A dynamic load balancing distribution system for a reduction gear motor

Abstract

The application relates to the technical field of speed reducer motor control, and discloses a dynamic load balancing distribution system for a speed reducer motor. The system comprises a load data acquisition module, a load rule verification module, an associated network construction module and a balancing distribution and abnormality identification module. The load data acquisition module carries out standardized processing on original load data to generate standardized load data units containing factors such as load values, time stamps and motor identifiers; the load rule verification module determines a load compliance state based on a standardized data searching rule database; the associated network construction module calculates an associated closeness index through shared entities to construct a multi-load associated graph; and the balancing distribution and abnormality identification module generates a dynamic load balancing distribution suggestion in combination with the associated graph and the compliance state. The system realizes associated analysis and dynamic balancing of multiple speed reducer motor loads, improves the accuracy and adaptability of load management, and is suitable for industrial scenes in which multiple motors work cooperatively.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of speed reduction motor control, in particular to a dynamic load balancing distribution system for speed reduction motor. BACKGROUND

[0002] In the process of industrial production and automation equipment operation, as the core component of power transmission, the load state of the speed reduction motor directly affects the running stability, energy consumption level and service life of the equipment. With the complication of industrial scenes, multiple speed reduction motors often need to work cooperatively in a single scene, and the rationality of load distribution among different motors becomes a key factor restricting the overall system efficiency.

[0003] At present, most load management methods of speed reduction motors still rely on independent monitoring and control of a single motor, and lack effective consideration of load correlation in the cooperative scene of multiple motors. For example, in a pipeline transmission system, the load changes of adjacent speed reduction motors have obvious time sequence correlation, but the existing system often only judges the load threshold of a single motor, and it is difficult to identify the chain reaction caused by unbalanced load distribution. At the same time, the collection and processing of load data mostly use non-standardized formats, and the sensor data structures of different motors differ greatly, which makes it difficult to efficiently carry out cross-motor load analysis.

[0004] The existing load balancing strategies are mostly based on preset fixed rules, and lack the ability of adaptive adjustment to dynamic working conditions. When the working condition changes suddenly, the fixed rules often cannot respond in time, which easily leads to an unbalanced state of some motors being overloaded and some motors being insufficiently loaded, and further causes problems such as accelerated equipment wear, increased energy consumption, and even system shutdown. In terms of load anomaly identification, due to the lack of deep mining of the load correlation characteristics of multiple motors, many potential abnormal states are difficult to be discovered in time, which increases the risk of system operation. SUMMARY

[0005] The purpose of the present application is to provide a dynamic load balancing distribution system for speed reduction motor to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a dynamic load balancing distribution system for speed reduction motor, which comprises:

[0007] A load data acquisition module is used to acquire original load data from speed reduction motor sensors, identify and extract element information including load value, time stamp, motor identifier element, map the element information to a preset data field and establish an internal correlation identifier, and generate a standardized load data unit;

[0008] The load rule verification module retrieves rule conditions clauses matching the load type and the allowed threshold value in the rule database based on the standardized load data unit, calculates a rule application complexity factor according to the number of matching clauses and the hierarchical depth, compares the load data with the clause constraint value, the time window and the location range item by item based on the rule application complexity factor, determines attribute compliance conditions, and generates a load compliance state judgment;

[0009] The associated network construction module finds motor identifiers, associated locations and time stamp entities shared between different speed reduction motors based on a plurality of the standardized load data units, obtains a cross-motor association closeness index according to the number and category of shared entities, constructs a graph structure with load entities as nodes and shared relationships as edges based on the cross-motor association closeness index, and establishes a multi-load association graph.

[0010] The balanced allocation and abnormality identification module calculates a balanced score by combining the load compliance state judgment of the associated load based on the multi-load association graph, and generates a dynamic load balanced allocation suggestion.

[0011] Preferably, the obtaining step of the standardized load data unit is:

[0012] Based on the original load data of the speed reduction motor sensor, the sensor field content in the data is scanned and identified item by item, field category identification matching is performed, and time series splitting is performed, and load values, time stamps and motor identifiers are extracted one by one to generate an original load data element set;

[0013] Based on the original load data element set, the mapping matching verification of the fields is performed one by one, and the numerical value format conversion and field reconstruction are performed on the field content matched successfully to generate a load data standardized mapping field set;

[0014] Based on the load data standardized mapping field set, internal association matching is performed between each standardized mapping field, and internal association relationship identifiers are established according to the matching results, and index construction and relationship binding of the standardized mapping field are performed;

[0015] Based on the index construction and relationship binding results, all fields are integrated to generate a standardized load data unit.

[0016] Preferably, the obtaining step of the rule application complexity factor is:

[0017] Based on the load type field and the allowed threshold value field in the standardized load data unit, the original field value and the field definition library are extracted for field content normalization matching, missing fields are completed and field expression methods are unified to obtain a standardized load type field and allowed threshold value field combination set;

[0018] According to the combination set of the normalized load type field and the allowed threshold field, the rule condition clauses in the rule database matching the load type are retrieved in sequence, and the clauses matching the combination set of the allowed threshold field are screened, and the nested level, the number of logical judgments, the clause reference frequency, the clause activation period coverage, and the clause applicable position coincidence degree of each clause are extracted;

[0019] Based on the extraction result, the rule application complexity factor is calculated.

[0020] Preferably, the obtaining step of the load compliance state determination is:

[0021] Based on the rule application complexity factor, the load type field value in the standardized load data unit is called item by item, and the threshold comparison, numerical range verification, and time matching judgment between the field value and the constraint value are performed;

[0022] According to the comparison verification judgment result, the position field value in the standardized load data unit is obtained, the load position field and the rule clause applicable position range are mapped item by item, and the clauses with successful position mapping are screened;

[0023] Based on the screening result, the attribute compliance of each item of load data is determined.

[0024] According to the attribute compliance, the load compliance state determination is generated.

