Multi-brand air conditioner compatible intelligent control panel and protocol conversion method and system

Abstract

The application relates to the technical field of communication protocol conversion, and discloses a multi-brand air conditioner compatible intelligent control panel, a protocol conversion method and a system, wherein the protocol conversion method comprises the following steps: extracting a frame fingerprint feature vector of a candidate data frame and performing a density clustering operation to separate a mixed frame stream into a plurality of device frame groups; an address mapping between the device frame groups and a physical air conditioner host is established through polling; frame structure segmentation is independently performed on each device frame group to generate a protocol conversion rule; for a device frame group that fails to generate an effective protocol conversion rule, protocol field semantics are inferred through control terminal level change time sequence correlation analysis; protocol drift is jointly determined according to a frame stream dispersion rate and an abnormal frame proportion, and variant frame recovery and protocol conversion rule incremental updating are performed.

Description

Technical Field

[0001] This invention relates to the field of communication protocol conversion technology, specifically to a smart control panel compatible with multiple air conditioners and a protocol conversion method and system. Background Technology

[0002] Base stations and data center server rooms typically connect multiple air conditioning units of different brands via a shared RS-485 communication bus for centralized monitoring and control. Each brand of air conditioner uses its own proprietary communication protocol, with different frame lengths, frame header and trailer identifiers, and data encoding methods.

[0003] Existing protocol conversion methods perform protocol template matching on the entire mixed frame stream on the shared bus, relying on a pre-stored brand protocol template library to complete frame parsing and conversion rule generation.

[0004] The above method has the following problems: First, the communication frames of multiple brands are intertwined and mixed on the bus. During the overall matching, the frame data of different brands interfere with each other, resulting in protocol identification failure and the inability to generate the correct protocol conversion rules for each air conditioner. Second, for air conditioners of niche brands not included in the template library or those replaced with non-original motherboards, the template matching process is completely ineffective. Third, when the protocol frame format changes due to firmware upgrades, the frame attribution determination and protocol conversion rules fail simultaneously. The failures at the two levels are intertwined, forming a detection blind spot, which causes the panel to interrupt the monitoring and control of the device. Summary of the Invention

[0005] This invention provides a smart control panel and protocol conversion method and system compatible with multiple brands of air conditioners, solving the technical problems in related technologies such as the inability to automatically identify heterogeneous protocol frames, the failure of conversion rules after protocol drift, and the inability to adaptively update in the scenario of multiple brands of air conditioners sharing a bus.

[0006] This invention discloses a smart control panel compatible with multiple brands of air conditioners and a protocol conversion method, including: monitoring the original data stream on the shared communication bus, dividing the data stream into candidate data frames at frame intervals, extracting the frame fingerprint feature vector of each candidate data frame, performing density clustering operation on all frame fingerprint feature vectors, grouping candidate data frames belonging to the same cluster into the same device frame group, and temporarily storing frames with a confidence level lower than the allocation threshold into an unassigned frame buffer pool. Frame structure segmentation is performed independently for each device frame group, generating frame structure templates and protocol conversion rules; The frame-level compliance score of the real-time frame stream is calculated based on the frame structure template, and the frame stream escape rate of the unassigned frame buffer pool is calculated. When the proportion of abnormal frames in a certain device frame group exceeds the first preset threshold and the frame stream escape rate simultaneously exceeds the second preset threshold, it is determined that the air conditioning host corresponding to the device frame group has experienced protocol drift. Unassigned frames after the drift alarm trigger time are extracted from the unassigned frame buffer pool. The byte-level structural similarity between the unassigned frames and the reference frame samples of the device frame group where the protocol drift occurred is calculated. Unassigned frames that exceed the structural association threshold are recycled as mutated frames. Based on the mutated frames, the protocol conversion rules are locally incrementally updated and the cluster centers are merged and updated.

[0007] Furthermore, the frame fingerprint feature vector includes frame length, first byte value, last byte value, and histogram of byte distribution within the frame. Figure 4 Each component has an intra-frame byte distribution histogram, which is a distribution feature obtained by performing frequency statistics on all byte values ​​within the frame according to a preset interval. Before combining them into a frame fingerprint feature vector, Z-score standardization is performed on each component.

[0008] Furthermore, after generating the device frame group, the method also includes an address mapping step between the device frame group and the physical host: For each device frame group, polling request frames carrying different address codes are sent sequentially to the bus via the panel; the response frames received within the preset response time window after each poll are recorded; the device frame group to which the response frame belongs is determined based on the distance between the frame fingerprint feature vector of the response frame and the existing cluster center; the device frame group to which the response frame belongs is bound to the bus address currently polled, and an address mapping relationship between the device frame group and the physical air conditioning host is generated.

[0009] Furthermore, the step of independently performing frame structure segmentation for each device frame group, generating frame structure templates and protocol conversion rules, includes: The frame boundary patterns are determined by frame header and tail feature detection and byte alignment analysis. The data within the frame is divided into a fixed header region, a variable data region, and a check tail region to generate a frame structure template. Statistical analysis of byte position value changes is performed on all frame samples of the device frame group for the variable data region. Byte positions with constant values ​​are classified as fixed constant fields, byte positions with values ​​that switch only between finite discrete values ​​are classified as status flag fields, and byte positions with values ​​that show a continuous gradual change trend are classified as numerical fields.

[0010] Furthermore, the generation protocol conversion rules also include: The Pearson correlation coefficient is calculated after performing Z-score standardization on the changing trend of the numerical fields and the measured ambient temperature values ​​collected by the panel sensor terminals. When the Pearson correlation coefficient between a certain numerical field and the measured ambient temperature value exceeds the correlation threshold, the field is marked as a temperature data field. The temperature encoding method and conversion formula are determined based on the correspondence between the temperature data field value and the measured ambient temperature value. Calculate the Pearson cross-correlation coefficients for similar numerical fields across different device frame groups. When the Pearson cross-correlation coefficients exceed the coordination threshold, cross-validation is performed using cross-device source fields to correct misjudgments of encoding methods caused by insufficient frame samples or data noise in a single device frame group. The verified temperature encoding method and conversion formula are then written into the protocol conversion rules of the corresponding device frame group.

