Adaptive device vibration data lossless compression method and system based on dynamic bit width

Abstract

The application discloses a dynamic bit-width-based adaptive device vibration data lossless compression method and system, and particularly relates to the field of vibration data processing, and comprises the following steps: acquiring device vibration sequences collected by edge nodes; reading a first vibration sample in a sampling order; determining a first non-sign bit sequence of the first vibration sample and recording it as a current bit width; and establishing a current section based on the first vibration sample; reading a next vibration sample in the sampling order; determining a first non-sign bit sequence of the next vibration sample; and writing the next vibration sample into the current section according to the current bit width when the first non-sign bit sequence is not higher than the current bit width. The application separates and losslessly compresses high-bit information that can be continued across samples and low-bit information that changes locally and temporarily, so as to solve the problem that a small number of abnormal sampling points or short-time change segments continuously raise the dynamic bit width of a section in the existing device vibration data lossless compression process.

Description

Technical Field

[0001] This invention relates to the field of vibration data processing technology, and more specifically, to a method and system for lossless compression of adaptive device vibration data based on dynamic bit width. Background Technology

[0002] In the lossless compression processing of equipment vibration data, existing solutions mainly focus on compressing the data volume without changing the sampled values ​​and sampling order. In actual implementation, edge computing nodes write the vibration sampled values ​​continuously output by the sensor into a fixed word length, or first count the range of values ​​occupied by continuous sample segments, and then configure the matching dynamic bit width for the corresponding sample segments to complete the compression and storage. For example, when the main shaft bearings, gearboxes, or industrial motors of a wind turbine are in continuous operation, the edge side needs to continuously collect, compress, and write high-frequency vibration data to disk. This must be done while being limited by terminal computing power, cache capacity, and transmission capabilities, and ensuring that the compressed results can be restored intact for subsequent spectrum analysis, fault backtracking, and status diagnosis. In this situation, many adjacent sampled values ​​often only show slight changes in the low-order bits while the high-order bits remain unchanged over a continuous period. However, existing compression mechanisms still repeatedly write this high-order portion along with each sampled value. Furthermore, when a small number of impact sampling points are mixed in during this continuous period... When there are local disturbance sampling points or short-term fine change segments, the bit width configuration of the entire data segment will be raised as a whole, causing a large number of background samples that could have been stored at a lower bit width to be widened synchronously. In practice, this manifests as a significant drop in compression ratio after an impact or disturbance occurs. When the compression segment is shortened, additional control information increases, and when the compression segment is lengthened, it is more easily widened by a small number of abnormal samples. The root cause of the above problems is that the existing compression organization still uses the integer range as the basis for dynamic bit width allocation. It neither utilizes the high bit information that can be directly continued between continuous samples, nor does it organize the low bit active information that only appears locally and for a short time separately. Therefore, the technical problem to be solved by this application is: how to perform separate lossless compression organization of high-level information that can continue across samples and low-level information that changes locally in a short time in equipment vibration data under edge computing conditions, so as to avoid a small number of abnormal sampling points or short-time change segments from continuously increasing the dynamic bit width of the whole segment. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an adaptive lossless compression method and system for device vibration data based on dynamic bit width. This method solves the problems mentioned in the background art by separating and losslessly compressing high-bit information that can continue across samples and low-bit information that changes locally in a short time in device vibration data.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a lossless compression method for adaptive device vibration data based on dynamic bit width, comprising: S1. Obtain the device vibration sequence collected by the edge node, read the first vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the first vibration sample and record it as the current bit width, and establish the current segment with the first vibration sample; S2. Read the next vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the next vibration sample, and write the first non-signed valid bit sequence into the current segment according to the current bit width if the first non-signed valid bit sequence is not higher than the current bit width. If the first non-signed valid bit sequence is higher than the current bit width, rewrite the current bit width to the first non-signed valid bit sequence and then write it into the current segment. S3. During the current segment writing process, check the value of each vibration sample in the corresponding bit of the current bit width according to the sampling order. When the value is zero, it is recorded as empty; when the value is one, it is recorded as occupied, and a high-bit state sequence is formed. S4. Starting from the end of the current segment, check the high-level state sequence backwards. Locate the empty sample intervals at the end. Delete the corresponding position of the current position width for each vibration sample in the sample interval and recalculate the value. When the recalculation result is consistent with the value before deletion, record the starting position of the sample interval as the width reduction position and decrement the current position width by one as the subsequent segment width. S5. At the width reduction position, end the current segment, write the vibration samples before the width reduction position into the current compression segment according to the current bit width, create the subsequent segment according to the bit width of the subsequent segment, and repeat the vibration sample writing, high bit state formation and width reduction position determination to output the lossless compression result.

[0005] In a preferred embodiment, S1 includes: S1-1. Read the sign bit and value bit of the first vibration sample. Keep the value bit unchanged when the sign bit is zero. When the sign bit is one, invert the value bit bit by bit and output the judgment bit string corresponding to the first vibration sample. S1-2. Check the value of each bit from the high end to the low end of the judgment bit string, determine the bit sequence corresponding to the first bit with a value of 1 as the first non-signed valid bit sequence, and record the first non-signed valid bit sequence as the current bit width; S1-3. Extract the numerical bits of the first vibration sample according to the current bit width and write them together with the sign bit into the first sample position. Establish the current segment with the first sample position, the current bit width and the sample count value.

[0006] In a preferred embodiment, S2 includes: S2-1. Read the sign bit and value bit of the next vibration sample. When the sign bit is zero, keep the value bit unchanged. When the sign bit is one, invert the value bit bit bit by bit. Output the judgment bit string corresponding to the next vibration sample. Check the value of each bit bit by bit from the high bit end to the low bit end of the judgment bit string. Determine the bit sequence corresponding to the first bit with a value of one as the first non-sign valid bit sequence. S2-2. When the first non-signed valid bit sequence is not higher than the current bit width, extract the numerical bits of the next vibration sample according to the current bit width and write them together with the sign bit to the end of the current segment, and output the current segment after writing.

[0007] In a preferred embodiment, S2 further includes: S2-3. When the first non-signed valid bit sequence is higher than the current bit width, calculate the bit difference between the first non-signed valid bit sequence and the current bit width, and add the same extension bit as the sign bit of each vibration sample to the high bit end of each vibration sample written in the current segment according to the bit difference. S2-4. Rewrite the current bit width as the first non-signed valid bit sequence, then extract the numerical bits of the next vibration sample according to the rewritten current bit width and write them together with the sign bit to the end of the current segment, and output the expanded current segment.

[0008] In a preferred embodiment, S3 includes: S3-1. When writing each vibration sample in the current segment, synchronously read the value of each vibration sample in the corresponding bit of the current bit width. When the value is zero, generate an empty state value; when the value is one, generate an occupied state value; and output the state value sequence arranged in the sampling order. S3-2. Compare adjacent state values ​​in the state value sequence according to the sampling order. When adjacent state values ​​are the same, keep the state interval to which the current position belongs unchanged. When adjacent state values ​​are different, take the previous position as the end position of the state interval to which the current position belongs, and take the next position as the start position of the next state interval. Output the state value, start position and end position of each state interval. S3-3. Write the state value, start position and end position in the order of the start position of each state interval to form the high-order state sequence corresponding to the current segment.