[0025] Preferably, the obtaining step of the cross-motor association closeness index is:

[0026] Based on the plurality of standardized load data units, the motor identifier field, the associated position field, and the timestamp field in each standardized load data unit are extracted, all load data are entity classified and matched and de-duplicated according to field type, and a load sharing entity set is generated;

[0027] According to the load sharing entity set, the number of occurrences of each entity in different motor load is counted and the entity category is marked, and the distribution information of each type of shared entity in each standardized load data unit is recorded;

[0028] Based on the distribution information, the cross-motor association closeness index is calculated.

[0029] Preferably, the obtaining step of the multi-load association graph is:

[0030] Based on the cross-motor association closeness index, the motor identifier field, the associated position field, and the timestamp field in the plurality of standardized load data units are extracted, and the entity normalization identifier coding and entity type annotation are performed on each field content;

[0031] Based on the annotation result, a load entity node set is generated.

[0032] According to the load entity node set, the shared relationship strength value between the load entities is calculated;

[0033] Based on the shared relationship strength value, a graph structure is constructed with the load entities as nodes and the shared relationship strength value as edge weights, and an entity edge connection relationship mapping table between the nodes is established;

[0034] Based on the entity edge connection relationship mapping table, a multi-load association graph is generated.

[0035] Preferably, the acquisition step of the dynamic load balancing distribution suggestion is:

[0036] Based on the multi-load association graph, bidirectional paths of all load entity nodes are extracted from the graph structure, and the corresponding load compliance state judgment result, path length, path endpoint node out-degree, path starting node field number and number of intermediate nodes passed through by the path are retrieved;

[0037] According to the retrieval result, a multi-load path attribute set is generated;

[0038] Based on the multi-load path attribute set, the balance score of the corresponding entity node is calculated;

[0039] According to the balance score, a dynamic load balancing distribution suggestion is generated.

[0040] Preferably, the element information identification and extraction step of the load data acquisition module is:

[0041] Receive the original load data stream of the speed reducer motor sensor and perform data stream segmentation processing;

[0042] Based on the segmentation processing result, identify the load value, timestamp and motor identifier element;

[0043] Map the identified elements to the preset data field;

[0044] Based on the mapping result, establish an internal association identifier;

[0045] Based on the internal association identifier, output the standardized load data unit.

[0046] Preferably, the balance distribution and abnormality identification module further comprises an abnormality processing step, specifically:

[0047] Call the load compliance state judgment in the multi-load association graph; based on the load compliance state judgment, detect abnormal load patterns;

[0048] According to the abnormal load pattern, adjust the dynamic load balancing distribution suggestion; based on the adjustment result, generate a final load balancing distribution output.

[0049] Preferably, the rule condition clause matching the load type with the allowed threshold in the retrieval rule database comprises:

[0050] Querying the load rule set in the rule database, based on the query result, matching the load type with the allowed threshold;

[0051] According to the matching result, extracting the rule condition clause, based on the rule condition clause, performing rule application complexity factor calculation.

[0052] Compared with the prior art, the beneficial effects of the present application are:

[0053] By constructing a standardized load data processing mechanism, the unified integration and correlation analysis of different speed reducer motor load information is realized, and the limitation of traditional single motor independent monitoring is broken. The load data acquisition module maps the original data to the preset field and establishes internal correlation identifier, so that the cross-motor load data has comparability and correlation, which lays a foundation for multi-motor collaborative analysis.

[0054] The load rule verification module introduces the rule application complexity factor, realizes the fine application of load determination rules through the calculation of the number of clauses and the depth of levels. This way avoids the mechanical application of fixed rules, can dynamically adjust the determination logic according to different load types and working condition characteristics, and improves the accuracy and flexibility of load compliance state determination.

[0055] The correlation network construction module calculates the cross-motor correlation closeness index and constructs a multi-load correlation graph based on motor identifiers, positions and time stamps and other shared entities, and intuitively presents the load correlation characteristics between multiple motors. This graph-based presentation makes the hidden load correlation relationship explicit, providing an effective way to identify potential problems caused by correlation imbalance.

[0056] The balanced allocation and abnormality identification module combines the correlation graph and the load compliance state to generate dynamic load balancing allocation suggestions, realizing the transition from static rules to dynamic adaptation. This module can respond to changes in working conditions in real time, adjust the balancing strategy according to the dynamic changes of multi-motor load correlation characteristics, and avoid the response lag problem of traditional fixed strategies when the working condition changes suddenly. At the same time, with the help of multi-load correlation graph, it can more accurately identify the load imbalance state caused by abnormal correlation relationship, expand the dimension of abnormality identification, and reduce the risk of missing judgment of implicit abnormalities. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The timing diagram of the dynamic load balancing distribution system for the speed reducer motor described in the present application;

[0058] Figure 2 The flowchart for obtaining the standardized load data unit;

[0059] Figure 3 Flowchart for obtaining rule applicability complexity factor;

[0060] Figure 4 Flowchart for obtaining multi-load association graph;

[0061] Figure 5 Flowchart for element information identification and extraction of load data acquisition module. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0063] Please refer to Figure 1 The present application provides a dynamic load balancing distribution system for a reduction motor, which comprises a load data acquisition module, a load rule verification module, an association network construction module, and a balancing distribution and abnormality identification module.

[0064] The load data acquisition module acquires original load data from a reduction motor sensor, identifies and extracts load values, time stamps, and motor identifier elements, maps them to preset data fields, establishes internal association identifiers, and generates standardized load data units. The load rule verification module retrieves a rule database based on the standardized load data units to match load type and allowed threshold rule condition clauses, calculates rule applicability complexity factors, compares load data with clause constraint values, time windows, and location ranges item by item, and generates load compliance state judgments. The association network construction module finds motor identifiers, associated locations, and time stamp entities shared between different reduction motors based on multiple standardized load data units, calculates cross-motor association closeness indexes, constructs a graph structure with load entities as nodes and shared relationships as edges, and establishes a multi-load association graph. The balancing distribution and abnormality identification module calculates balancing scores based on the multi-load association graph combined with load compliance state judgments, and generates dynamic load balancing distribution suggestions.