[0011] Furthermore, for device frame groups that fail to generate valid protocol conversion rules through frame structure segmentation, the following also applies: Collect timing data of level changes of the control terminals corresponding to the frame group of the device during the autonomous operation of the air conditioner; The timing data of the level changes of the control terminal are correlated with the sensor temperature change trend and the command message on the communication terminal within the same time window. When the level flip event of a certain control terminal precedes the start time of a specific temperature change trend and the time interval is stably within a preset range in multiple observations, it is determined that the function of the control terminal corresponds to the temperature change behavior. When the level flip event and a command message of a certain fixed format on the communication terminal have a stable temporal relationship, it is determined that the level state of the control terminal corresponds to the running instruction represented by the command message. Based on causal timing relationships, the terminal function definitions and protocol field semantics are inferred to generate protocol conversion rules for the device frame group.

[0012] Furthermore, the frame-level compliance score is calculated as follows: the real-time received data frames are compared field by field with the frame structure template of the corresponding device frame group, and the frame length matching degree component, the frame header and tail conformance component, and the fixed constant field consistency component are weighted and summed with preset weights; wherein, the frame length matching degree component is full marks when the actual frame length matches the template frame length, and zero marks otherwise; the frame header and tail conformance component is the ratio of the actual number of consistent bytes to the total number of bytes at the corresponding position in the template; the fixed constant field consistency component is the ratio of the number of fixed constant fields in the actual frame that are completely consistent with the template to the total number of fixed constant fields; The frame stream spillover rate is the ratio of the number of frames added to the unowned frame buffer pool per unit time to the total number of frames received on the bus in the same unit time.

[0013] Furthermore, the calculation method for the byte-level structural similarity is as follows: taking the frame structure template of the benchmark frame sample as a reference, searching for the byte value sequence corresponding to the fixed constant field of the frame structure template in the unassigned frame, and calculating the ratio of the number of successfully matched fixed constant byte value sequences to the total number of fixed constant fields of the frame structure template. The local incremental update includes: performing a field-by-field difference comparison between the mutated frame and the reference frame sample to identify the mutation types such as frame length change, field width change, and value domain offset; for frame length change, updating the frame length parameter and field offset table in the frame structure template; for field width change, updating the byte extraction position and length parameter of the corresponding field in the protocol conversion rule; and for value domain offset, updating the offset constant in the conversion formula of the corresponding field in the protocol conversion rule.

[0014] Furthermore, the method for fusion and updating the cluster centers is as follows: extract the frame fingerprint feature vector again using the mutated frames, calculate the mean of the frame fingerprint feature vectors of all mutated frames, and use a preset attenuation factor to weight and fuse the mean of the frame fingerprint feature vectors with the original cluster centers to generate updated cluster centers; wherein the value range of the preset attenuation factor is zero to one. The updated clustering centers and protocol conversion rules are used to re-execute the attribution calculation and protocol parsing for subsequent bus frames. When the frame flow escape rate and frame-level compliance score are both restored to the normal range, the drift alarm status is cleared, and the updated protocol conversion rules and frame fingerprint features are written into the protocol template library of the panel. If the frame flow escape rate and frame-level compliance score are not restored to the normal range within the preset recovery time window, the complete protocol analysis process is re-executed for the device frame group where protocol drift occurred.

[0015] This invention provides a smart control panel and protocol conversion system compatible with multiple air conditioner brands, comprising: The frame fingerprint extraction and separation module is used to monitor the raw data stream on the shared communication bus, divide the data stream into candidate data frames according to the frame interval time, extract the frame fingerprint feature vector of each candidate data frame, perform density clustering operation on all frame fingerprint feature vectors to classify the candidate data frames into device frame groups, and temporarily store frames with a home confidence score lower than the allocation threshold into the unassigned frame buffer pool. The protocol conversion rule generation module is used to independently perform frame structure segmentation for each device frame group, and generate frame structure templates and protocol conversion rules. The protocol drift detection module is used to calculate the frame-level compliance score of the real-time frame stream based on the frame structure template, calculate the frame stream escape rate of the unassigned frame buffer pool, and determine that the air conditioning host corresponding to the device frame group has experienced protocol drift when the proportion of abnormal frames in a certain device frame group exceeds the first preset threshold and the frame stream escape rate simultaneously exceeds the second preset threshold. The mutated frame recycling and update module is used to extract unassigned frames after the drift alarm trigger time from the unassigned frame buffer pool, calculate the byte-level structural similarity between the unassigned frames and the reference frame samples of the device frame group where the protocol drift occurred, recycle unassigned frames that exceed the structural association threshold as mutated frames, perform local incremental updates on the protocol conversion rules based on the mutated frames, and merge and update the cluster centers.

[0016] This invention extracts frame fingerprint feature vectors and performs density clustering operations to automatically separate interleaved and mixed multi-brand air conditioner communication frames on a shared bus into independent device frame groups based on the statistical differences in their frame fingerprint feature vectors. The frame samples of each device frame group originate from protocol frames of the same brand, eliminating mutual interference between frame data from different brands during protocol matching. This allows each device frame group to independently execute frame structure segmentation and protocol conversion rule generation, thereby solving the problem of protocol identification failure caused by overall matching of mixed frame streams.

[0017] This invention performs passive listening and timing correlation analysis of control terminal level changes on frame groups of unknown brand devices. It infers terminal function definitions and protocol field semantics by utilizing the causal timing relationship between control terminal level flip events, temperature change events, and command messages. It generates protocol conversion rules without relying on pre-stored templates, thereby solving the problem of protocol conversion being impossible for brands not covered by the template library.

[0018] This invention uses a combined determination of frame stream escape rate and the proportion of abnormal frames in a device frame group to synchronously detect failure signals at the frame attribution level and protocol conversion rule level. A drift alarm is triggered only when both indicators are synchronously abnormal, eliminating false positives caused by fluctuations in a single indicator. Furthermore, mutated frames are reclaimed from the unattributed frame buffer pool based on byte-level structural similarity, and local incremental updates to the protocol conversion rules and preset attenuation factors of the cluster centers are fused and updated in conjunction. This ensures that frame attribution determination and protocol parsing synchronously adapt to the changed frame format, thereby solving the problem of detection blind spots caused by the intertwining of frame attribution failures and protocol conversion rule failures. Attached Figure Description