[0009] In a preferred embodiment, S4 includes: S4-1. Read the high-order state sequence from the end of the current segment backward. When the state corresponding to the end position is occupied, record the position after the end of the current segment as the width-preserving position and maintain the current bit width. When the state corresponding to the end position is empty, continue reading forward until the first occupied position is read. Then, record the position after the occupied position as the empty start position and the end of the current segment as the empty end position. Use the empty start position and the empty end position to determine the empty interval. S4-2. Read each vibration sample in the order from the start position to the end position of the empty space. After deleting the corresponding bit of the current bit width for each vibration sample, recalculate the value. If the recalculated value is consistent with the value before deletion, it is recorded as a preserved value. If the recalculated value is inconsistent with the value before deletion, it is recorded as a changed value, and the sample result is formed.

[0010] In a preferred embodiment, S4 further includes: S4-3. Read two adjacent vibration samples in the order from the start position to the end position of the empty space. After deleting the corresponding bit of the current bit width for each vibration sample, recalculate the difference and the size relationship. When the difference and the size relationship are the same before and after deletion, it is recorded as adjacency maintenance. When the difference or the size relationship is inconsistent before and after deletion, it is recorded as adjacency mismatch, and the adjacency result is formed. S4-4. Starting from the empty end position, check the sample results and adjacency results in reverse from the empty start position. If the sample result corresponding to the current position is preserved and the adjacency result between the current position and the next position is maintained, continue the reverse check. If the sample result corresponding to the current position is a variable value, record the current position as the blocking position. If the adjacency result between the current position and the next position is an adjacency mismatch, record the next position as the blocking position. If the blocking position exists, record the blocking position as the width reduction position. Otherwise, record the empty start position as the width reduction position and reduce the current bit width by one and record it as the subsequent segment bit width.

[0011] In a preferred embodiment, S5 includes: S5-1. Read the current segment, the width reduction position, the current bit width, and the bit width of the subsequent segment. Record the position before the width reduction position as the end position of the current segment. Determine the vibration sample corresponding to the current segment start position to the current segment end position as the current compression segment sample. Determine the vibration sample corresponding to the current segment end position from the width reduction position to the current segment end as the subsequent segment sample. S5-2. Read the samples of the current compressed segment in the order from the start position to the end position of the current segment. Extract the numerical bits of each vibration sample according to the current bit width and write them together with the sign bits into the current compressed segment. At the same time, write the start position of the current segment, the end position of the current segment, the current bit width and the number of samples, and output the current compressed segment.

[0012] In a preferred embodiment, S5 further includes: S5-3. Using the width reduction position as the starting position of the subsequent segment, extract the numerical bits of the vibration sample corresponding to the width reduction position according to the width of the subsequent segment and write them together with the sign bits into the first sample position of the subsequent segment. At the same time, write the starting position of the subsequent segment, the width of the subsequent segment, and the first sample position, and output the subsequent segment. S5-4. Read the vibration samples after the width reduction position according to the sampling order. Repeat the vibration sample writing, high-level state formation and width reduction position determination for subsequent segments. When the end of the equipment vibration sequence is read, write the start position, end position, width of subsequent segments and number of samples into the lossless compression result. Combine the current compression segment and subsequent segments according to the start position of each compression segment.

[0013] In a preferred embodiment, an adaptive device vibration data lossless compression system based on dynamic bit width includes: The bit width determination module is used to acquire the device vibration sequence collected by the edge node, read the first vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the first vibration sample and record it as the current bit width, and establish the current segment with the first vibration sample. The bit width extension module reads the next vibration sample according to the sampling order, determines the first non-signed valid bit sequence of the next vibration sample, and writes it into the current segment according to the current bit width when the first non-signed valid bit sequence is not higher than the current bit width. When the first non-signed valid bit sequence is higher than the current bit width, the current bit width is rewritten to the first non-signed valid bit sequence and then written into the current segment. The status flag module is used to check the value of each vibration sample in the corresponding bit of the current bit width according to the sampling order during the current segment writing process. When the value is zero, it is marked as empty; when the value is one, it is marked as occupied, and a high-bit status sequence is formed. The bit width reduction module is used to check the high-level state sequence from the end of the current segment backwards, locate the empty sample intervals at the end, delete the corresponding bit of the current bit width for each vibration sample in the sample interval, and recalculate the value. When the recalculation result is consistent with the value before deletion, the starting position of the sample interval is recorded as the reduction position, and the current bit width is reduced by one and recorded as the bit width of the subsequent segment. The segmented compression module is used to end the current segment at the width reduction position, write the vibration samples before the width reduction position into the current compression segment according to the current bit width, create the subsequent segment according to the bit width of the subsequent segment for the vibration samples at the width reduction position and after, and repeat the vibration sample writing, high bit state formation and width reduction position determination to output lossless compression results.

[0014] The technical effects and advantages of this invention are as follows: By writing according to the current bit width, performing bit deletion and recalculation in the empty interval at the end, and determining the width reduction position, the high bit information that can be continued across samples is separated from the local short-term change part, thus relatively alleviating the problem of a small number of abnormal sampling points or short-term change segments continuously raising the dynamic bit width of the whole segment. By comparing the first non-signed valid bit sequence of the next vibration sample with the current bit width, and when the limit is exceeded, the vibration samples already written in the segment are first extended with extended bits before the current bit width is rewritten. This relatively suppresses the truncation distortion caused by local high values ​​during the sample writing process, while maintaining the connection of the encoding bit width within the segment. By generating empty and occupied states around the corresponding position of the current bit width, and compressing adjacent positions with the same state into a high-bit state sequence, the processing burden of repeatedly checking the entire vibration sample point by point is relatively reduced, while providing a direct basis for locating the empty section at the end. By recalculating the values ​​of vibration samples in the empty interval after deletion, and combining the recalculation of the difference between adjacent vibration samples and the recalculation of the magnitude relationship, the width reduction position is constrained. Thus, the risk of false width reduction caused by directly reducing the width based solely on the empty state at the end is relatively reduced. By ending the current segment at the width reduction position, writing the samples before the width reduction position as the current compressed segment, and continuing to build segments for the samples after the width reduction position according to the subsequent segment width, the bit width fitting degree under different fluctuation ranges is relatively improved, and the storage overhead caused by high bit width staying for a long time is reduced. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method steps of the present invention.