[0065] Embodiment 1: Please refer to Figure 2The load data acquisition module, as the initial link of the system, its core function is to extract effective information from the dispersed sensor raw data and standardize it, providing a unified format data basis for subsequent processing. The module receives raw load data streams from the speed reducer motor sensor. Such data streams usually exist in mixed formats, which may contain binary sensor signals or mixed with text format device state information. The transmission rate of the data stream is related to the motor operating state. When running stably, it is about 500 to 1000 data points per second, and when abnormally fluctuating, it may appear high-frequency pulses for a short time.

[0066] The original data stream is segmented. The basis for segmentation is the time characteristics and structural features of the data: for time-synchronized continuous data, fixed time intervals are used for cutting to ensure that each sub-data block corresponds to a complete sampling period; for non-time-synchronized discrete data, it is segmented according to the data length threshold to avoid processing delay caused by too long data. In the segmentation process, the possible sticky packet problem needs to be handled. When the data of two sampling periods are combined into one data block due to transmission delay, they are identified and split through the check code to ensure the independence and integrity of each sub-data block.

[0067] The segmented sub-data blocks enter the element identification stage. This stage needs to accurately extract the load value, timestamp, and motor identifier from the mixed binary and text data. The load value usually exists in numerical form, either in binary complement form (such as 16-bit signed integer, corresponding to the numerical range of -32768 to 32767) or text floating-point format (such as "150.5"). The pattern matching algorithm is used for identification: for binary data, the load value field is located by parsing the type identifier in the data header (such as 0x01 indicating load value), and then converted to decimal value; for text data, the numerical pattern is matched using regular expressions (such as "[0-9]+(.[0-9]+)?"), and the position of the load value is confirmed in combination with the context (such as the leading character "LOAD="). The identification of the timestamp relies on the time synchronization protocol of the device, common ones are the system time synchronized by NTP (Network Time Protocol) (such as "2025-08-11T10:13:43.430Z") or the counter value of the device local clock (such as the number of milliseconds since startup "123456789"). The timestamp field is located by keyword matching (such as "TS=" "Time:") and then converted to the unified ISO8601 format (such as "2025-08-11 10:13:43.430"). The motor identifier is used to distinguish different motors, usually as a string (such as "MOTOR_001" "DRV-2025A") or a number (such as "001" "2025A"), and is extracted by the fixed position character sequence (such as the string after "ID:" at the beginning of the data block) and verified for its format legality (such as length not exceeding 20 characters, containing only letters, numbers and underscores).

[0068] The identified elements need to be mapped to the preset data fields to ensure standardization. The preset data fields are defined during system initialization, including "load value (unit: N·m)" "timestamp (format: YYYY-MM-DD HH:MM:SS.sss)" "motor identifier (format: string, length ≤20)". The mapping process uses key-value pair matching rules: for load value, the identified value (such as 150.5) is directly assigned to the "load value" field, and its physical reasonableness is checked (such as whether it is within the rated load range of the motor, assuming the rated load is 200 N·m, then 150.5 N·m is a valid value, while -50 N·m or 250 N·m will be marked as abnormal); for timestamp, the converted ISO format string is assigned to the "timestamp" field, and its timeliness is verified (such as the deviation from the current system time is not more than 1 second, otherwise it is considered as outdated data); for motor identifier, the extracted string is assigned to the "motor identifier" field, and the device account is checked (such as whether there is a registration information of "MOTOR_001"), and the identifier of the unregistered device will be marked as unknown device.

[0069] After the mapping is completed, an internal association identifier needs to be established to uniquely identify each standardized load data unit and support fast retrieval. The association identifier is generated using a hash algorithm, specifically the SHA-256 algorithm, to hash the concatenated string of "load value + timestamp + motor identifier". The standardized load data unit is composed of the standardized mapping field set and the internal association identifier.

[0070] During the entire processing process, the processing mechanism for abnormal situations needs to be considered. For example, when there are garbled characters or format errors in the sensor data stream, the data stream segmentation module will trigger the error detection program, mark the sub-data block as "invalid data" and record the error log, to avoid invalid data entering the subsequent processing link. For repeated motor identifiers, the system distinguishes them through the timestamp field to ensure that each data corresponds to a unique sampling time, avoiding data redundancy caused by repeated sending of devices. In addition, to cope with high-concurrency data collection scenarios, the load data collection module uses a multi-threaded processing mechanism, with each data stream allocated an independent processing thread to ensure that the real-time performance and throughput of data processing meet the system requirements.

[0071] Through the above process, the load data collection module converts scattered and heterogeneous sensor raw data into standardized load data units with unified format and complete information, providing a standardized data input for subsequent load rule verification, associated network construction and balanced allocation, and is the basic prerequisite for the operation of the entire dynamic load balancing and allocation system.

[0072] Example 2: see Figure 3 After the standardized load data unit is generated, the system enters the processing phase of rule applicability complexity factor calculation and load compliance state determination. These two links are closely related, the former is used to evaluate the complexity of matching the current load data with the rule conditions, and the latter verifies the compliance of the load data based on the results of the former.

[0073] The calculation of the rule application complexity factor begins with the extraction and normalization of key fields in the standardized load data unit. The standardized load data unit contains a load type field and an allowable threshold field, which are the core basis for rule matching. The system normalizes the original value of the load type field: if the original value is "variable torque load" or "variable torque load", the system will unify it to the standard expression "variable torque load" through a predefined synonym mapping table; if the original value is missing or ambiguous, the system will automatically complete the load type according to the time series characteristics of the load value. For the allowable threshold field, the system checks its compliance with units and dimensions. For example, if the load value unit is "N·m" (Newton-meter), and the unit of the allowable threshold field is "kgf·m" (kilogram-force-meter), the system will convert it to a unified unit through the unit conversion formula (1 kgf·m ≈ 9.8 N·m); if the allowable threshold field is missing dimension information, the system will refer to the standard dimension of the load of this type in the equipment manual to complete it. After the normalization process is completed, the combination of the load type and the allowable threshold forms a standardized input for subsequent rule retrieval.

