CN-122022018-A - Full-chain health prediction and energy efficiency optimization method for track traffic rotating equipment

CN122022018ACN 122022018 ACN122022018 ACN 122022018ACN-122022018-A

Abstract

The invention relates to the technical field of equipment health prediction, in particular to a full-chain health prediction and energy efficiency optimization method of rail transit rotating equipment, which comprises the steps of stage marking, converting and transferring a proportion by a Markov chain, recursively estimating a stage proportion, converting a life remaining proportion by a life occupation proportion, weighting and summarizing the life remaining proportion and a stage punishment coefficient to obtain a continuously extrapolated health index track, updating a multi-unit state according to time steps, executing cutting and writing back on a boundary of direction consistency, increment homodromous, phase difference threshold value and power consumption, the Kalman filtering selects an updating path according to residual mean square and outputs a posterior interval set, interval boundaries and sensitivity difference linkage generate marginal contribution sequences and merge and sort, meanwhile, a trigger list is formed by recording crossing mileage points, power, retrogradation, limiting files and energy consumption under gradient, curve and station distance indexes are calculated and linked, posterior interval slope penalty items are overlapped to screen out boundary crossing items, a working condition action index table is formed, chain level degradation direction consistency is enhanced, and false triggering and missed triggering probability convergence is achieved.

Inventors

HE SHAOMING
CHEN JIANGLIN
Pan Zhichun
HE HUA
CHEN LINXIAO

Assignees

中交(广州)铁道设计研究院有限公司
广州鹏远智能设备有限公司

Dates

Publication Date: 20260512
Application Date: 20260107

Claims (10)