[0019] Figure 1 This is a flowchart of the intelligent control panel and protocol conversion method compatible with multi-brand air conditioners provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the distribution and clustering of candidate data frame fingerprint features provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the classification and distribution of the frame structure fields of Brand A provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of timing correlation analysis of the control terminals of Brand C provided in an embodiment of the present invention; Figure 5 This is a schematic diagram comparing the monitoring indicators of Brand A group before and after protocol drift provided in an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating an example of calculating the frame-level compliance score for Brand A provided in an embodiment of the present invention; Figure 7 This is a schematic diagram illustrating the changing trends of key indicators before and after drift recovery in Brand A group, provided in an embodiment of the present invention. Figure 8 This is a schematic diagram of the retrieval and structural similarity distribution of variant frames of Brand A provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the distribution of brand A frame fingerprint feature vectors before and after cluster center fusion update provided in an embodiment of the present invention. Detailed Implementation

[0020] In base station or data center environments, intelligent control panels connect multiple air conditioners of different brands via a shared RS-485 communication bus. Each brand of air conditioner uses its own proprietary communication protocol, with different frame lengths, frame header and trailer identifiers, and data encoding methods. When multiple heterogeneous brand air conditioners are connected to the shared bus simultaneously, the communication frames from each brand are intertwined and mixed on the bus. Existing protocol conversion methods perform protocol matching on the mixed frame stream as a whole, but the frame data from different brands interfere with each other, leading to protocol recognition failure and making it impossible to generate correct protocol conversion rules for each air conditioner. Furthermore, for air conditioners from niche brands not included in the template library or those with replaced motherboards from non-original manufacturers, the template matching process completely fails. In addition, when the air conditioner's protocol frame format changes due to firmware upgrades, frame attribution determination and protocol conversion rules simultaneously fail. The panel cannot correctly attribute the changed frames to the corresponding device, nor can it determine when the protocol conversion rules have failed. These two levels of failure intertwine, causing the panel to interrupt its monitoring and control of the device.

[0021] According to an embodiment of the present invention, a multi-brand air conditioner compatible smart control panel and protocol conversion method are provided, such as... Figure 1 As shown, it includes the following steps: Step 1: Frame fingerprint extraction and device frame group separation; The system continuously monitors the raw data stream on the shared communication bus and segments it into candidate data frames based on the frame interval. It extracts the frame fingerprint feature vector for each candidate data frame, performs density clustering on all frame fingerprint feature vectors, and groups candidate data frames belonging to the same cluster into the same device frame group, generating multiple device frame groups and corresponding cluster centers. Frames with a home assignment confidence score below the assigned threshold are temporarily stored in the unassigned frame buffer pool.

[0022] It should be noted that the frame fingerprint feature vector mentioned above includes frame length, first byte value, last byte value, and histogram of byte distribution within the frame. Figure 4The four components are: Intra-frame byte distribution histogram (NTD) and Intra-frame byte distribution histogram. The Frame length represents the inherent length differences between different brand protocol frames. The first and last byte values ​​correspond to the numerical characteristics of the frame header and trailer identifiers or check bytes, respectively. The NTD reflects the statistical differences in the data encoding methods of different brand protocol frames. Since the four components have different dimensions, Z-score normalization is performed on each component before combining them into a frame fingerprint feature vector. This eliminates the influence of dimensional differences on distance measurement, ensuring that each component has a comparable scale in the feature space. The combination of the normalized four components allows protocol frames from different brands to form distinguishable clustered distributions in the feature space, enabling automatic separation of mixed frame streams by brand through density clustering operations.

[0023] It should be noted that the density clustering operation described above takes the frame fingerprint feature vectors of all candidate data frames as input and outputs cluster labels and corresponding cluster centers. Density clustering groups density-connected frame fingerprint feature vectors in the feature space into the same cluster, while outliers that are density-inaccessible are marked as unassigned frames. The assignment confidence score reflects the distance between the candidate data frame's frame fingerprint feature vector and its nearest cluster center; the smaller the distance, the higher the assignment confidence score. When the assignment confidence score is lower than the allocation threshold, it indicates that the frame cannot be reliably assigned to any existing cluster and is temporarily stored in the unassigned frame buffer pool for further processing.

[0024] Step 2: Address mapping between device frame groups and physical hosts; For each device frame group, polling request frames carrying different address codes are sequentially sent to the bus via the panel. The response frames received within a preset response time window after each poll are recorded. The device frame group to which the response frame belongs is bound to the corresponding bus address, generating an address mapping relationship between the device frame group and the physical air conditioning unit.

[0025] It should be noted that the above polling process traverses the available address space of the RS-485 bus in an address-incrementing manner. The preset response time window is the maximum waiting time for a response after sending a polling request frame. The duration of the preset response time window covers the maximum bus transmission delay and the air conditioning host response delay. When a response frame is received within the preset response time window, the device frame group to which the response frame belongs is determined based on the distance between the frame fingerprint feature vector of the response frame and the existing cluster centers, thereby binding the device frame group to which the response frame belongs to the currently polled bus address.

[0026] Step 3: Frame structure segmentation and protocol conversion rule generation; Frame structure segmentation is performed independently for each device frame group. Frame boundary patterns are determined through frame header and trailer feature detection and byte alignment analysis. The data within the frame is segmented into a fixed header region, a variable data region, and a checksum trailer region, generating a frame structure template for each device frame group.

[0027] Perform statistical analysis on the byte position value changes of the variable data region on all frame samples of each device frame group, and classify the byte positions into status flag fields, numeric fields and fixed constant fields.

[0028] It should be noted that the above byte position classification is based on the following criteria: byte positions whose values ​​remain unchanged throughout all frame samples are classified as fixed constant fields; byte positions whose values ​​switch only between a finite number of discrete values ​​are classified as status flag fields; and byte positions whose values ​​show a continuous and gradual trend in the frame sample sequence are classified as numeric fields.

[0029] The Pearson correlation coefficient is calculated between the changing trends of numerical fields in different device frame groups and the measured ambient temperature values ​​collected by the panel sensor terminals. Before calculation, Z-score normalization is performed on the original byte values ​​of the numerical fields and the measured ambient temperature values ​​to eliminate the influence of the difference in their dimensions on the correlation coefficient calculation. When the Pearson correlation coefficient between the changing trend of a certain numerical field and the measured ambient temperature value exceeds the correlation threshold, the field is marked as a temperature data field. Based on the correspondence between the temperature data field values ​​and the measured ambient temperature values, the temperature encoding method and conversion formula are determined, and the protocol conversion rules corresponding to each device frame group are generated.