[0074] The system retrieves the rule database based on the normalized combination set. The rule database stores multi-dimensional and multi-level load rule condition clauses, each clause containing load type matching conditions, allowable threshold constraints, logical judgment relationships, nesting levels, logical judgment quantities, clause reference frequencies, clause activation period coverage, and clause application location coincidence degrees. The system matches the load type field with the load type classification in the rule database to preliminarily match all clauses containing the current load type as a candidate clause set. The allowable threshold field is further matched in the candidate clause set. If the unit and dimension of the current allowable threshold are completely consistent with the constraint value in the clause, or are compatible after unit conversion, the clause is included in the list of clauses to be evaluated.

[0075] For each clause in the list of clauses to be evaluated, the system extracts parameters such as its nesting level, number of logical judgments, clause reference frequency, clause activation period coverage, and clause applicable location coincidence. The nesting level is determined by analyzing the internal structure of the clause: if the clause contains sub-clauses, the nesting level of the main clause is 2 (main clause + sub-clauses); if the clause has no nested structure, the nesting level is 1. The number of logical judgments is obtained by counting the number of judgment conditions in the clause: the above-mentioned clause involving time window and load type involves 2 judgments (load type matching, time window matching), so the number of logical judgments is 2. The clause reference frequency is obtained by querying historical matching records, for example, a certain clause is matched 150 times in the past 7 days, so its reference frequency is 150 times / cycle. The clause activation period coverage is determined by calculating the overlap ratio of the clause applicable time window and the current time window: if the clause applicable time is "2025-08-10 to 2025-08-20" and the current time window is "2025-08-11 10:00 to 2025-08-11 11:00", the coverage is 100%; if the current time window is "2025-08-09 09:00 to 2025-08-09 10:00", the coverage is 0%. The clause applicable location coincidence is calculated by comparing the spatial relationship between the clause applicable location and the current motor location: if the clause applicable location is "Workshop A area" and the current motor location is "Workshop A area B production line", the coincidence is 100%; if the current motor location is "Workshop B area", the coincidence is 0%.

[0076] Based on the above extracted parameters, the system calculates the rule application complexity factor. This factor reflects the overall difficulty of matching the current load data with the rule conditions, and its calculation process takes into account the relative importance of each parameter: the higher the nesting level, the more complex the clause structure, and more sub-conditions need to be considered during matching; the more logical judgments, the more cumbersome the conditions to be verified; the higher the reference frequency, the more versatile the clause, and more careful verification may be required; the lower the activation period coverage, the weaker the applicability of the clause to the current time; the lower the location coincidence, the less targeted the clause is to the current motor location. By weighting these parameters according to pre-set weights (such as nesting level accounting for 30%, number of logical judgments accounting for 20%, reference frequency accounting for 20%, activation period coverage accounting for 20%, and location coincidence accounting for 10%), the rule application complexity factor is finally obtained by weighted summation. The value of this factor usually ranges from 0 to 1, and the closer the value is to 1, the more complex the matching process, and more computing resources need to be invested for verification; the closer the value is to 0, the simpler the matching process, and it can be directly verified by basic conditions.

[0077] The generation of the load compliance status determination is based on the specific values of the rule application complexity factor and the standardized load data unit, aiming to determine whether the current load data meets the rule conditions. The system divides different matching strategies according to the level of the complexity factor: low complexity factor (≤0.3) corresponds to loose matching, only the verification of whether the load value is within the allowed threshold range is required; medium complexity factor (0.3<factor≤0.7) corresponds to standard matching, the load value range, time window (whether the timestamp of the load data is within the effective period specified by the clause) and location range (whether the location corresponding to the motor identifier is within the geographic area specified by the clause) need to be verified at the same time; high complexity factor (>0.7) corresponds to strict matching, in addition to the above verification, the fluctuation trend of the load value (such as whether it exceeds 80% of the threshold value within 3 consecutive sampling periods) and the reference situation of the associated clause (such as whether there is a logical conflict with other matched clauses) also need to be checked.

[0078] In specific implementation, the system compares the load type field value in the standardized load data unit with the clause constraint value by threshold comparison: if the load value is 190 N·m and the maximum allowed load value of the clause constraint is 180 N·m, it is determined that it exceeds the upper limit; if the load value is 170 N·m and within the time window specified by the clause (such as 8:00-18:00), and the motor position belongs to the "Workshop A area" specified by the clause, it is preliminarily determined to be in compliance. The system checks whether the timestamp of the load data falls within the clause activation period based on the timestamp field value of the load data: for example, the clause requires the time window to be "2025-08-10 to 2025-08-20", and the current timestamp is "2025-08-11 10:13:43", the time matching is successful; if the timestamp is "2025-08-09 09:00:00", the time matching fails. The system checks whether the motor identifier field value belongs to the location range applicable to the clause through the position mapping table (such as "MOTOR_003" corresponds to "Workshop A area"): if the applicable location of the clause is "Workshop A area" and the current motor position is "Workshop A area", the location matching is successful; if the motor position is "Workshop B area", the location matching fails.