[0016] Figure 2 This is a schematic diagram of the system modules of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Refer to the instruction manual appendix Figure 1-2 The present invention provides a lossless compression method for adaptive device vibration data based on dynamic bit width, comprising: S1. Obtain the device vibration sequence collected by the edge node, read the first vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the first vibration sample and record it as the current bit width, and establish the current segment with the first vibration sample; In this implementation, the first vibration sample is used to provide the initial bit width and the basis for writing the first sample to the current segment. Therefore, this part not only needs to determine the first non-signed valid bit sequence, but also needs to establish the segment fields to be called in subsequent writing, bit expansion, high-bit state formation, and width reduction processing. During processing, the first vibration sample is first read into fixed-length binary, with the highest bit as the sign bit and the remaining bits as the value bits. The bit sequence is numbered sequentially from the side closest to the sign bit to the side furthest from the sign bit, and subsequent references are made in this numbering manner. For the first vibration sample with a sign bit of zero, the value bits are directly used to determine the first non-signed valid bit sequence. For the first vibration sample with a sign bit of one, the value bits are inverted bit by bit before being used to determine the first non-signed valid bit sequence. The inversion action is only used to locate the first non-signed valid bit and does not participate in the subsequent first sample writing. After the first non-signed valid bit sequence is determined, this bit sequence is used as the current bit width, and the first sample writing and current segment establishment are completed accordingly. The implementation process includes the following steps: In S1-1, the original binary data of the first vibration sample is first received from the edge node and split into one sign bit and multiple value bits. When the sign bit is zero, the values ​​and order of the value bits remain unchanged, and the value bits are directly used as the basis for judgment. When the sign bit is one, each bit is inverted according to the original order of the value bits, changing zero to one and one to zero. The inverted bits are then rearranged in their original order to form the judgment bit string. For example, if the first vibration sample is represented by one sign bit and seven value bits, when the sign bit is zero and the value bits are 0010110, the judgment bit string is still 0010110. When the sign bit is one and the value bits are 0010110, the judgment bit string becomes 1101001. After this processing, the first vibration samples with different signs can enter the same bit order judgment process. In S1-2, starting from the bit closest to the sign bit in the decision bit string, the values ​​are checked bit by bit in the order of the bit sequence, moving away from the sign bit. When the first bit is found to be 1, the bit sequence corresponding to that bit is determined as the first non-sign valid bit sequence, and the first non-sign valid bit sequence is recorded as the current bit width. If all bits in the decision bit string are zero, the bit sequence corresponding to the bit furthest from the sign bit in the numerical bits is recorded as the current bit width, so that the first vibration sample still occupies a definite write bit width and subsequent processing can continue. For example, when the decision bit string is 0010110, the first bit with a value of 1 appears in the third check position, so the bit sequence corresponding to the third check position is the current bit width. When the decision bit string is 0000000, the bit sequence corresponding to the last bit of the numerical bits is recorded as the current bit width. After this process, the first vibration sample can obtain a definite current bit width regardless of whether the value is zero, positive, or negative. In S1-3, the format for writing the first sample is determined by the current bit width. During writing, the original sign bit of the first vibration sample is retained, and the numerical bits corresponding to the current bit width are truncated from the original numerical bits and written to the first sample position. The truncating range is from the least significant bit of the numerical bits to the corresponding bit of the current bit width. Bits higher than the corresponding bit of the current bit width are not written to the first sample position. The sign bit and the truncated numerical bits are combined in the original order to form the first sample writing value, and the first sample writing value is written to the first sample position of the current segment. After the first sample is written, the first sample position is used as the start position of the current segment, the current bit width is used as the current segment bit width, the first sample position is used as the end position of the current segment, and the sample count value of one is used as the initial count of the current segment, thus establishing the current segment. The current segment established in this way includes the first sample writing result, as well as the start position, end position, current bit width, and sample count value directly called in subsequent steps. Through the above processing, the first vibration sample is first converted into a bit string that can be used for judgment, and then the current bit width is determined by checking bit by bit. Then, the first sample is written according to the current bit width and the current segment is formed. When subsequent vibration samples enter the current segment, they can be directly written or expanded based on the current bit width. When performing the current bit width corresponding bit check, empty interval location and width reduction position judgment on the current segment, the start position of the current segment, the end position of the current segment, the current bit width and the sample count value can be directly called. In practical applications: When the edge node samples the vibration signal of the motor bearing housing, if the sign bit of the original binary data of the first vibration sample is zero and the value bit is 00010110, the judgment bit string is kept as 00010110. After checking from the high bit end to the low bit end, the bit sequence corresponding to the first bit with a value of one is recorded as the current bit width. Then, the original sign bit is retained and the value bit corresponding to the current bit width is truncated and written to the first sample position. At the same time, the first sample position is recorded as the start position and end position of the current segment, and the sample count value is recorded as one. If the sign bit of the original binary data of the first vibration sample is one and the value bit is 00010110, the value bit is first inverted to obtain the judgment bit string 11101001, and then the current bit width is determined in the same way. When writing the first sample, the original sign bit and the original value bit truncated result are still used for writing. In this way, the processing of the first vibration sample entering the current segment is completed, and the starting bit width and the recording basis within the segment are provided for the continuous compression of subsequent vibration samples.