[0030] In this embodiment, to improve the accuracy of temperature coding method calibration, Pearson cross-correlation coefficients for similar numerical fields across different device frame groups are calculated. When the Pearson cross-correlation coefficient exceeds a coordination threshold, it indicates that the corresponding fields in multiple device frame groups exhibit a consistent trend of change within the same time period, which can be determined as independent coding of the same environmental physical quantity. Cross-validation is performed using cross-device source fields where the Pearson cross-correlation coefficient exceeds the coordination threshold. This involves cross-verifying the independent coding results of multiple air conditioners for the same temperature quantity, correcting misjudgments of coding methods in a single device frame group due to insufficient frame samples or data noise, and incorporating the verified temperature coding method and conversion formula into the protocol conversion rules of the corresponding device frame group.

[0031] Step 4, real-time protocol conversion; Based on the address mapping relationship and protocol conversion rules of each device frame group, for subsequent data frames arriving on the bus, the distance between the frame fingerprint feature vector of the data frame and each cluster center is calculated, and the data frame is distributed to the nearest device frame group. Using the protocol conversion rules of each device frame group, the proprietary protocol frames are parsed into monitoring data of a unified format, and the unified format control commands are encoded into proprietary protocol frames of each brand and sent to the bus.

[0032] Steps 1 to 4 above are used for protocol separation and conversion of known brand air conditioners on the shared bus. For unknown brand air conditioners not covered by the template library, step 3 may also include: In this embodiment of the application, in order to handle device frame groups that fail to generate valid protocol conversion rules through frame structure analysis, the following steps are also included in step 3: Step 5, inference of unknown brand agreements; For device frame groups that fail to generate valid protocol conversion rules in step 3, passive listening is performed on the communication terminals corresponding to the device frame groups that fail to generate valid protocol conversion rules in step 3. Communication physical layer parameters and message structure features are extracted from spontaneous data frames and classified into protocol families.

[0033] Step 5 may specifically include: Step 501: Collect the timing data of level changes of the control terminal corresponding to the device frame group that failed to generate a valid protocol conversion rule in step 3 during the autonomous operation of the air conditioner.

[0034] Step 502: Perform timing correlation analysis on the level change timing data of the control terminals with the sensor temperature change trend and command messages on the communication terminals within the same time window to identify the causal timing relationship between the level flip of each control terminal and the temperature change event or command message.

[0035] It should be noted that the specific process of the above-mentioned time-series correlation analysis is as follows: Under the premise of aligning the time axis, the occurrence time of the control terminal level reversal event is detected, and within a preset observation window after the occurrence time of the control terminal level reversal event, it is detected whether a sensor temperature change event or a command message on the communication terminal occurs. When the level reversal event of a certain control terminal precedes the start time of a specific temperature change trend in time, and the time interval between the two stably falls within a preset range in multiple observations, it is determined that the function of the control terminal corresponds to the temperature change behavior. When the level reversal event and a command message of a fixed format on the communication terminal have a stable temporal sequence, it is determined that the level state of the control terminal corresponds to the running instruction represented by the command message.

[0036] Step 503: Based on the above causal timing relationship, infer the terminal function definition and protocol field semantics, and generate the protocol conversion rule for the device frame group that failed to generate a valid protocol conversion rule in step 3.

[0037] In this embodiment of the application, in order to automatically detect protocol drift and restore protocol conversion capability when the protocol frame format changes due to an air conditioner firmware upgrade, the following steps are also included in step 4: Step 6: Frame stream compliance monitoring and protocol drift determination; For each device frame group's real-time frame stream, a frame-level compliance score is calculated for each frame based on the frame structure template of each device frame group. The frame-level compliance score sequence is then input into a sliding window for statistical analysis to calculate the proportion of abnormal frames within the window and the moving variance of the frame-level compliance score.

[0038] It should be noted that the aforementioned frame-level compliance score refers to a comprehensive score obtained by comparing the real-time received data frames with the frame structure template of the corresponding device frame group field by field, and then weighting and summing the scores based on frame length matching, frame header and trailer conformity, and consistency of fixed constant fields. Frame-level compliance score The calculation method is as follows:

[0039] in, Frame-level compliance scoring; This is the frame length matching component, which takes the value when the actual frame length matches the template frame length. Otherwise, the value is ; This is the frame header / tail alignment component. It takes the value [value] when the actual frame header / tail byte values ​​are exactly the same as the corresponding byte values ​​in the template. When there is partial consistency, the value is taken as the ratio of the actual number of consistent bytes to the total number of bytes at the corresponding position in the template; For the consistency component of fixed constant fields, the value is the ratio of the number of fixed constant fields in the actual frame that are completely consistent with the corresponding fields in the template to the total number of fixed constant fields; , , The preset weights for the corresponding components, and satisfying the following conditions: The above three components , , These are all dimensionless ratios and can be directly weighted and summed. A higher frame-level compliance score indicates a higher degree of consistency between the frame and the frame structure template. An abnormal frame refers to a data frame with a frame-level compliance score lower than the frame-level compliance threshold.

[0040] Simultaneously monitor the cumulative rate of frames in the unassigned frame buffer pool, calculate the ratio of the number of newly added unassigned frames per unit time to the total number of bus frames, and obtain the frame stream escape rate index.

[0041] It should be noted that the frame stream escape rate The calculation method is as follows:

[0042] in, Frame stream escape rate; The number of frames added to the unassigned frame buffer pool per unit of time; This represents the total number of frames received on the bus within the same unit of time. Frame stream escape rate reflects the overall failure level of density clustering attribution determination. When the protocol frame format changes, the changed frames deviate from the original cluster center and are rejected by density clustering attribution determination into the unassigned frame buffer pool, resulting in an increase in frame stream escape rate.

[0043] When the proportion of abnormal frames in a certain device frame group exceeds the first preset threshold and the frame stream leakage rate increases synchronously to exceed the second preset threshold, it is determined that the air conditioning host corresponding to the device frame group whose proportion of abnormal frames exceeds the first preset threshold has experienced protocol drift, a drift alarm is generated, and the trigger time of the drift alarm and the corresponding device frame group identifier are recorded.