[0079] According to the results of threshold comparison, time window matching and location range check, the system generates a load compliance status determination. If all three verifications pass (load value compliance, time matching, location matching), it is determined to be "fully compliant"; if only part of it passes (such as load value compliance but time matching fails), it is determined to be "partially compliant", and the specific reason for the mismatch is recorded (such as "time window mismatch"); if any one does not pass (such as load value exceeds the upper limit), it is determined to be "non-compliant", and the specific violation point is marked (such as "load value exceeds the maximum allowed threshold 180 N·m"). For strict matching corresponding to high complexity factors, the system further analyzes the fluctuation trend of the load value: if the current load value is 185 N·m (exceeding the threshold by 5 N·m), and the load values of the previous two sampling periods are 182 N·m and 178 N·m respectively, it is determined to be "continuous over-limit", and the compliance status is marked as "non-compliant"; if the load value only exceeds the threshold temporarily in a single sampling period (such as 185 N·m), and the previous and subsequent periods are below the threshold, it may be marked as "occasional over-limit", which needs to be further verified in combination with other clauses. The load compliance status determination result contains the compliance level (fully / partially / non-compliant), the non-compliant reason (if any) and the corresponding rule clause reference information (such as "rule clause R003"), which provides a key compliance basis for subsequent associated network construction and balanced allocation.

[0080] Embodiment 3: refer to Figure 4 The acquisition process of the cross-machine association tightness index starts with the extraction and classification of key entities in multiple standardized load data units. The standardized load data unit contains a motor identifier field, an associated location field and a timestamp field. The system performs entity classification matching and deduplication processing on these fields: for the motor identifier, the string fuzzy matching algorithm is used to unify "Workshop A" and "A Workshop" into a standard name, generating a location entity set; for the timestamp, the continuous time is divided into discrete windows according to the preset time window (such as 5-minute interval), "2025-08-11 10:13:43" and "2025-08-11 10:14:00" belong to "10:10-10:15" time window, generating a time entity set. After deduplication processing, the load sharing entity set is obtained, which contains the unique identifier of the motor entity, the location entity and the time entity and the number of occurrences in each standardized load data unit.

[0081] The system counts the number of occurrences of each type of shared entity in different motor load data units. For example, the motor entity "MOTOR_001" appears in 10 standardized load data units, the location entity "Workshop A area" appears in 15 units, and the time entity "10:10-10:15" appears in 8 units. At the same time, mark the entity category (motor / position / time), and record the distribution information of each type of shared entity in each standardized load data unit, that is, which motor entities, position entities and time entities are included in each unit. These distribution information is stored in the form of a hash table, with the unique identifier of the standardized load data unit as the key and the entity set associated with the unit as the value.

[0082] The calculation of the cross-motor association closeness index is based on the above distribution information, using an improved Jaccard similarity formula. For any two motors (denoted as motor i and motor j), extract the set of load sharing entities associated with them (that is, the entities that appear in the load data units of both motor i and motor j), and then calculate the proportion of the size of this set to the total number of entities associated with motor i and motor j. The specific formula is:

[0083]

[0084] Where: represents the cross-motor association closeness index between motor i and motor j; represents the number of shared entities associated with both motor i and motor j; represents the total number of entities associated with motor i (including the size of the union of motor entities, position entities and time entities); represents the total number of entities associated with motor j. This formula balances the influence of the difference in the total number of entities associated with motor i and motor j on the result through the square root of the denominator, avoiding the similarity being underestimated due to the excessive number of entities associated with a motor. For example, if motor i is associated with 10 entities and motor j is associated with 16 entities, and they are associated with 6 entities in common, then . Normalize the of all motor pairs (range 0 to 1) to obtain the cross-motor association closeness index matrix, where each element represents the association closeness between two motors. This index reflects the degree of cooperation of different motor loads at the entity level, and the higher the index, the stronger the correlation between the loads of the two motors in the motor, position and time dimensions.

[0085] When constructing the multi-load correlation graph based on the cross-machine association tightness index, a set of load entity nodes is generated. Each load entity node corresponds to a standardized load data unit, and the node attributes include motor identifier, load value, timestamp, compliance status determination result, etc. The unique identifier of the node is generated by combining the hash value of the motor identifier, timestamp, and load value, ensuring the global uniqueness of each node. The shared relationship strength value between load entities is calculated, which integrates the cross-machine association tightness index and the overlap degree of shared entities. The specific calculation method is as follows: for two load entity nodes (correlation motor i) and (correlation motor j), the shared relationship strength value between them is equal to the cross-machine association tightness index multiplied by the proportion of the number of shared entities to the total number of entities of each, that is:

[0086]

[0087] Wherein: represents the shared relationship strength value between load entity node (load data unit of correlation motor i) and load entity node (load data unit of correlation motor j), which quantifies the association tightness between the two load entities; represents the cross-machine association tightness index between gear motor and gear motor ; represents the size of the common association entity set ; represents the total number of entities associated with motor ; represents the total number of entities associated with motor ; the denominator is the size of the union of the total number of entities of motor i and motor j, ensuring that the proportion item reflects the proportion of shared entities in the union. For example, if , , , , then . This strength value quantifies the association strength between two load entities, and the larger the value, the tighter the association.

[0088] The graph structure with load entities as nodes and shared relationship strength values as edge weights is stored through an adjacency list, where each edge connects two load entity nodes, and the edge weight is the corresponding shared relationship strength value. At the same time, a mapping table of entity edge connection relationships between nodes is established, recording the source node, target node, edge weight value and associated shared entity type (such as motor entity, location entity or time entity) of each edge. For example, edge connection node U001 (motor M001, time T01) and node U002 (motor M002, time T01), edge weight is 0.142, and the associated shared entity type is "time entity T01" and "location entity A01". The finally generated multi-load association graph intuitively reflects the association relationship between different speed reducer loads, providing a structured analysis basis for subsequent balanced allocation. The graph supports efficient path query and relationship analysis, and can quickly identify the collaborative influence between loads, thereby optimizing the dynamic load balancing allocation strategy.

[0089] Example 4: see Figure 5 Balanced score calculation and dynamic allocation suggestion generation based on multi-load association graph are the core links to realize precise load regulation. The following will combine specific application scenarios to explain the implementation of this process in detail.