[0019] S2. Read the next vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the next vibration sample, and write the first non-signed valid bit sequence into the current segment according to the current bit width if the first non-signed valid bit sequence is not higher than the current bit width. If the first non-signed valid bit sequence is higher than the current bit width, rewrite the current bit width to the first non-signed valid bit sequence and then write it into the current segment. In this implementation, after the next vibration sample enters the current segment, it is necessary to first determine whether the next vibration sample still falls within the coverage range of the current bit width, and then perform direct writing or extended bit writing respectively. This part, on the one hand, inherits the current segment and current bit width formed by the first vibration sample, and on the other hand, provides the basis for intra-segment writing for the formation of subsequent high-bit states. During processing, the next vibration sample is still split into sign bit and value bit according to the determination method of the first vibration sample, and the first non-signed valid bit sequence is determined by the determination bit string. Then, the first non-signed valid bit sequence is compared with the current bit width item by item. When the first non-signed valid bit sequence does not exceed the current bit width, the next vibration sample is directly written to the end of the current segment according to the current bit width. When the first non-signed valid bit sequence exceeds the current bit width, the extended bit is first added to the vibration samples already written in the current segment according to the bit difference, and then the current bit width is rewritten. Subsequently, the next vibration sample is written according to the rewritten current bit width, so that all vibration samples in the current segment correspond to the same writing bit width. The implementation process includes the following steps: In S2-1, the original binary data of the next vibration sample is first read from the equipment vibration sequence according to the sampling order, and the original binary data is split into one sign bit and multiple value bits. When the sign bit is zero, the value and arrangement order of each value bit remain unchanged, and the value bit is directly used as the basis for judgment. When the sign bit is one, each bit is inverted bit by bit according to the original arrangement order of the value bits, and the inverted bits are rearranged according to their original positions to form a judgment bit string. After the judgment bit string is formed, starting from the bit closest to the sign bit, the value is checked bit by bit from the bit furthest away from the sign bit. When the first bit is found to be one, the bit sequence corresponding to that bit is determined as the first non-sign valid bit sequence. If all the values ​​in the judgment bit string are zero, the bit sequence corresponding to the bit furthest from the sign bit in the value bits is determined as the first non-sign valid bit sequence. After this process, the next vibration sample can be converted into the first non-sign valid bit sequence regardless of the sign change, and is used for comparison with the current bit width. In S2-2, when the first non-signed valid bit sequence is not higher than the current bit width, the current bit width of the current segment remains unchanged, and the position after the end of the current segment is used as the write position. The numerical bits of the original numerical bits of the next vibration sample are truncated according to the current bit width. The truncation range is from the least significant bit of the numerical bits to the corresponding bit of the current bit width. Bits higher than the corresponding bit of the current bit width are not written. Then, the original sign bit of the next vibration sample is retained, and the original sign bit and the truncated numerical bits are combined in the original order to form the write value. The write value is then written to the position after the end of the current segment. After the writing is completed, the write position is changed to the new end position of the current segment, and the sample count value is incremented by one. This results in the current segment after the writing. At this time, the starting position of the current segment remains unchanged, the current bit width remains unchanged, and the bit layout of the vibration samples written earlier in the segment does not change. In S2-3, when the first non-signed valid bit sequence is higher than the current bit width, the bit difference between the first non-signed valid bit sequence and the current bit width is first calculated. The bit difference is a positive integer obtained by subtracting the current bit width from the first non-signed valid bit sequence. After the bit difference is determined, the vibration samples written in the current segment are read one by one in the order from the start position to the end position of the current segment. For each written vibration sample, the same number of extension bits as the sign bit of the vibration sample are added to the high bit of the original written value. The number of extension bits is equal to the bit difference. Among them, vibration samples with a sign bit of zero are extended by... For vibration samples with a zero-value extension bit, a one-value extension bit is added. After the extension, the original sign bit of each written vibration sample remains in its original position, and the overall position of the original value bits is shifted to the lower bit direction. The newly added extension bit is located after the sign bit and before the original value bits. After the extension is completed for all written vibration samples in the current segment, the write bit length of each sample in the current segment increases synchronously with the bit difference. Through this process, the previously written vibration samples in the current segment are rewritten on the basis of the new bit width, providing the segment conditions for subsequent rewriting of the current bit width and writing the next vibration sample. In S2-4, after all the vibration samples written in the current segment have completed the extension bit filling, the current bit width is rewritten to the first non-signed valid bit sequence. Then, the position after the end of the current segment is used as the writing position. The numerical bits of the original numerical bits of the next vibration sample are truncated according to the rewritten current bit width, and the original sign bit of the next vibration sample is retained. The original sign bit and the truncated numerical bits are combined in the original order to form the writing value, and then the writing value is written to the position after the end of the current segment. After the writing is completed, the writing position is rewritten to the new end position of the current segment, and the sample count value is incremented by one based on the original bit width before the extension, thus obtaining the extended current segment. After the current bit width is rewritten, when the high-bit state is formed in the current segment, the corresponding bit of the rewritten current bit width is used as the inspection object. In this way, the extended current segment contains both the original vibration samples after the bit filling and the next vibration sample written according to the rewritten current bit width. Through the above processing, the next vibration sample is first converted into the first non-signed valid bit sequence, and then compared with the current bit width, thus splitting into two paths: direct writing or extended bit writing. In the direct writing path, the current segment bit width remains unchanged, and only one end sample is added to the segment structure. In the extended bit writing path, the vibration samples already written in the current segment are first rewritten according to the bit difference, and then the current bit width is rewritten and the next vibration sample is written, so that all vibration samples in the current segment are always based on the same bit width. The subsequent high bit state formation and the determination of the width reduction position can continue to be executed on this basis. In practical applications: After the edge node has established the current segment with a current bit width of three bits, if the first non-signed valid bit sequence obtained by the judgment bit string check of the next vibration sample is still three bits, then the original current bit width is directly retained, the next vibration sample is written to the end of the current segment with three numerical bits and a sign bit, and the sample count value is incremented by one; if the first non-signed valid bit sequence obtained by the judgment bit string check of the next vibration sample is five bits, then the bit difference is first calculated to be two, and then two extension bits are added to the high bit end of the numerical bits of each vibration sample already written in the current segment. Positive samples are added with two zero-value extension bits, and negative samples are added with two one-value extension bits. Then the current bit width is rewritten to five bits, and the next vibration sample is written to the end of the current segment with five numerical bits and a sign bit. This ensures that each vibration sample before and after the bit expansion is based on five-bit writing.

[0020] S3. During the current segment writing process, check the value of each vibration sample in the corresponding bit of the current bit width according to the sampling order. When the value is zero, it is recorded as empty; when the value is one, it is recorded as occupied, and a high-bit state sequence is formed. In this implementation, after each vibration sample in the current segment is written, a segment state record is formed around the corresponding bit of the current bit width for subsequent width reduction processing. Here, instead of processing each bit of the entire written value, only the corresponding bit of the current bit width is checked: if the bit value is zero, it means that the corresponding bit of the current bit width is not actually occupied by the current vibration sample; if the bit value is one, it means that the corresponding bit of the current bit width has been actually occupied by the current vibration sample. Based on this, adjacent positions with the same value are merged into the same state interval according to the sampling order, and each state interval is written in positional order to form a high-bit state sequence corresponding to the current segment, which can be directly called when identifying empty intervals from the end of the current segment backward. The implementation process includes the following steps: In S3-1, when writing each vibration sample in the current segment, the sign bit and value bit in the written value of the vibration sample are read synchronously, and the corresponding bit of the current bit width is located in the value bit. The corresponding bit of the current bit width is the highest value bit within the coverage of the current bit width, excluding the sign bit and bits below the current bit width. After reading the corresponding bit of the current bit width, an empty state value is generated when the value of the bit is zero, and an occupied state value is generated when the value of the bit is one. The state values ​​corresponding to each vibration sample are arranged in the order of writing the vibration samples to form a state value sequence. For example, when the current bit width is four bits, the value bits of each vibration sample are arranged from the lowest bit to the corresponding bit of the current bit width to form four value bits. The object to be checked is the highest bit among these four value bits. If the bit is zero, the vibration sample corresponds to an empty state value. If the bit is one, the vibration sample corresponds to an occupied state value. After this processing, each vibration sample in the current segment obtains a state value that corresponds one-to-one with its position. In S3-2, adjacent state values ​​in the state value sequence are compared according to the sampling order. The state value corresponding to the starting position of the current segment is taken as the state value of the first state interval, and the starting position of the current segment is recorded as the starting position of the first state interval. When the state value corresponding to the next position is the same as the state value corresponding to the previous position, the next position is merged into the state interval to which the previous position belongs, the starting position of the state interval remains unchanged, and the ending position is advanced to the next position. When the state value corresponding to the next position is different from the state value corresponding to the previous position, the previous position is recorded as the ending position of the state interval to which the current position belongs, and the next position is recorded as the starting position of the next state interval. At the same time, the state value corresponding to the next position is recorded as the state value of the next state interval. After comparing to the end of the current segment, the ending position is recorded as the ending position of the last state interval. Thus, the state value, starting position, and ending position corresponding to each state interval in the current segment are obtained. Through this process, the same state values ​​that appear consecutively in the state value sequence are compressed into a state interval, and the position of state change within the segment is also located. In S3-3, the state intervals are arranged from front to back according to their starting positions. The state value, starting position, and ending position corresponding to each state interval are sequentially written into the state record area to form a high-order state sequence corresponding to the current segment. During writing, each state interval is treated as a state record. The state record first writes the state value, then writes the starting and ending positions. This allows subsequent readings of the high-order state sequence to directly determine whether the current position is empty or occupied based on the state value, and to determine the state coverage range based on the starting and ending positions. For example, if the first two vibration samples correspond to empty state values, the third to fifth vibration samples correspond to occupied state values, and the sixth to eighth vibration samples correspond to empty state values, then the high-order state sequence after writing contains three state records, corresponding to the empty state interval, the occupied state interval, and the empty state interval, respectively. Through this process, the sample-by-sample state values ​​within the current segment are converted into a high-order state sequence organized by state intervals. Through the above processing, each vibration sample in the current segment first generates a state value around the corresponding position of the current bit width, then forms a state interval by comparing adjacent state values, and finally writes it into the high-bit state sequence according to the position order of the state interval. When performing empty interval identification in the future, it is not necessary to check the corresponding position of the current bit width of all vibration samples in the current segment again. Instead, the end state interval and its position range can be read directly from the high-bit state sequence to determine whether there is an empty interval at the end and which vibration sample positions are covered by the empty interval. In practical applications: After the edge node completes writing to the eight vibration samples in the current segment, if the state value sequence obtained by checking the corresponding bit width is empty, empty, occupied, occupied, occupied, empty, empty, empty, empty, the first empty state interval is formed by the first two positions, the occupied state interval is formed by the third to fifth positions, and the final empty state interval is formed by the sixth to eighth positions. Then, the three state records are written to the high-bit state sequence in the order of the starting position. When performing width reduction processing later, it is only necessary to read the empty state interval corresponding to the sixth to eighth positions from the end of the high-bit state sequence to directly obtain the start and end positions of the final empty interval, without having to return to the sample-by-sample bit value checking process.