[0044] It should be noted that the above joint judgment requires both the abnormal frame ratio and the frame stream escape rate to simultaneously meet their respective threshold conditions. An increase in the abnormal frame ratio indicates a deviation between the existing frame structure template and the actual frame format, which is a failure signal at the protocol conversion rule level. An increase in the frame stream escape rate indicates a synchronous failure in density clustering attribution judgment, meaning that the changed frames have not been correctly assigned to the corresponding device frame groups, which is a failure signal at the frame attribution level. A drift alarm is triggered only when the indicators at both levels are synchronously abnormal, thus eliminating fluctuations in a single indicator caused by bus noise or occasional transmission errors, and accurately pointing to a scenario where the protocol frame format has undergone a systemic change.

[0045] Step 7: Mutant frame recycling and incremental update of protocol conversion rules; In response to a drift alarm, all unassigned frames accumulated since the trigger time of the drift alarm are retrieved from the unassigned frame buffer. A byte-level structural similarity calculation is performed between all unassigned frames and the baseline frame samples retained by the device frame group experiencing protocol drift before the drift alarm was triggered. Unassigned frames with a byte-level structural similarity exceeding the structural association threshold are marked as mutated frames and recycled to the device frame group experiencing protocol drift, forming the drift device extended frame set.

[0046] It should be noted that the above-mentioned byte-level structural similarity is calculated as follows: Using the frame structure template of the baseline frame sample as a reference, the byte value sequences corresponding to the fixed constant fields of the baseline frame sample's frame structure template are searched in the unassigned frames. The ratio of the number of successfully matched fixed constant byte value sequences to the total number of fixed constant fields in the baseline frame sample's frame structure template is calculated; this ratio is the byte-level structural similarity. A higher byte-level structural similarity indicates a higher consistency between the unassigned frame and the baseline frame sample in the fixed structure portion, and is more likely to belong to a variant of the same brand's protocol frame format after a change.

[0047] Step 7 may specifically include: Step 701: Perform a field-by-field difference comparison between the mutated frames and the reference frame samples in the extended frame set of the drifting device. Through position search of fixed constant fields and byte alignment analysis, identify mutation types such as frame length changes, field width changes and value range offsets.

[0048] It should be noted that the above field-by-field difference comparison uses the frame structure template of the baseline frame sample as a reference. The mutated frame searches for the byte value sequence corresponding to the fixed constant fields of the baseline frame sample's frame structure template. The offset of the matching position determines whether the frame length has changed and the specific location of the change. For variable regions between successfully matched fixed constant fields, the byte width difference between the mutated frame and the baseline frame sample at corresponding positions is compared to identify field width changes. For numeric fields whose width has not changed, the value range of numeric fields in the mutated frame is compared with the historical value range of numeric fields in the baseline frame sample. When the overall value range offset exceeds a preset deviation, it is identified as a value range offset.

[0049] Step 702: Perform a local incremental update on the protocol conversion rules of the device frame group where protocol drift has occurred based on the mutation type.

[0050] It should be noted that the above-mentioned local incremental update strategy corresponds to the mutation type: for frame length changes, the frame length parameter and field offset table in the frame structure template are updated; for field width changes, the byte extraction position and length parameter of the corresponding field in the protocol conversion rule are updated; for value range offsets, the offset constant in the conversion formula of the corresponding field in the protocol conversion rule is updated. The above-mentioned local incremental update is only performed on the mutated local fields; fields that have not mutated retain their original protocol conversion rules.

[0051] Step 703: Extract the frame fingerprint feature vector again using the mutated frame, and fuse the original cluster centers with a preset attenuation factor to generate updated cluster centers.

[0052] It should be noted that the updated cluster centers The calculation method is as follows:

[0053] in, The updated cluster centers; The original cluster centers; The mean of the frame fingerprint feature vectors of all mutated frames; The preset attenuation factor has a value range of 1. to . , and All are Z-score-normalized feature vectors of the same dimension, which can be directly weighted and fused. The larger the value of , the more the updated cluster centers are biased towards the feature distribution represented by the mean of the frame fingerprint feature vector of the mutated frame. By adjusting The value of controls the rate at which the cluster center transitions from the original frame features to the changed frame features, so that subsequent frame attribution determinations gradually adapt to the frame fingerprint features after the protocol change.

[0054] Step 8: Drift recovery verification and template library update; The updated cluster centers and protocol conversion rules are used to re-perform the attribution calculation and protocol parsing for subsequent bus frames. The updated frame stream escape rate and frame-level compliance score are calculated.

[0055] Once the frame stream escape rate and frame-level compliance score have both returned to normal ranges, clear the drift alarm status. Write the updated protocol conversion rules and frame fingerprint features into the panel's protocol template library as new template entries, for direct use when encountering the same model of air conditioner in the future.

[0056] In one embodiment, when the frame stream escape rate or frame-level compliance score fails to recover to the normal range within a preset recovery time window, the complete protocol analysis process of step 3 or step 5 is re-executed for the device frame group where protocol drift has occurred, in order to deal with the scenario where the protocol frame format has undergone fundamental reconstruction rather than partial change.

[0057] This implementation extracts frame fingerprint feature vectors and performs density clustering operations to automatically separate multi-brand air conditioner communication frames interleaved and mixed on the shared bus into independent device frame groups based on the statistical differences in frame fingerprint feature vectors. The frame samples of each device frame group come from the protocol frames of the same brand, eliminating mutual interference between frame data from different brands during the protocol matching process. Therefore, each device frame group can independently execute frame structure segmentation and protocol conversion rule generation, solving the problem of protocol identification failure caused by overall matching of mixed frame streams.

[0058] This implementation method performs passive listening and timing correlation analysis of control terminal level changes on frame groups of unknown brand devices. It infers the terminal function definition and protocol field semantics by using the causal timing relationship between control terminal level flip events, temperature change events, and command messages. It generates protocol conversion rules without relying on pre-stored templates, thus solving the problem that protocol conversion cannot be performed for brands not covered by the template library.

[0059] This implementation method uses a combined determination of frame stream escape rate and the proportion of abnormal frames in the device frame group to synchronously detect failure signals at the frame attribution level and protocol conversion rule level. An increased frame stream escape rate indicates that the modified frame has deviated from the cluster center and flowed into the unassigned frame buffer pool; an increased proportion of abnormal frames indicates that the frame structure template does not match the actual frame format. When both indicators are synchronously abnormal, the protocol drift event is accurately located. Furthermore, mutated frames are recovered from the unassigned frame buffer pool based on byte-level structural similarity, and a local incremental update of the protocol conversion rule and a fusion update of the preset attenuation factor of the cluster center are executed in conjunction, ensuring that frame attribution determination and protocol parsing synchronously adapt to the modified frame format. Therefore, after the air conditioner firmware upgrade, the panel can automatically restore its protocol conversion capability for the device without manual intervention, solving the problem of detection blind spots caused by the intertwining of frame attribution failure and protocol conversion rule failure.