[0090] Suppose there are three speed reducer motors running in a factory, MOTOR_001 (Workshop A area), MOTOR_002 (Workshop A area) and MOTOR_003 (Workshop B area), and the running time range is from 10:00 to 10:30 on August 11, 2025. The system has constructed a multi-load association graph based on the running data of the previous three hours, containing 12 load entity nodes (U001 to U012), each node records the motor identifier, load value, timestamp, compliance status determination result (fully compliant / partially compliant / non-compliant) and associated location information.

[0091] The bidirectional path extraction of the multi-load entity node starts from the full connection analysis of the graph structure. The system traverses all possible node pairs and identifies all connected paths between any two nodes. The path length is defined as the number of edges in the path (i.e. the number of intermediate nodes + 1), for example, node U001 (MOTOR_001, time T01, load 150 N·m) is directly connected with U002 (MOTOR_002, time T01, load 160 N·m), and the path length is 1; node U002 is indirectly connected with U003 (MOTOR_003, time T02, load 170 N·m) through U004 (MOTOR_002, time T02, load 165 N·m), and the path length is 2. The out-degree of the path end node refers to the number of edges (i.e. the number of associations with other nodes) connected outward in the graph structure, for example, U002 is connected with U001, U004, U005, and the out-degree is 3; the field number of the path start node refers to the number of attribute fields (such as motor identifier, load value, timestamp, compliance status, associated position) contained by the node, for example, the field number of U001 is 5; the number of intermediate nodes refers to the number of nodes in the path except the start and end nodes, for example, the number of intermediate nodes of the path U002→U004→U003 is 1 (U004). The system sorts all the extracted path attributes into a multi-load path attribute set, and part of the data is shown in Table 1.

[0092] Table 1: The system sorts all the extracted path attributes into a multi-load path attribute set, and part of the data is shown as follows.

[0093]

[0094] Based on the multi-load path attribute set, the system calculates the balance score of each node. The scoring process considers four dimensions: path length, end node out-degree, start node field number, and intermediate node number. The shorter the path length, the smaller the impact of load adjustment on other nodes, and the higher the weight of the score; the higher the out-degree of the end node, the closer the association between the node and other nodes, the greater the impact of adjusting its load on the overall system, and the higher the weight; the more the field number of the start node, the more comprehensive the load information of the node (containing more dimensions of operation data), the more sufficient the basis for adjustment decision, and the higher the weight; the more the number of intermediate nodes, the more complex the path, and the higher the risk of chain reaction in the adjustment process, and the lower the weight.

[0095] In a specific calculation, the system assigns a weight coefficient to each dimension (path length 0.4, end node out-degree 0.3, start node field number 0.2, and intermediate node number 0.1), and normalizes the original values of each dimension (range 0 to 1). For example, the path length of path P001 is 1 (the maximum path length is 3), and after normalization, it is 1 / 3 ≈ 0.333; the end node out-degree is 3 (the maximum out-degree is 4), and after normalization, it is 3 / 4 = 0.75; the start node field number is 5 (a fixed value), and after normalization, it is 1; the intermediate node number is 0 (the maximum intermediate node number is 2), and after normalization, it is 0. The balanced score of path P001 = 0.4 x 0.333 + 0.3 x 0.75 + 0.2 x 1 + 0.1 x 0 ≈ 0.133 + 0.225 + 0.2 + 0 = 0.558. Similarly, the balanced scores of other paths are calculated, as shown in Table 2.

[0096] Table 2: The balanced scores of other paths are calculated as follows.

[0097]

[0098] According to the balanced scores, the system generates dynamic load balancing distribution suggestions. The node pairs corresponding to paths with higher scores are considered to have lower adjustment costs and more controllable impacts, and therefore are preferred as adjustment targets. For example, path P003 (score 0.712) connects U003 (MOTOR_003, load 170 N·m) and U004 (MOTOR_002, load 165 N·m), both of which have a compliance status of “fully compliant” (the load values do not exceed the respective allowed thresholds 180 N·m and 175 N·m), and the path length is short and the number of intermediate nodes is small, so the system suggests maintaining the current load distribution and no adjustment is needed.

[0099] Path P004 (score 0.485) connects U004 (MOTOR_002, load 165 N·m) and U005 (MOTOR_003, load 190 N·m), where the compliance status of U005 is “partially compliant” (the load value 190 N·m is close to the allowed threshold 195 N·m), and the path length is relatively long (3) and the number of intermediate nodes is large (2), so the system suggests reducing the load value of U005 by 5% (from 190 N·m to 180.5 N·m) to reduce the risk of exceeding the threshold and reduce the impact on associated nodes.

[0100] Path P002 (score 0.621) connects U002 (MOTOR_002, load 160 N·m) and U003 (MOTOR_003, load 170 N·m), the compliance state of U002 is "fully compliant", and the compliance state of U003 is "fully compliant", but the load value of U003 (170 N·m) is much lower than its allowed threshold (200 N·m), the system detects that its load utilization is low, and suggests that part of the load of U002 (about 8 N·m) be transferred to U003 to improve overall load balancing. After adjustment, the load value of U002 is adjusted to 152 N·m (still lower than the threshold 175 N·m), the load value of U003 is adjusted to 178 N·m (still lower than the threshold 200 N·m), and the balancing score of path P002 is expected to be improved to 0.685.

[0101] The abnormality processing step runs throughout the whole process of generating balancing distribution suggestions. The system calls the load compliance state judgment results in the multi-load association graph, and detects abnormal load patterns through pattern recognition algorithms. For example, the compliance state of node U001 (MOTOR_001, time T01 to T03) is "non-compliant" for three consecutive sampling periods, the reason is that the load value (150 N·m, 155 N·m, 160 N·m) continuously exceeds the allowed threshold (140 N·m), and the load value of the associated node U002 (MOTOR_002) (160 N·m, 165 N·m, 170 N·m) also shows a synchronous upward trend, and the system judges it as "persistent associated threshold value abnormality".