[0021] S4. Starting from the end of the current segment, check the high-level state sequence backwards. Locate the empty sample intervals at the end. Delete the corresponding position of the current position width for each vibration sample in the sample interval and recalculate the value. When the recalculation result is consistent with the value before deletion, record the starting position of the sample interval as the width reduction position and decrement the current position width by one as the subsequent segment width. In this embodiment, this part is used to determine whether the end of the current segment has the conditions for width reduction after a high-level state sequence has been formed in the current segment, and to give the width reduction position and the width of the subsequent segment if the conditions for width reduction are met. During processing, the empty interval at the end is located in reverse from the state record corresponding to the end of the current segment. Then, each vibration sample in the empty interval is recalculated by deleting positions to determine whether the value of a single vibration sample changes after deleting the position corresponding to the current width. On this basis, the difference and magnitude relationship between two adjacent vibration samples in the empty interval are recalculated after deleting positions to determine whether the deletion action destroys the sequential relationship between adjacent samples. Finally, the sample-by-sample results and the results of adjacent samples are merged and checked in the order from the end to the front. When a change in value or adjacency relationship is found, the blocking position is determined. When no blocking position is found, the empty starting position is used as the width reduction position, and the current width is reduced by one and recorded as the width of the subsequent segment. The implementation process includes the following steps: In S4-1, the high-order state sequence corresponding to the current segment is read first, and the process starts from the state record with the largest starting position in the high-order state sequence and proceeds backward. If the state record corresponding to the end position of the current segment is occupied, it means that the vibration sample at the end of the current segment is still occupied at the position corresponding to the current bit width. In this case, no empty interval is formed at the end, and the position after the end of the current segment is recorded as the width-preserving position, while the current bit width remains unchanged. If the state record corresponding to the end position of the current segment is empty, the end position of the current segment is taken as the empty end position, and the process continues to proceed forward along the high-order state sequence. The system checks the previous status record until it finds the first status record that is occupied. The position after the end of that status record is recorded as the empty start position. If no occupied status record appears before the start of the current segment, the start of the current segment is recorded as the empty start position. After the empty start position and empty end position are determined, the corresponding vibration sample position range between them is determined as the empty interval. After this processing, whether the end position is in an empty state, the start position of the empty interval, and the end position of the empty interval all have direct intra-segment position basis. In S4-2, the written values ​​of each vibration sample are read sequentially from the start position to the end position of the empty space. For each vibration sample, the corresponding bit of the current bit width is deleted and the value is recalculated. During deletion, the sign bit is kept unchanged, the corresponding bit of the current bit width is deleted from the value bits, and the remaining value bits are reassembled in the order after deletion to form the value bits after deletion. Then, the value after deletion is restored by the sign bit and the value bits after deletion. Among them, the value before deletion is the original value corresponding to the vibration sample before the current segment is written, and the recalculated value is the value restored after deletion. When the recalculated value is equal to the value before deletion, the vibration sample is recorded as a preserved value. When the values ​​are not equal, the vibration sample is recorded as a variable value and written into the sample result according to the sample position order. Taking one sign bit and four value bits as an example, if the current bit width corresponds to the highest bit among the four value bits, and a vibration sample is written with a sign bit of zero and a value bit of 0101, then the corresponding bit of the current bit width is deleted to obtain a value bit of 101. The value after deletion is then restored from the sign bit of zero and the value bit of 101 and compared with the value of the vibration sample before deletion. If the values ​​before and after deletion are equal, the vibration sample is recorded as a preserved value. If the values ​​before and after deletion are not equal, the vibration sample is recorded as a variable value. Through this process, each vibration sample in the empty interval obtains a deletion result at the single sample level. In S4-3, after the sample results are formed, two adjacent vibration samples are read sequentially from the empty start position to the empty end position, and the same deletion and recalculation process as in S4-2 is performed on each vibration sample. For each pair of adjacent vibration samples, the difference before deletion is obtained by subtracting the value of the vibration sample corresponding to the previous position from the value before deletion of the vibration sample corresponding to the next position. Then, the difference after deletion is obtained by subtracting the recalculated value of the vibration sample corresponding to the previous position from the recalculated value of the vibration sample corresponding to the next position. At the same time, the size relationship between the vibration samples corresponding to the previous and next positions before and after deletion is compared. When the difference before deletion is equal to the difference after deletion, and the size relationship before and after deletion is the same, this pair of adjacent vibration samples is... The results between adjacent samples are recorded as adjacency preserved. When the difference before deletion is not equal to the difference after deletion, or the size relationship before deletion is different from the size relationship after deletion, the results between this pair of adjacent vibration samples are recorded as adjacency mismatch, and the adjacency results are written in the order from the previous position to the next position. Taking the seventh and eighth vibration samples in the empty interval as an example, if the difference between the two before deletion is two, and the difference between the two after deletion is still two, and the value of the eighth vibration sample is greater than the value of the seventh vibration sample before and after deletion, then the two are recorded as adjacency preserved. If the difference after deletion is changed to one, or the size relationship after deletion is changed to equal or reversed, then the two are recorded as adjacency mismatch. Through this process, the relationship between adjacent samples in the empty interval before and after deletion is also recorded. In S4-4, the sample results and adjacency results are checked in reverse from the empty end position to the empty start position. During the check, the sample result corresponding to the current position is read first, and then the adjacency result corresponding to the current position and the next position is read. If the sample result corresponding to the current position is preserved and the adjacency result corresponding to the current position and the next position is adjacency maintained, the current position is moved forward one position and the check continues. If the sample result corresponding to the current position is a variable value, the current position is recorded as the blocking position. If the adjacency result corresponding to the current position and the next position is an adjacency mismatch, the next position is moved forward one position and the check continues. The position is recorded as the blocking position; if multiple blocking positions appear simultaneously during the reverse check, the blocking position closest to the end of the empty position is taken as the width reduction position; if no blocking position is formed after the reverse check is completed, the empty start position is recorded as the width reduction position; after the width reduction position is determined, the current position width is reduced by one and recorded as the subsequent segment position width; through this process, the width reduction position is not directly given by the end empty state, but is constrained by the single sample deletion result and the adjacent sample deletion result, so that the generation of the subsequent segment position width corresponds to the actual deletion result of the vibration sample at the end of the current segment; Through the above processing, the empty interval at the end is first located by the high-level state sequence. Then, the vibration samples in the empty interval are recalculated by deleting positions and recalculating the relationship between adjacent samples. Then, starting from the end position of the empty interval, the results of single samples and adjacent samples are merged and checked. Finally, the width reduction position and the width of the subsequent segment are given. When the end of the current segment is still occupied, the current width is maintained and the width preservation position is given. When the end of the current segment is empty but there is a change in value or adjacency relationship inside the interval after deletion, the width reduction boundary is constrained by the blocking position. When no blocking position appears in the empty interval after deletion, the starting position of the empty interval is used as the width reduction position. In practical applications: When the high-level state record corresponding to the end of the current segment is empty, and this empty state continues from the fifteenth vibration sample to the twentieth vibration sample, the fifteenth position can be recorded as the empty start position and the twentieth position as the empty end position. Then, the corresponding bits of the current bit width are deleted sequentially for the fifteenth to the twentieth vibration samples, and the values ​​after deletion are restored. If the values ​​before and after deletion are equal for the fifteenth to the seventeenth vibration samples, and the values ​​before and after deletion change for the eighteenth vibration sample, the eighteenth position forms a blocking position during reverse inspection, and the eighteenth position is used as the width reduction position. If the values ​​before and after deletion do not change for the fifteenth to the twentieth vibration samples, but the difference between the sixteenth and seventeenth vibration samples changes, the seventeenth position is used as the blocking position. If the values ​​before and after deletion do not change for the fifteenth to the twentieth vibration samples, and the difference and magnitude relationship between all adjacent samples before and after deletion do not change, the fifteenth position is used as the width reduction position, and the current bit width is reduced by one and recorded as the bit width of the subsequent segment.