[0060] A data center's intelligent control panel connects to three precision air conditioners of different brands simultaneously via a shared RS-485 bus: Brand A (address 01H), Brand B (address 02H), and Brand C (address 03H). The Brand C unit has a replaced motherboard (not the original manufacturer), and no corresponding record exists in the protocol template library. The data center completed a firmware upgrade for the Brand A unit in Q1 of 20XX. The upgrade caused a partial change in the protocol frame format, resulting in an interruption of the panel's monitoring of the Brand A unit. The following describes the complete handling process in this scenario, step by step.

[0061] like Figures 2-9 As shown, it also includes the following steps: Step 1: Frame fingerprint extraction and device frame group separation; The intelligent control panel continuously monitors the raw data stream of the bus, segmenting candidate data frames based on the frame interval, and collects a total of several candidate data frames. For each frame, the frame length, first byte value, last byte value, and histogram of byte distribution within the frame are extracted. Figure 4 Each component is Z-score standardized and then combined to form a frame fingerprint feature vector. Density clustering is then performed on all feature vectors. The clustering results divide the candidate data frames into three device frame groups, corresponding to air conditioners A, B, and C, respectively. A small number of frames whose features deviate from the cluster center are temporarily stored in the unassigned frame buffer pool.

[0062] Table 1. Frame fingerprint features and clustering results of candidate data frames:

[0063] The density clustering operation assigns frames 001 and 004 to brand A, frame 002 to brand B, and frame 003 to brand C. Frame 005 is temporarily stored in the unassigned frame buffer pool because its distance from each cluster center exceeds the assignment threshold, reflecting the buffering logic when the assignment confidence is insufficient.

[0064] Step 2: Address mapping between device frame groups and physical hosts; The panel sequentially sends polling request frames carrying addresses 01H, 02H, and 03H to the bus, recording each response frame within a preset response time window. Based on the distance between the frame fingerprint feature vector of the response frame and each cluster center, the response frame is assigned to the corresponding device frame group, completing the address binding.

[0065] Table 2 Polling Response and Address Mapping Results:

[0066] All three device frame groups completed address binding with the physical host through polling responses. In subsequent steps, the protocol processing results of each device frame group can be accurately mapped to the specific air conditioning unit in the computer room.

[0067] Step 3: Frame structure segmentation and protocol conversion rule generation (Brand A group and Brand B group); Frame structure segmentation was performed independently for Brand A and Brand B respectively. Taking Brand A as an example, through frame header and tail feature detection and byte alignment analysis, the 18-byte frame was divided into a fixed header (bytes 1 to 2), a variable data area (bytes 3 to 16), and a checksum tail (bytes 17 to 18). Value variation statistics were performed on each byte position in the variable data area across all frame samples. Bytes 3 and 4 consistently had fixed values, classified as fixed constant fields; byte 5 only switched between 0x01 and 0x02, classified as a status flag field; and bytes 6 and 7 showed a continuous gradual change in value, classified as numeric fields.

[0068] Subsequently, the Pearson correlation coefficient was calculated between the value sequence of the numerical fields and the measured ambient temperature values ​​collected by the panel sensor. Z-score standardization was performed on both sets of data before calculation. The correlation coefficients of bytes 6 to 7 in Brand A's data set exceeded the correlation threshold, and were therefore identified as temperature data fields. The conversion formula was determined to be the actual temperature. ,in This is an unsigned integer value consisting of 6 to 7 bytes. Similarly, the protocol conversion rules for Brand B are generated.

[0069] Further calculation of the Pearson cross-correlation coefficients of the temperature data fields of Brand A and Brand B showed that the results exceeded the co-correlation threshold, confirming that the two sets of temperature fields independently encoded the ambient temperature of the same computer room. The accuracy of the conversion formula was verified by cross-validating the two sets of data, and the offset estimation error of Brand A caused by insufficient early frame samples was corrected.

[0070] Table 3. Segmentation and Field Classification Results of Brand A's Frame Structure:

[0071] Step 4: Inference of Unknown Brand Agreement (Brand C Group); The C-brand unit is a unit with a non-original motherboard replacement; the frame structure analysis in step 3 failed to generate a valid protocol conversion rule. The panel passively listens to the control terminals corresponding to the C-brand unit, collecting timing data of level changes during the air conditioner's autonomous operation.

[0072] A timing correlation analysis was performed between the control terminal level transition events, the sensor temperature change trend, and the command messages on the communication terminals. Observations revealed that after the level of terminal P1 transitioned from low to high, the sensor recorded a temperature decrease trend indicating the start of cooling within a stable time interval of approximately 4.2 to 5.8 seconds, and this pattern consistently appeared in multiple observations. The level transition event of terminal P2 showed a stable sequential relationship with the fixed-format shutdown message on the communication terminals. Based on the above causal timing relationship, it was inferred that terminal P1 corresponds to the cooling start command, and terminal P2 corresponds to the shutdown command. Furthermore, a protocol conversion rule for Brand C was generated by combining the byte semantics of the command messages.

[0073] Table 4. Results of timing correlation analysis of control terminals for Brand C group:

[0074] Step 5: Real-time protocol conversion; After Brand A, Brand B, and Brand C have all completed the generation of protocol conversion rules, the panel calculates the distance between the frame fingerprint feature vector and each cluster center for subsequent bus frames. Data frames are then distributed to the nearest device frame group, and their respective protocol conversion rules are used to parse the proprietary protocol frames into monitoring data in a unified format. Control commands are then encoded according to each brand's protocol and sent to the corresponding address. At this point, all three units are under normal monitoring.

[0075] Step 6: Frame stream compliance monitoring and protocol drift determination (after firmware upgrade for Brand A group); After the firmware upgrade of the Brand A unit was completed in QB of 20XX, the panel calculated the frame-level compliance score for each frame in real time.

[0076] Taking a frame received at a certain moment as an example, the frame length changes from 18 bytes to 20 bytes. The header and footer byte values ​​must match the template. ; There are 3 matches in the fixed constant field, out of a total of 4. Take weights , , ,but Frames below the frame-level compliance threshold of 0.80 are marked as abnormal frames.