[0102] For this abnormal pattern, the system adjusts the dynamic load balancing distribution suggestion: in the original suggestion, the load of U001 is maintained at 150 N·m, which is now adjusted to be reduced by 12% (to 132 N·m), and the monitoring frequency is increased (from every 100 milliseconds to every 50 milliseconds); at the same time, the load adjustment range of U002 is increased from 5% to 8% (from 160 N·m to 147.2 N·m) to share the threshold value load of U001. After adjustment, the load value of U001 (132 N·m) is lower than the threshold (140 N·m), and the compliance state is restored to "fully compliant"; the load value of U002 (147.2 N·m) is still lower than the threshold (175 N·m), and the system running risk is significantly reduced.

[0103] The dynamic load balancing distribution suggestion is output in a structured data form, including the target load value of each motor, the adjustment priority (high-score nodes are given priority), the adjustment time window (such as "complete adjustment before 10:30"), and the abnormality processing explanation (such as "U001 needs to strengthen temperature monitoring"). The suggestion is transmitted to the motor controller through industrial Ethernet to guide it to perform load adjustment operations, ensuring that the speed reducer system runs in a safe and balanced state.

[0104] Embodiment 5: Retrieving rule database matching load type with rule condition clause of allowable threshold and calculating rule applicable complexity factor is the key link of connecting load data collection and compliance status determination, which directly affects the accuracy of subsequent balanced allocation strategy.

[0105] When the standardized load data unit is generated, the system extracts the load type field and the allowable threshold field from the unit as the core input of rule matching. The load type field reflects the current operating mode of the motor, such as "constant torque load", "variable torque load", "impact load", etc., which may come from feature analysis of sensor data (such as the rate of change of load value over time) or pre-set classification of equipment manual. The allowable threshold field contains safety or performance constraints related to the load type, such as "maximum allowable load value", "minimum allowable load value", "load fluctuation threshold", etc., which is usually specified by equipment manufacturers or industry standards.

[0106] To ensure the accuracy of matching, the system normalizes the original value of the load type field. If the original value is "variable torque load" or "variable torque load", the system will unify it to the standard expression "variable torque load" through a predefined synonym mapping table; if the original value is missing or ambiguous (such as only marked as "variable torque"), the system will automatically complete the load type according to the time series characteristics of the load value (such as whether the load value changes periodically over time). After completing the normalization process, the combination set of load type and allowable threshold is formed, which provides standardized input for subsequent rule retrieval.

[0107] The system retrieves the rule database based on the normalized combination set. The rule database stores multi-dimensional, multi-level load rule condition clauses, each clause containing the following key information: load type matching condition, allowable threshold constraint, logical judgment relationship, nesting level, logical judgment quantity, clause reference frequency, clause activation period coverage, clause applicable position coincidence degree, etc.

[0108] The system matches the load type field with the load type classification in the rule database to preliminarily match all clauses containing the current load type as the candidate clause set. For example, if the current load type is "variable torque load", the candidate clause set contains all clauses with "variable torque load" as the matching condition. Further match the allowable threshold field in the candidate clause set, if the unit and dimension of the current allowable threshold are completely consistent with the constraint value in the clause, or can be compatible after unit conversion, then the clause is included in the list of clauses to be evaluated.

[0109] For each clause in the list of clauses to be evaluated, the system extracts parameters such as its nesting level, number of logical judgments, clause reference frequency, clause activation period coverage, and clause applicable location coincidence. The nesting level is determined by analyzing the internal structure of the clause: if the clause contains sub-clauses, the nesting level of the main clause is 2 (main clause + sub-clauses); if the clause has no nested structure, the nesting level is 1. The number of logical judgments is obtained by counting the number of judgment conditions in the clause: the above-mentioned clause involving time window and load type involves 2 judgments (load type matching, time window matching), so the number of logical judgments is 2. The clause reference frequency is obtained by querying historical matching records, for example, a certain clause is matched 150 times in the past 7 days, so its reference frequency is 150 times / cycle. The clause activation period coverage is determined by calculating the overlap ratio of the clause applicable time window and the current time window: if the clause applicable time is "2025-08-10 to 2025-08-20" and the current time window is "2025-08-11 10:00 to 2025-08-11 11:00", the coverage is 100% (complete overlap); if the current time window is "2025-08-09 09:00 to 2025-08-09 10:00", the coverage is 0% (no overlap). The clause applicable location coincidence is calculated by comparing the spatial relationship between the clause applicable location and the current motor location: if the clause applicable location is "Workshop A area" and the current motor location is "Workshop A area B production line", the coincidence is 100% (complete inclusion); if the current motor location is "Workshop B area", the coincidence is 0% (no association).

[0110] Based on the above extracted parameters, the system calculates the rule application complexity factor. This factor reflects the overall difficulty of matching the current load data with the rule conditions, and its calculation process takes into account the relative importance of each parameter: the higher the nesting level, the more complex the clause structure, and more sub-conditions need to be considered during matching; the more logical judgments, the more cumbersome the conditions to be verified; the higher the reference frequency, the more versatile the clause, and more careful verification may be required; the lower the activation period coverage, the weaker the applicability of the clause to the current time; the lower the location coincidence, the less relevant the clause to the current motor location. By weighting and summing these parameters according to pre-set weights, the rule application complexity factor is finally obtained. The value range of this factor is usually between 0 and 1, and the closer the value is to 1, the more complex the matching process and the more computing resources need to be invested for verification; the closer the value is to 0, the simpler the matching process and the basic conditions can be directly verified.

[0111] The calculation result of the rule application complexity factor directly affects the strategy selection of the subsequent load compliance state determination. For example, if the complexity factor is 0.2 (low complexity), the system only verifies whether the load value is within the allowed threshold range; if it is 0.5 (medium complexity), the load value range, time window and location range need to be verified at the same time; if it is 0.8 (high complexity), the fluctuation trend of the load value and the reference situation of the associated clauses also need to be checked. Through such differentiated verification strategies, the system can improve the efficiency of load balancing allocation under the premise of ensuring compliance, and ensure that the speed-reducing motor system operates in a safe and stable state.