[0022] S5. End the current segment at the width reduction position, write the vibration samples before the width reduction position into the current compression segment according to the current bit width, create the subsequent segment according to the bit width of the subsequent segment, and repeat the vibration sample writing, high bit state formation and width reduction position determination to output the lossless compression result. In this embodiment, this part is used to split the current segment into a current compression segment and a subsequent segment based on the determined width reduction position and the given subsequent segment width, and write the split result into the lossless compression result. During processing, the vibration samples in the current segment are first divided into two parts according to the width reduction position. The vibration samples before the width reduction position are retained in the current compression segment, and the vibration samples after the width reduction position are transferred to the subsequent segment. Then, the current compression segment is written according to the current width, and the first sample of the subsequent segment is written according to the subsequent segment width. Starting from this first sample, the subsequent vibration samples are written, high-level state is formed, and width reduction position is determined. When the end of the equipment vibration sequence is read, the end segment information is written into the lossless compression result, and each compression segment is arranged in order of the starting position of the compression segment. The implementation process includes the following steps: In S5-1, the current segment, the width reduction position, the current bit width, and the bit width of the subsequent segment are read first, and the position before the width reduction position is taken as the end position of the current segment. When the width reduction position does not coincide with the start position of the current segment, the vibration samples corresponding to the start position to the end position of the current segment are determined as the current compressed segment samples, and the vibration samples corresponding to the width reduction position to the end of the current segment are determined as the subsequent segment samples. When the width reduction position coincides with the start position of the current segment, the current compressed segment samples are empty, and all vibration samples corresponding to the width reduction position to the end of the current segment are determined as the subsequent segment samples. After this division, the vibration samples inside the current segment are divided into two parts, the first part corresponds to the current compressed segment, and the second part corresponds to the subsequent segment. The two parts of samples are no longer cross-called during subsequent writing. In S5-2, the samples of the current compressed segment are read sequentially from the start position to the end position of the current segment. For each vibration sample, the original sign bit is retained, and the numerical bits are truncated from the original numerical bits according to the current bit width. The truncation range is from the least significant bit to the corresponding bit of the current bit width, and higher bits beyond the corresponding bit width are not written. Then, the original sign bit and the truncated numerical bits are combined in the original order to form the sample write value, and written continuously in the current compressed segment according to the sample position order. After all the samples of the current compressed segment have been written, the start position, end position, current bit width, and number of samples of the current compressed segment are written together into the current compressed segment record. The number of samples is counted according to the number of vibration samples actually written in the current compressed segment. Through this process, the current compressed segment contains both the sample encoding value within the segment and the segment record information used to locate the segment boundary and restore the bit width during subsequent decompression. In S5-3, the width reduction position is used as the starting position of the subsequent segment, and the vibration sample corresponding to the width reduction position is used as the first sample of the subsequent segment. During writing, the original sign bit of the vibration sample corresponding to the width reduction position is retained, and the value bits are truncated from the original value bits according to the width of the subsequent segment. The truncation range is from the least significant bit of the value bits to the corresponding bit of the width of the subsequent segment. Bits higher than the corresponding bit of the width of the subsequent segment are not written. Then, the original sign bit and the truncated value bits are combined in the original order to form the first sample write value, and the first sample write value is written to the first sample position of the subsequent segment. After the first sample is written, the starting position of the subsequent segment, the width of the subsequent segment, and the position of the first sample are written to the subsequent segment record, and the sample count value of the subsequent segment is recorded as one. After this processing, the subsequent segment obtains a new segment starting point and a new bit width basis. When the subsequent vibration sample enters the subsequent segment, it uses the subsequent segment record as the reference basis. In S5-4, vibration samples after the width reduction position are read again according to the sampling order. The subsequent segment established in S5-3 is used as the current processing segment. Vibration sample writing, high-level state formation, and width reduction position determination are repeated for the subsequent vibration samples. When a new width reduction position is formed again within the subsequent segment, the subsequent segment is further divided into a new current compression segment and a new subsequent segment in the same processing method as the current segment. The compression segment that has been written is written into the lossless compression result. When the end of the equipment vibration sequence is read, the start position of the current subsequent segment is recorded as the start position of the last segment, and the end position of the equipment vibration sequence is recorded as the end position of the last segment. The width and number of samples of the subsequent segment are written into the lossless compression result. Finally, the records of each compression segment and the corresponding sample code value in the lossless compression result are arranged from front to back according to the start position of each compression segment. Through this process, all compression results of the equipment vibration sequence from the first segment to the last segment are written segment by segment and organized according to the original sampling order. Through the above processing, the current segment is split into the current compression segment and the subsequent segment at the width reduction position. The first part is written according to the current bit width, and the second part is re-constructed according to the bit width of the subsequent segment and continues to be compressed until the end of the device vibration sequence is read. In this way, the writing results of vibration samples before the width reduction position under the current bit width are preserved, and the vibration samples at the width reduction position and the subsequent position are continued to be written under a lower bit width. Thus, the lossless compression result saves the sample encoding values ​​and segment record information under different bit widths in segments. In practical applications: When the starting position of a current segment is the 1st vibration sample, the ending position is the 20th vibration sample, and the width reduction position is the 15th vibration sample, the 1st to 14th vibration samples can be first determined as the current compressed segment samples and written according to the current bit width. Then, the 15th to 20th vibration samples are determined as the subsequent segment samples, and the 15th vibration sample is used as the first sample of the subsequent segment and written according to the subsequent segment bit width. If the 15th to 20th vibration samples do not form a new width reduction position after further processing, then during the read... After obtaining the 20th vibration sample, the 15th position is recorded as the start position of the last segment, and the 20th position is recorded as the end position of the last segment. The width of the subsequent segments and the number of samples are written into the lossless compression result. If the 15th to 20th vibration samples form a new width reduction position in the 18th vibration sample after further processing, the 15th to 17th vibration samples are first written into the lossless compression result as a segment, and the 18th to 20th vibration samples are then processed as a new subsequent segment. This completes the segmented compression writing of the entire equipment vibration sequence.