[0077] Sliding window statistics show that the percentage of abnormal frames in Brand A group has risen to 0.73%, while the number of newly added frames in the unassigned frame buffer pool is accumulating rapidly, and the frame stream leakage rate is increasing. The calculated result rose to 0.41, exceeding the second preset threshold of 0.30. Both indicators simultaneously exceeded their respective thresholds, triggering a drift alarm for Brand A group, and the alarm trigger time was recorded as T1.

[0078] Table 5. Criteria for Determining Agreement Drift in Brand A Group:

[0079] Step 7: Mutant frame recycling and incremental update of protocol conversion rules; In response to a drift alarm, all unassigned frames accumulated since time T1 are retrieved from the unassigned frame buffer. Using the frame structure template of the baseline frame sample as a reference, the byte value sequence corresponding to fixed constant fields is searched in each frame, and byte-level structural similarity is calculated. Unassigned frames with a similarity exceeding the structural association threshold of 0.70 are marked as mutated frames and recycled to Brand A group, forming the drift device extended frame set.

[0080] Step 701, through field-by-field difference comparison, revealed that the matching position of the fixed constant field shifted 2 bytes to the right, the judgment frame length increased from 18 bytes to 20 bytes, and the newly added 2 bytes were located at positions 5 and 6; the byte width of the numerical temperature field was expanded from 2 bytes to 3 bytes. Step 702 performed a local incremental update: the frame length parameter in the updated frame structure template was set to 20, the byte extraction position of the temperature field was updated to bytes 8 and 10, and the temperature conversion formula was adjusted to... ,in It is an unsigned integer value consisting of 8 to 10 bytes, while the remaining unmutated fields retain the original protocol conversion rules.

[0081] Step 703: Re-extract the frame fingerprint feature vector using the mutated frame and calculate the mean. Take the preset attenuation factor ,according to Integrate existing cluster centers Generate updated cluster centers This allows subsequent frame attribution determinations to gradually adapt to the upgraded frame fingerprint feature distribution.

[0082] Table 6. Identification of Variation Types and Incremental Updates for Brand A Group:

[0083] Step 8: Drift recovery verification and template library update; The updated cluster centers and protocol conversion rules were used to re-perform attribution calculations and protocol parsing for subsequent Brand A group bus frames. After a period of sliding window statistics, the frame flow escape rate decreased to 0.02, and the average frame-level compliance score rebounded to 0.96, both indicators returning to the normal range, and the drift alarm status was cleared. The panel wrote the updated Brand A group protocol conversion rules and frame fingerprint features into the protocol template library, forming a new template entry for direct use when encountering Brand A groups with the same model and upgraded firmware.

[0084] Throughout the process, data originates from the original bus frame stream. Frame fingerprint extraction forms distinguishable feature vectors, density clustering separates the mixed frame stream into three independent device frame groups, address mapping precisely maps these device frame groups to physical units, and frame structure segmentation and temporal correlation analysis generate protocol conversion rules for known and unknown brands, respectively. During the real-time protocol conversion phase, all three units enter normal monitoring status. After a firmware upgrade triggers a drift alarm, the frame stream escape rate and abnormal frame ratio simultaneously pinpoint the drift event at both the device frame group and frame attribution levels. Mutant frame recycling re-incorporates the drifted frames into the Brand A group, local incremental updates only modify the mutated field parameters, and cluster center fusion updates ensure that attribution determination adapts synchronously to the new frame characteristics. Finally, after recovery verification confirms that both indicators have returned to normal ranges, the template library is updated, achieving a complete closed loop from protocol drift detection to automatic recovery.

[0085] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A smart control panel compatible with multiple air conditioners and a protocol conversion method, characterized in that, include: Listen to the raw data stream on the shared communication bus, divide the data stream into candidate data frames according to the frame interval time, extract the frame fingerprint feature vector of each candidate data frame, perform density clustering operation on all frame fingerprint feature vectors, classify candidate data frames belonging to the same cluster into the same device frame group, and temporarily store frames with a home confidence score lower than the allocation threshold into the unassigned frame buffer pool. Frame structure segmentation is performed independently for each device frame group, generating frame structure templates and protocol conversion rules; The frame-level compliance score of the real-time frame stream is calculated based on the frame structure template, and the frame stream escape rate of the unassigned frame buffer pool is calculated. When the proportion of abnormal frames in a certain device frame group exceeds the first preset threshold and the frame stream escape rate simultaneously exceeds the second preset threshold, it is determined that the air conditioning host corresponding to the device frame group has experienced protocol drift. Unassigned frames after the drift alarm trigger time are extracted from the unassigned frame buffer pool. The byte-level structural similarity between the unassigned frames and the reference frame samples of the device frame group where the protocol drift occurred is calculated. Unassigned frames that exceed the structural association threshold are recycled as mutated frames. Based on the mutated frames, the protocol conversion rules are locally incrementally updated and the cluster centers are merged and updated.

2. The intelligent control panel and protocol conversion method compatible with multiple air conditioners according to claim 1, characterized in that, The frame fingerprint feature vector includes four components: frame length, first byte value, last byte value, and intra-frame byte distribution histogram. The intra-frame byte distribution histogram is the distribution feature obtained by performing frequency statistics on all byte values ​​in the frame according to a preset interval. Before combining them into the frame fingerprint feature vector, Z-score standardization is performed on each component.

3. The intelligent control panel and protocol conversion method compatible with multi-brand air conditioners according to claim 1, characterized in that, After generating the device frame group, the process also includes an address mapping step between the device frame group and the physical host: For each device frame group, polling request frames carrying different address codes are sent sequentially to the bus via the panel; Record the response frames received within the preset response time window after each polling. Determine the device frame group to which the response frame belongs based on the distance between the frame fingerprint feature vector of the response frame and the existing cluster center. Bind the device frame group to which the response frame belongs to the bus address of the current polling to generate the address mapping relationship between the device frame group and the physical air conditioning host.

4. The intelligent control panel and protocol conversion method for multi-brand air conditioner compatibility according to claim 1, characterized in that, The process of independently performing frame structure segmentation for each device frame group, generating frame structure templates and protocol conversion rules, includes: The frame boundary patterns are determined by frame header and tail feature detection and byte alignment analysis. The data within the frame is divided into a fixed header region, a variable data region, and a check tail region to generate a frame structure template. Statistical analysis of byte position value changes is performed on all frame samples of the device frame group for the variable data region. Byte positions with constant values ​​are classified as fixed constant fields, byte positions with values ​​that switch only between finite discrete values ​​are classified as status flag fields, and byte positions with values ​​that show a continuous gradual change trend are classified as numerical fields.