[0112] It should be noted that, in this document, the terms such as first and second are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between such entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus including a series of elements includes not only those elements, but also other elements not explicitly listed, or inherent to such a process, method, article or apparatus.

[0113] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A dynamic load balancing distribution system for a geared motor, characterized by, The system comprises: a load data acquisition module for acquiring raw load data from a speed reducer motor sensor, identifying and extracting element information including load value, timestamp, motor identifier element, mapping the element information to a preset data field and establishing an internal correlation identifier, and generating a standardized load data unit; a load rule verification module for retrieving, based on the standardized load data unit, rule condition clauses matching load type and allowed threshold value in a rule database, calculating rule application complexity factors according to the number and level depth of matched clauses, and comparing, based on the rule application complexity factors, load data with clause constraint values, time windows and location ranges, determining attribute compliance conditions, and generating a load compliance status determination; a correlation network construction module for finding, based on a plurality of the standardized load data units, motor identifiers, associated locations and timestamp entities shared between different speed reducer motors, calculating a cross-motor association tightness index according to the number and category of shared entities, constructing a graph structure with load entities as nodes and shared relationships as edges based on the cross-motor association tightness index, and establishing a multi-load association graph; a balanced allocation and anomaly identification module for calculating, based on the multi-load association graph, a balanced score in combination with the load compliance status determination of associated loads, and generating a dynamic load balanced allocation suggestion.

2. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The obtaining step of the standardized load data unit is: based on raw load data of a speed reducer motor sensor, scanning and identifying sensor field content in the data item by item, performing field category identification matching, and performing time series splitting, extracting load value, timestamp and motor identifier one by one, and generating a raw load data element set; based on the raw load data element set, performing field mapping matching verification one by one, and performing numerical format conversion and field reconstruction on the field content of the successfully matched fields, generating a load data standardized mapping field set; based on the load data standardized mapping field set, performing internal correlation matching between each standardized mapping field, and establishing an internal correlation relationship identifier according to the matching result, performing index construction and relationship binding of the standardized mapping field; based on the index construction and relationship binding result, integrating all fields to generate a standardized load data unit.

3. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The obtaining step of the rule application complexity factor is: based on the load type field and the allowed threshold value field in the standardized load data unit, extracting the original field value and the field definition library for field content normalization matching, completing the missing fields and unifying the field expression method, obtaining a standardized load type field and allowed threshold value field combination set; according to the standardized load type field and allowed threshold value field combination set, sequentially retrieving rule condition clauses matching the load type in the rule database, and screening clauses matching the allowed threshold value field and the combination set, extracting the nesting level, logic judgment number, clause reference frequency, clause activation period coverage and clause application location coincidence degree of each clause; based on the extraction result, calculating the rule application complexity factor.

4. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The obtaining step of the load compliance status determination is: Based on the rule application complexity factor, the load type field value in the standardized load data unit is called item by item with the clause constraint value, and threshold comparison, numerical range verification and time matching judgment are performed between the field value and the constraint value. According to the comparison verification judgment result, the position field value in the standardized load data unit is obtained, the load position field is mapped with the application position range of the rule clause item by item, and the clauses with successful position mapping are screened. Based on the screening result, the attribute compliance of each item of load data is determined. According to the attribute compliance, the load compliance state judgment is generated.

5. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The acquisition step of the cross-motor association closeness index is: Based on the plurality of standardized load data units, the motor identifier field, the associated position field and the timestamp field in each standardized load data unit are extracted, all load data are matched and de-duplicated according to field types, and a load sharing entity set is generated; According to the load sharing entity set, the number of occurrences of each entity in different motor load is counted and the entity category is marked, and the distribution information of each type of shared entity in each standardized load data unit is recorded; Based on the distribution information, the cross-motor association closeness index is calculated.

6. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The acquisition step of the multi-load association graph is: Based on the cross-motor association closeness index, the motor identifier field, the associated position field and the timestamp field in the plurality of standardized load data units are extracted, and the entity normalization identification coding and entity type annotation are performed on each field content; Based on the annotation result, a load entity node set is generated; According to the load entity node set, the sharing relationship strength value between load entities is calculated; Based on the sharing relationship strength value, a graph structure is constructed with load entities as nodes and sharing relationship strength values as edge weights, and an entity edge connection relationship mapping table between nodes is established; Based on the entity edge connection relationship mapping table, a multi-load association graph is generated.

7. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The acquisition step of the dynamic load balancing distribution suggestion is: Based on the multi-load association graph, all load entity nodes are extracted from the graph structure, and the load compliance state judgment result, the path length, the out-degree of the path endpoint node, the field number of the path starting node and the number of intermediate nodes passed through by each path are retrieved; According to the retrieval result, a multi-load path attribute set is generated; Based on the multi-load path attribute set, the balance score of the corresponding entity node is calculated; According to the balance score, a dynamic load balancing distribution suggestion is generated.

8. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The element information identification and extraction step of the load data acquisition module is: Receive the original load data stream of the motor sensor, and perform data stream segmentation processing; Based on the segmentation processing result, identify the load value, timestamp and motor identifier elements; Map the identified elements to the preset data field; Based on the mapping result, establish internal association identification; Based on the internal association identification, output the standardized load data unit.

9. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The balance distribution and abnormality identification module further includes an abnormality processing step, specifically: Call the load compliance state judgment in the multi-load association graph; based on the load compliance state judgment, detect the abnormal load mode; According to the abnormal load mode, adjust the dynamic load balancing distribution suggestion; Based on the adjustment result, a final load balancing allocation output is generated.

10. The dynamic load sharing distribution system for a reduction gear motor of claim 1, wherein, The rule condition clause matching the load type with the allowed threshold in the search rule database includes: Querying a load rule set in the rule database, and based on the query result, matching the load type with the allowed threshold; According to the matching result, a rule condition clause is extracted, and based on the rule condition clause, a rule application complexity factor calculation is performed.

Images