[0023] Furthermore, an adaptive lossless compression system for device vibration data based on dynamic bit width includes: The bit width determination module is used to acquire the device vibration sequence collected by the edge node, read the first vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the first vibration sample and record it as the current bit width, and establish the current segment with the first vibration sample. The bit width extension module reads the next vibration sample according to the sampling order, determines the first non-signed valid bit sequence of the next vibration sample, and writes it into the current segment according to the current bit width when the first non-signed valid bit sequence is not higher than the current bit width. When the first non-signed valid bit sequence is higher than the current bit width, the current bit width is rewritten to the first non-signed valid bit sequence and then written into the current segment. The status flag module is used to check the value of each vibration sample in the corresponding bit of the current bit width according to the sampling order during the current segment writing process. When the value is zero, it is marked as empty; when the value is one, it is marked as occupied, and a high-bit status sequence is formed. The bit width reduction module is used to check the high-level state sequence from the end of the current segment backwards, locate the empty sample intervals at the end, delete the corresponding bit of the current bit width for each vibration sample in the sample interval, and recalculate the value. When the recalculation result is consistent with the value before deletion, the starting position of the sample interval is recorded as the reduction position, and the current bit width is reduced by one and recorded as the bit width of the subsequent segment. The segmented compression module is used to end the current segment at the width reduction position, write the vibration samples before the width reduction position into the current compression segment according to the current bit width, create the subsequent segment according to the bit width of the subsequent segment for the vibration samples at the width reduction position and after, and repeat the vibration sample writing, high bit state formation and width reduction position determination to output lossless compression results.

[0024] Working Principle: This solution is designed for lossless compression of device vibration data in edge computing environments. During processing, the current bit width is first determined from the first vibration sample, and the current segment is established based on this first sample. Then, subsequent vibration samples are read sequentially according to the sampling order. It is determined whether each subsequent sample is still within the current bit width range. If it is within the range, it is directly written into the current segment according to the current bit width. If it is outside the range, the existing vibration samples in the current segment are first padded with extension bits, and then the current bit width is rewritten before writing subsequent vibration samples. During the continuous writing of the current segment, the values ​​of each vibration sample at the corresponding bits of the current bit width are checked simultaneously to form a high-bit state sequence. Then, starting from the end of the current segment, the empty interval at the end is located, and the vibration samples in the empty interval undergo bit deletion, value recalculation, and size relationship recalculation to determine the width reduction position. Finally, the current segment ends at the width reduction position, the previous part is written as the current compressed segment, and the next part is used to establish a subsequent segment with a lower bit width for further processing. This allows the bit width to be adjusted segment by segment as the vibration data changes, reducing storage overhead without altering the original data. For example, in an edge computing architecture, when performing online vibration monitoring on acquisition nodes installed on motors, bearing housings, or gearboxes, the vibration sample amplitudes generated during the stable operation phase of the equipment are concentrated, and the current segment is continuously written with a lower bit width. When the equipment experiences start-stop switching, load fluctuations, or local impacts, the effective bits of subsequent vibration samples increase, and the system expands the current segment and continues writing to prevent vibration data loss. After the impact ends, if the last segment of vibration samples remains empty at the highest bit of the current bit width, the system checks from the end backwards whether the values ​​and adjacency relationships of these vibration samples after deleting that bit are still the same. If they are, this segment is used as the starting point for width reduction, and subsequent segments are established with a lower bit width to continue compression. Through this process, the equipment vibration sequence is automatically split into multiple compressed segments with different bit widths at the edge, which can expand the bit width in time when the operating conditions change, and reduce the bit width in time after the fluctuations subside, thus completing the storage in a way that better matches the data changes.

[0025] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A lossless compression method for adaptive device vibration data based on dynamic bit width, characterized in that, include: S1. Obtain the device vibration sequence collected by the edge node, read the first vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the first vibration sample and record it as the current bit width, and establish the current segment with the first vibration sample; S2. Read the next vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the next vibration sample, and write the first non-signed valid bit sequence into the current segment according to the current bit width if the first non-signed valid bit sequence is not higher than the current bit width. If the first non-signed valid bit sequence is higher than the current bit width, rewrite the current bit width to the first non-signed valid bit sequence and then write it into the current segment. S3. During the current segment writing process, check the value of each vibration sample in the corresponding bit of the current bit width according to the sampling order. When the value is zero, it is recorded as empty; when the value is one, it is recorded as occupied, and a high-bit state sequence is formed. S4. Starting from the end of the current segment, check the high-level state sequence backwards. Locate the empty sample intervals at the end. Delete the corresponding position of the current position width for each vibration sample in the sample interval and recalculate the value. When the recalculation result is consistent with the value before deletion, record the starting position of the sample interval as the width reduction position and decrement the current position width by one as the subsequent segment width. S5. At the width reduction position, end the current segment, write the vibration samples before the width reduction position into the current compression segment according to the current bit width, create the subsequent segment according to the bit width of the subsequent segment, and repeat the vibration sample writing, high bit state formation and width reduction position determination to output the lossless compression result.

2. The method for lossless compression of adaptive device vibration data based on dynamic bit width according to claim 1, characterized in that: S1 includes: S1-1. Read the sign bit and value bit of the first vibration sample. Keep the value bit unchanged when the sign bit is zero. When the sign bit is one, invert the value bit bit by bit and output the judgment bit string corresponding to the first vibration sample. S1-2. Check the value of each bit from the high end to the low end of the judgment bit string, determine the bit sequence corresponding to the first bit with a value of 1 as the first non-signed valid bit sequence, and record the first non-signed valid bit sequence as the current bit width; S1-3. Extract the numerical bits of the first vibration sample according to the current bit width and write them together with the sign bit into the first sample position. Establish the current segment with the first sample position, the current bit width and the sample count value.