5. The intelligent control panel and protocol conversion method for multi-brand air conditioner compatibility according to claim 4, characterized in that, The generation protocol conversion rules also include: The Pearson correlation coefficient is calculated after performing Z-score standardization on the changing trend of the numerical fields and the measured ambient temperature values ​​collected by the panel sensor terminals. When the Pearson correlation coefficient between a certain numerical field and the measured ambient temperature value exceeds the correlation threshold, the field is marked as a temperature data field. The temperature encoding method and conversion formula are determined based on the correspondence between the temperature data field value and the measured ambient temperature value. Calculate the Pearson cross-correlation coefficients for similar numerical fields across different device frame groups. When the Pearson cross-correlation coefficients exceed the coordination threshold, cross-validation is performed using cross-device source fields to correct misjudgments of encoding methods caused by insufficient frame samples or data noise in a single device frame group. The verified temperature encoding method and conversion formula are then written into the protocol conversion rules of the corresponding device frame group.

6. The intelligent control panel and protocol conversion method for multi-brand air conditioner compatibility according to claim 1, characterized in that, For device frame groups that fail to generate valid protocol conversion rules through frame structure segmentation, the following are also included: Collect timing data of level changes of the control terminals corresponding to the frame group of the device during the autonomous operation of the air conditioner; The timing data of the level changes of the control terminal are correlated with the sensor temperature change trend and the command message on the communication terminal within the same time window. When the level flip event of a certain control terminal precedes the start time of a specific temperature change trend and the time interval is stably within a preset range in multiple observations, it is determined that the function of the control terminal corresponds to the temperature change behavior. When the level flip event and a command message of a certain fixed format on the communication terminal have a stable temporal relationship, it is determined that the level state of the control terminal corresponds to the running instruction represented by the command message. Based on causal timing relationships, the terminal function definitions and protocol field semantics are inferred to generate protocol conversion rules for the device frame group.

7. The intelligent control panel and protocol conversion method compatible with multi-brand air conditioners according to claim 1, characterized in that, The frame-level compliance score is calculated as follows: the real-time received data frames are compared field by field with the frame structure template of the corresponding device frame group, and the frame length matching degree component, the frame header and tail conformance component, and the fixed constant field consistency component are weighted and summed according to preset weights. Among them, the frame length matching degree component is full marks when the actual frame length matches the template frame length, and zero marks otherwise; the frame header and tail conformance component is the ratio of the actual number of consistent bytes to the total number of bytes at the corresponding position in the template; the fixed constant field consistency component is the ratio of the number of fixed constant fields in the actual frame that are completely consistent with the template to the total number of fixed constant fields. The frame stream spillover rate is the ratio of the number of frames added to the unowned frame buffer pool per unit time to the total number of frames received on the bus in the same unit time.

8. The intelligent control panel and protocol conversion method for multi-brand air conditioner compatibility according to claim 1, characterized in that, The calculation method for the byte-level structural similarity is as follows: taking the frame structure template of the benchmark frame sample as a reference, searching for the byte value sequence corresponding to the fixed constant field of the frame structure template in the unassigned frame, and calculating the ratio of the number of successfully matched fixed constant byte value sequences to the total number of fixed constant fields of the frame structure template. The local incremental update includes: performing a field-by-field difference comparison between the mutated frame and the reference frame sample to identify the mutation types such as frame length change, field width change, and value domain offset; for frame length change, updating the frame length parameter and field offset table in the frame structure template; for field width change, updating the byte extraction position and length parameter of the corresponding field in the protocol conversion rule; and for value domain offset, updating the offset constant in the conversion formula of the corresponding field in the protocol conversion rule.

9. The intelligent control panel and protocol conversion method for multi-brand air conditioner compatibility according to claim 1, characterized in that, The method for fusion and updating cluster centers is as follows: extract frame fingerprint feature vectors again using mutated frames, calculate the mean of frame fingerprint feature vectors of all mutated frames, and use a preset attenuation factor to weight and fuse the mean of the frame fingerprint feature vectors with the original cluster centers to generate updated cluster centers; wherein the value of the preset attenuation factor ranges from zero to one. The updated clustering centers and protocol conversion rules are used to re-execute the attribution calculation and protocol parsing for subsequent bus frames. When the frame flow escape rate and frame-level compliance score are both restored to the normal range, the drift alarm status is cleared, and the updated protocol conversion rules and frame fingerprint features are written into the protocol template library of the panel. If the frame flow escape rate and frame-level compliance score are not restored to the normal range within the preset recovery time window, the complete protocol analysis process is re-executed for the device frame group where protocol drift occurred.

10. A multi-brand air conditioner compatible intelligent control panel and protocol conversion system, used to execute the multi-brand air conditioner compatible intelligent control panel and protocol conversion method according to any one of claims 1 to 9, characterized in that, include: The frame fingerprint extraction and separation module is used to monitor the raw data stream on the shared communication bus, divide the data stream into candidate data frames according to the frame interval time, extract the frame fingerprint feature vector of each candidate data frame, perform density clustering operation on all frame fingerprint feature vectors to classify the candidate data frames into device frame groups, and temporarily store frames with a home confidence score lower than the allocation threshold into the unassigned frame buffer pool. The protocol conversion rule generation module is used to independently perform frame structure segmentation for each device frame group, and generate frame structure templates and protocol conversion rules. The protocol drift detection module is used to calculate the frame-level compliance score of the real-time frame stream based on the frame structure template, calculate the frame stream escape rate of the unassigned frame buffer pool, and determine that the air conditioning host corresponding to the device frame group has experienced protocol drift when the proportion of abnormal frames in a certain device frame group exceeds the first preset threshold and the frame stream escape rate simultaneously exceeds the second preset threshold. The mutated frame recycling and update module is used to extract unassigned frames after the drift alarm trigger time from the unassigned frame buffer pool, calculate the byte-level structural similarity between the unassigned frames and the reference frame samples of the device frame group where the protocol drift occurred, recycle unassigned frames that exceed the structural association threshold as mutated frames, perform local incremental updates on the protocol conversion rules based on the mutated frames, and merge and update the cluster centers.

Images