whether the number of memory units classified in the strong error area is smaller than the number of memory units classified in the strong correct area. If it is judged that the proportion of strong correct errors is smaller than the critical ratio of strong correct errors, or if the number of memory cells classified in the strong error area is smaller than the number of memory cells classified in the strong correct area, step S701 may be repeatedly executed. On the contrary, step S713 is executed.

[Eighth Embodiment]

Please refer to FIG. 8, which is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to an eighth embodiment of the present invention. As shown in FIG. 8, the method for training artificial intelligence to estimate the service life of a storage device in this embodiment includes the following steps S801 to S817, which are applicable to a storage device, and the storage device includes multiple memory units. It should be understood that the steps of the multiple embodiments described herein can be implemented in combination according to actual needs.

In step S801, the decoder decodes the bit value stored in the memory unit of the storage device.

In step S803, it is determined whether the decoder successfully decodes the bit value stored in the memory unit of the storage device. In order to accurately estimate the service life of the storage device, the decoder can decode all or most of the bit values stored in multiple memory units of the storage device and record the time spent in each decoding, but the present invention does not use this Limited. In practice, in order to save the estimation time, only part of the bit values stored in the memory unit can be decoded.

If the decoder fails to decode the memory unit of the storage device, step S805 is executed: record the decoding time of the decoder bit value. Then, by repeating step S803, the decoder can flip the storage device to store the wrong bit value with a certain probability, or the decoder can decode the bit value re-accessed by the storage device. On the contrary, if the decoder successfully decodes the memory unit of the storage device, step S807 is executed.

In step S807, it is determined whether the decoding time of the decoder's current decoding bit value is less than the preset first decoding critical time.

In step S809, the artificial intelligence neural network system is activated.

In step S811, an artificial intelligence neural network system is used to determine whether the decoding time is less than a preset second decoding critical time using machine learning analysis. This second decoding critical time may be less than the first decoding critical time.

In step S813, the artificial intelligence neural network system is used to estimate the correction ability of the decoder to the storage device based on the above judgment result using machine learning.

In detail, under ideal circumstances, the decoder can perform a one-bit value correction procedure when decoding the bit value stored in the memory unit of the storage device, and flip the error bit value stored in the memory unit with a specific probability The correct bit value, for example, the storage device wants to store the bit value of logic 0 is misjudged and stored as a logic 1 during access, the decoder executes the bit value correction process to flip the logic 1 back to the bit value of logic 0 to achieve Correct the bit value and store the correct bit value in the memory unit of the storage device.

However, if the performance of the storage device is too poor, the storage device misjudges a large number of data bit values or stores invalid data, that is, the error rate of the miscalculated bit value is higher than the error rate threshold, or the decoder performance becomes too poor, resulting in exceeding the decoder The load level of the correction capability or the decoder does not have a good correction capability, resulting in the decoder cannot effectively correct the erroneous data stored in the storage device, it can be estimated that the service life of the decoder or storage device will expire, and a new one needs to be replaced. Decoder, storage, or both.

In step S815, an artificial intelligence neural network system is used to define a continuous use group and a correction limit group using machine learning.

In step S817, when the artificial intelligence neural network system is used to analyze the storage device according to the above estimation result using machine learning and the storage device can be used continuously, the storage device is classified into the continuous use group. On the contrary, if it is judged that the storage device can no longer be used, the storage device is classified into the correction limit group.

[Beneficial effect of embodiment]

In summary, the method for training artificial intelligence provided by the present invention to estimate the service life of a storage device can be based on the operating parameter values of the storage device such as the number of reads and writes, the number of erasures and the busy time, and the storage device access data bits The error rate of the meta value (that is, the number of memory cells classified in a strong error area or other storage state), and the decoder's ability to correct the data bit value stored in the storage device, to effectively estimate the service life of the storage device.

The content provided above is only a preferred and feasible embodiment of the present invention, and therefore does not limit the scope of the patent application of the present invention, so all equivalent technical changes made by using the description and drawings of the present invention are included in the application of the present invention. Within the scope of the patent.

S101~S119, S201~S215, S301~S317, S401~S409, S501~S513, S601~S609, S701~S713, S801~S817: Steps

FIG. 1 is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a first embodiment of the invention.

2 is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a second embodiment of the invention.

3 is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a third embodiment of the present invention.

4 is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a fourth embodiment of the present invention.

5 is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a fifth embodiment of the present invention.

6 is a flowchart of steps in a method for training artificial intelligence to estimate the service life of a storage device according to a sixth embodiment of the invention.

7 is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to a seventh embodiment of the present invention.

8 is a flowchart of steps of a method for training artificial intelligence to estimate the service life of a storage device according to an eighth embodiment of the present invention.