3. The method for lossless compression of adaptive device vibration data based on dynamic bit width according to claim 2, characterized in that: S2 includes: S2-1. Read the sign bit and value bit of the next vibration sample. When the sign bit is zero, keep the value bit unchanged. When the sign bit is one, invert the value bit bit bit by bit. Output the judgment bit string corresponding to the next vibration sample. Check the value of each bit bit by bit from the high bit end to the low bit end of the judgment bit string. Determine the bit sequence corresponding to the first bit with a value of one as the first non-sign valid bit sequence. S2-2. When the first non-signed valid bit sequence is not higher than the current bit width, extract the numerical bits of the next vibration sample according to the current bit width and write them together with the sign bit to the end of the current segment, and output the current segment after writing.

4. The method for lossless compression of adaptive device vibration data based on dynamic bit width according to claim 3, characterized in that: S2 further includes: S2-3. When the first non-signed valid bit sequence is higher than the current bit width, calculate the bit difference between the first non-signed valid bit sequence and the current bit width, and add the same extension bit as the sign bit of each vibration sample to the high bit end of each vibration sample written in the current segment according to the bit difference. S2-4. Rewrite the current bit width as the first non-signed valid bit sequence, then extract the numerical bits of the next vibration sample according to the rewritten current bit width and write them together with the sign bit to the end of the current segment, and output the expanded current segment.

5. The method for lossless compression of adaptive device vibration data based on dynamic bit width according to claim 4, characterized in that: S3 includes: S3-1. When writing each vibration sample in the current segment, synchronously read the value of each vibration sample in the corresponding bit of the current bit width. When the value is zero, generate an empty state value. When the value is one, generate an occupied state value. Output the state value sequence arranged in the sampling order. S3-2. Compare adjacent state values ​​in the state value sequence according to the sampling order. When adjacent state values ​​are the same, keep the state interval to which the current position belongs unchanged. When adjacent state values ​​are different, take the previous position as the end position of the state interval to which the current position belongs, and take the next position as the start position of the next state interval. Output the state value, start position and end position of each state interval. S3-3. Write the state value, start position and end position in the order of the start position of each state interval to form the high-order state sequence corresponding to the current segment.

6. The method for lossless compression of adaptive device vibration data based on dynamic bit width according to claim 5, characterized in that: S4 includes: S4-1. Read the high-order state sequence from the end of the current segment backward. When the state corresponding to the end position is occupied, record the position after the end of the current segment as the width-preserving position and maintain the current bit width. When the state corresponding to the end position is empty, continue reading forward until the first occupied position is read. Then, record the position after the occupied position as the empty start position and the end of the current segment as the empty end position. Use the empty start position and the empty end position to determine the empty interval. S4-2. Read each vibration sample in the order from the start position to the end position of the empty space. After deleting the corresponding bit of the current bit width for each vibration sample, recalculate the value. If the recalculated value is consistent with the value before deletion, it is recorded as a preserved value. If the recalculated value is inconsistent with the value before deletion, it is recorded as a changed value, and the sample result is formed.

7. The method for lossless compression of adaptive device vibration data based on dynamic bit width according to claim 6, characterized in that: S4 further includes: S4-3. Read two adjacent vibration samples in the order from the start position to the end position of the empty space. After deleting the corresponding bit of the current bit width for each vibration sample, recalculate the difference and the size relationship. When the difference and the size relationship are the same before and after deletion, it is recorded as adjacency maintenance. When the difference or the size relationship is inconsistent before and after deletion, it is recorded as adjacency mismatch, and the adjacency result is formed. S4-4. Starting from the empty end position, check the sample results and adjacency results in reverse from the empty start position. If the sample result corresponding to the current position is preserved and the adjacency result between the current position and the next position is maintained, continue the reverse check. If the sample result corresponding to the current position is a variable value, record the current position as the blocking position. If the adjacency result between the current position and the next position is an adjacency mismatch, record the next position as the blocking position. If the blocking position exists, record the blocking position as the width reduction position. Otherwise, record the empty start position as the width reduction position and reduce the current bit width by one and record it as the subsequent segment bit width.

8. The method for lossless compression of adaptive device vibration data based on dynamic bit width according to claim 7, characterized in that: S5 includes: S5-1. Read the current segment, the width reduction position, the current bit width, and the bit width of the subsequent segment. Record the position before the width reduction position as the end position of the current segment. Determine the vibration sample corresponding to the current segment start position to the current segment end position as the current compression segment sample. Determine the vibration sample corresponding to the current segment end position from the width reduction position to the current segment end as the subsequent segment sample. S5-2. Read the samples of the current compressed segment in the order from the start position to the end position of the current segment. Extract the numerical bits of each vibration sample according to the current bit width and write them together with the sign bits into the current compressed segment. At the same time, write the start position of the current segment, the end position of the current segment, the current bit width and the number of samples, and output the current compressed segment.

9. The method for lossless compression of adaptive device vibration data based on dynamic bit width according to claim 8, characterized in that: The S5 also includes: S5-3. Using the width reduction position as the starting position of the subsequent segment, extract the numerical bits of the vibration sample corresponding to the width reduction position according to the width of the subsequent segment and write them together with the sign bits into the first sample position of the subsequent segment. At the same time, write the starting position of the subsequent segment, the width of the subsequent segment, and the first sample position, and output the subsequent segment. S5-4. Read the vibration samples after the width reduction position according to the sampling order. Repeat the vibration sample writing, high-level state formation and width reduction position determination for subsequent segments. When the end of the equipment vibration sequence is read, write the start position, end position, width of subsequent segments and number of samples into the lossless compression result. Combine the current compression segment and subsequent segments according to the start position of each compression segment.

10. A lossless compression system for adaptive device vibration data based on dynamic bit width, characterized in that, include: The bit width determination module is used to acquire the device vibration sequence collected by the edge node, read the first vibration sample according to the sampling order, determine the first non-signed valid bit sequence of the first vibration sample and record it as the current bit width, and establish the current segment with the first vibration sample. The bit width extension module reads the next vibration sample according to the sampling order, determines the first non-signed valid bit sequence of the next vibration sample, and writes it into the current segment according to the current bit width when the first non-signed valid bit sequence is not higher than the current bit width. When the first non-signed valid bit sequence is higher than the current bit width, the current bit width is rewritten to the first non-signed valid bit sequence and then written into the current segment. The status flag module is used to check the value of each vibration sample in the corresponding bit of the current bit width according to the sampling order during the current segment writing process. When the value is zero, it is marked as empty; when the value is one, it is marked as occupied, and a high-bit status sequence is formed. The bit width reduction module is used to check the high-level state sequence from the end of the current segment backwards, locate the empty sample intervals at the end, delete the corresponding bit of the current bit width for each vibration sample in the sample interval, and recalculate the value. When the recalculation result is consistent with the value before deletion, the starting position of the sample interval is recorded as the reduction position, and the current bit width is reduced by one and recorded as the bit width of the subsequent segment. The segmented compression module is used to end the current segment at the width reduction position, write the vibration samples before the width reduction position into the current compression segment according to the current bit width, create the subsequent segment according to the bit width of the subsequent segment for the vibration samples at the width reduction position and after, and repeat the vibration sample writing, high bit state formation and width reduction position determination to output lossless compression results.

Images