S201~S215: Steps

Claims (8)

A method for training artificial intelligence to estimate the service life of a storage device is suitable for a storage device. The storage device includes a plurality of memory units. The method for training artificial intelligence to estimate the service life of a storage device includes the following steps: (a) Use the storage device to execute a data bit processing procedure for one or more bit values; (b) Calculate an operating parameter of the data bit processing program performed by the storage device on the multiple bit values; (c) Determine whether the operation parameter is less than a first operation critical parameter value, if yes, go to step (a), if not, go to the next step (d); (d) use a decoder to decode each bit value stored in the storage device; (e) Determine whether the decoder successfully decodes the multiple bit values stored in the storage device, if so, perform step (a), if not, perform the next step (f); (f) Define multiple storage status areas, which include a strong correct area, a weak correct area, a strong error area, and a weak error area; (g) According to the storage status of each memory unit, classify each memory unit as belonging to the strong correct area, the weak correct area, the strong error area, or the weak error area; (h) Calculate the number of one memory unit of the multiple memory units classified in the storage state area; (i) Determine whether the number of the memory unit is within a permitted range of the first number, if yes, go to step (a), if not, go to the next step (j); (j) Activate artificial intelligence neural network system, use machine learning to analyze whether the operating parameter is less than a second operating critical parameter value, whether the decoder successfully decodes the multiple bit values stored in the storage device and the memory Whether the number of units is within a second number allowable range, if so, perform step (a), if not, estimate the service life of the storage device. The method for training artificial intelligence as described in item 1 of the patent application scope to estimate the service life of a storage device further includes the following steps: (k) Count the number of reads and writes by the storage device to read and write the multiple bit values; (l) Compare whether the number of reads and writes is less than a critical value of the number of reads and writes. If yes, go to step (a); if not, go to the next step (m); (m) Count the number of times the storage device erases the multiple bit values; (n) Compare whether the number of erasure times is less than the critical value of the number of erasure times. If yes, go to step (a); if not, go to the next step (o); (o) counting a busy time for the storage device to read, write or erase the multiple bit values; and (p) Compare whether the busy time is less than a busy time threshold, if so, perform step (a), if not, perform step (d); The operation parameters include the number of reads and writes, the number of erases, and the busy time. The method for training artificial intelligence as described in item 1 of the patent application scope to estimate the service life of a storage device further includes the following steps: (q) Calculate the number of strong errors of the memory unit classified in the strong error area; (r) Determine whether the number of strong errors is within the allowable range of the number of first strong errors. If yes, go to step (a); if not, go to the next step (s); (s) Activate the artificial intelligence neural network system, use machine learning to analyze whether the number of strong errors is within the allowable range of the number of the second strongest error, if so, perform step (a), if not, estimate the storage device’s Life time; and The number of memory units includes the number of strong errors, the first number allowed range includes the first strong error number allowed range, and the second number allowed range includes the second strong error number allowed range. The method for training artificial intelligence as described in item 1 of the patent application scope to estimate the service life of a storage device further includes the following steps: (t) Calculate the number of the multiple memory units classified in the strong error area, accounting for a strong error ratio of the number of the multiple memory units classified in the strong error area and the weak error area; (u) determine whether the strong error rate is less than a first strong error critical rate, if so, proceed to step (a), if not, proceed to the next step (v); and (v) Activate the artificial intelligence neural network system, use machine learning to analyze whether the strong error rate is less than a second strong error critical rate, if so, perform step (a), if not, estimate the life time of the storage device . The method for training artificial intelligence to estimate the service life of a storage device as described in item 1 of the patent scope further includes the following steps: (w) Calculate a strong correct number of the memory cells classified in the strong correct area; (x) determine whether the strong correct number is greater than a first strong correct number threshold, if so, go to step (a), if not, go to the next step (y); and (y) Activate the artificial intelligence neural network system, use machine learning to analyze whether the strong correct number is greater than a second strong correct number threshold, if so, perform step (a), if not, estimate the service life of the storage device time; The number of memory cells includes the strong correct number. The method for training artificial intelligence as described in item 1 of the patent application scope to estimate the service life of a storage device further includes the following steps: (z) Calculate the number of memory units classified in the strong correct area, accounting for a strong correct proportion of the number of memory units classified in the sum of the strong correct area and the weak correct area; (aa) determine whether the strong correct ratio is greater than a first strong correct critical ratio, if so, perform step (a), if not, perform the next step (bb); and (bb) Activate the artificial intelligence neural network system, use machine learning to analyze whether the strong correct ratio is greater than a second strong correct critical ratio, if so, perform step (a), if not, estimate the life time of the storage device . The method for training artificial intelligence as described in item 1 of the patent application scope to estimate the service life of a storage device further includes the following steps: (cc) Determine whether the decoder successfully decodes the one or more bit values stored in the storage device, if yes, go to step (a), if not, go to the next step (dd); (dd) Record a decoding time for the decoder to decode each bit value; and (ee) Determine whether the decoding time is less than a first decoding critical time, if yes, go to step (a), if not, go to the next step (ff); (ff) Activate the artificial intelligence neural network system, use machine learning to analyze whether the decoding time is less than the second decoding critical time, if so, perform step (a), if not, estimate the storage of the decoder to the storage device The ability to correct bit values. The method for training artificial intelligence as described in item 1 of the patent application scope to estimate the service life of a storage device further includes the following steps: (gg) Activate artificial intelligence neural network system, use machine learning to define a continuous use group and a correction limit group; and (hh) Use machine learning to analyze whether the life time of the storage device is greater than a service life critical time. If so, classify the storage device to the continuous use group. If not, classify the storage device to the correction limit group.

Images