1. The full-chain health prediction and energy efficiency optimization method for the track traffic rotating equipment is characterized by comprising the following steps of: S1, addressing a table for an axle box bearing, a gear pair, a traction motor rotor and a brake disc based on a full chain unit list of rail transit rotary equipment, pairing state components and observed quantity item by item, solidifying symbol constraint, and establishing a constraint set; s2, based on the constraint set, sequentially sorting key performance index sequences along mileage and cycle times, calculating neighboring point change rate summarizing attenuation rate parameters, marking degradation phase numbers according to change rate segmentation rules, marking conversion ratios by stages and pushing a stage duty ratio sequence by a Markov chain, fusing life occupation ratios and stage penalty coefficients, and summarizing to obtain a health index track; s3, based on the health index track, updating the multi-unit state according to time steps and calculating a residual sequence, performing consistent direction, increment homodromous, phase difference threshold value and power consumption boundary cutting and writing back item by item, adopting Kalman filtering, and selecting an updating path according to a residual mean square comparison result to obtain a posterior interval set; s4, based on the posterior interval set, taking life difference in the perturbation state to generate a sensitivity table, multiplying and adding posterior mean value to obtain a marginal contribution sequence, merging a coupling, a gear box, an axle box and a traction motor, and sequencing and marking crossing mileage points to obtain a trigger list; And S5, selecting power, regeneration and limiting files according to the gradient, the curve and the station distance based on the trigger list, calculating energy consumption, adding a multi-unit degradation posterior interval set slope penalty term, and eliminating a boundary crossing term to obtain a working condition action index table.
2. The method of claim 1, wherein the constraint set comprises a unit identification set, a state quantity naming set, an observed quantity naming set, a symbol constraint set, and a load level segmentation set, the healthy finger trajectory comprises a chain health index sequence, a unit health index sequence, a degradation phase number sequence, a phase duty cycle sequence, and a life remaining duty cycle sequence, the posterior interval comprises a rotor unbalance interval, a bearing radial clearance interval, a tooth flank clearance drift interval, a brake friction attenuation interval, and a confidence boundary set, the trigger list comprises a crossing mileage point set, a pre-warning threshold set, a shutdown threshold set, a unit contribution ordering set, and a unit priority set, and the action index comprises a gradient level entry, a curve radius level entry, an inter-station distance level entry, a traction power level entry, a regenerative braking ratio level entry, and a torque limiting level entry.
3. The method for predicting full chain health and optimizing energy efficiency of a track traffic rotating equipment according to claim 1, wherein the specific step of generating the constraint set is: Based on a full chain unit list of the track traffic rotating equipment, addressing a table for an axle box bearing, a gear pair, a traction motor rotor and a brake disc, writing a state component field into each unit, pairing with an observed quantity field item by item, writing a symbol constraint value according to a pairing relation, and generating a pairing constraint table; Based on the pairing constraint table, performing one-to-one correspondence verification on the state component field and the observed field, deleting symbol constraint conflict lines, writing missing lines in a complementary mode, writing a load level segmentation boundary value, locking a segmentation interval, and establishing a constraint set.
4. The method for predicting the health and optimizing the energy efficiency of a full chain of track traffic rotating equipment according to claim 1, wherein the specific step of generating the health index track is as follows: Based on the constraint set, sorting key performance index sequences along the mileage and cycle times sequence, grouping according to unit numbers, sorting according to time indexes, deleting missing points and reverse sequence points, locking a continuous interval, writing interval start-stop indexes, and generating a performance sequence table; based on the performance sequence table, generating a change rate sequence by adjacent point difference, aggregating the change rate according to window length, outputting window statistics values, eliminating mutation windows according to abnormal threshold values, reserving continuous windows, summarizing attenuation rate parameters, writing into unit rows, and generating a degradation annotation table; Based on the degradation labeling table, a Markov chain is adopted, a phase labeling count is used for converting the transfer proportion and pushing a phase duty sequence, a life occupation proportion is used for converting the life residual duty ratio, a phase punishment coefficient and the life residual duty ratio are added according to weight and written into a chain assembly line, and a health index track is obtained.
5. The method for predicting the health and optimizing the energy efficiency of the full chain of the track traffic rotating equipment according to claim 4, wherein the Markov chain is characterized in that a phase set is firstly limited to be normal, light degradation, moderate degradation and heavy degradation, a phase labeling sequence is read along the sequence of mileage and circulation times, the occurrence times of two adjacent time phases are counted and written into a transfer count matrix, the transfer count matrix is normalized line by line to obtain a transfer proportion matrix, a matrix multiplication is carried out on a current time phase duty vector and the transfer proportion matrix, the number of prediction steps is iteratively updated, a phase duty sequence of each prediction time phase is output, and the phase duty sequence is converted into a phase punishment coefficient sequence according to a preset mapping table and written into a chain summary line.
6. The method for predicting full chain health and optimizing energy efficiency of a track traffic rotating equipment according to claim 1, wherein the specific step of generating the posterior interval set is: Based on the health index track, writing rotor equivalent unbalance amount, bearing equivalent radial clearance, tooth side clearance drift amount and brake friction attenuation amount according to time steps, aligning traction current fluctuation amount, shaft end vibration energy, meshing impact index and brake power consumption index according to unit numbers, calculating an observation difference value, writing into a residual error row, and generating a state residual error table; based on the state residual error table, performing direction consistency judgment on the unbalance increment, performing homodromous judgment on the radial clearance increment, performing phase difference judgment on the tooth side clearance drift and the impact index, performing boundary judgment on friction attenuation and power consumption, performing write-back on out-of-limit items, marking cutting zone bits, and generating a constraint state table; Based on the constraint state table, adopting Kalman filtering, calculating residual mean square according to time steps, comparing the residual mean square with a threshold value, writing multi-point samples above the threshold value, taking the sub-points to form an upper boundary and a lower boundary, writing single-point updating below the threshold value, expanding the upper boundary and the lower boundary according to accumulated deviation, and summarizing the upper boundary and the lower boundary according to unit numbers to obtain a posterior interval set.
7. The method for predicting the health and optimizing the energy efficiency of the full chain of the track traffic rotating equipment according to claim 6, wherein the Kalman filtering is characterized in that firstly, a last time step state estimation value, a last time step covariance matrix, a current time step observation vector, an observation matrix, a state transition matrix, a process noise covariance matrix and an observation noise covariance matrix in a constraint state table are read, a prediction state value is calculated, a prediction covariance matrix is calculated, an observation residual vector is calculated, a residual covariance matrix is calculated, a Kalman gain matrix is calculated, the prediction state value is added with a Kalman gain matrix and an observation residual vector multiplication term to obtain an update state value, the Kalman gain matrix is combined with the observation matrix and the prediction covariance matrix to obtain an update covariance matrix, a mean value is obtained after the square of the update state value and the observation vector difference value is compared with a threshold value, the update state value and a multi-point sample set are selected according to the comparison result, an upper boundary and a lower boundary are calculated according to the update covariance matrix and the unit number are summarized to form a posterior interval set.
8. The method for predicting full chain health and optimizing energy efficiency of a track traffic rotating equipment according to claim 1, wherein the specific steps of generating the trigger list are as follows: Taking a disturbance value between the upper bound and the lower bound of each state component interval based on the posterior interval set, calculating the difference of the lifetime shortening amount before and after disturbance, writing into a unit row, and summarizing the difference result according to the state component fields to generate a sensitivity table; Based on the sensitivity table, multiplying the sensitivity value with the posterior mean value item by item and accumulating to obtain a marginal contribution sequence, aggregating and summing according to a coupler, a gear box, an axle box and a traction motor, arranging in descending order, comparing an extrapolation sequence with a threshold lower boundary point by point and recording a crossing mileage point to obtain a trigger list.
9. The method for predicting the full chain health and optimizing the energy efficiency of the track traffic rotating equipment according to claim 8, wherein the passing mileage points, in particular an extrapolated sequence, are converted from a position higher than a threshold lower limit to a position lower than the threshold lower limit on a mileage axis, the extrapolated sequence is scattered into adjacent mileage points and corresponding values according to fixed mileage steps, when a certain mileage point value is higher than the threshold lower limit and a next mileage point value is lower than the threshold lower limit, and when the value of the certain mileage point is equal to the threshold lower limit and the value of the next mileage point is lower than the threshold lower limit, the passing mileage point is judged to exist between the adjacent mileage points, the position of the passing mileage point is determined according to the change ratio of the values in the mileage of the adjacent mileage points, when a plurality of times of passes exist, the first passing mileage point is recorded, the following passing mileage point is not recorded, and the passing mileage point is marked as an empty value when no passing occurs.
10. The method for predicting full chain health and optimizing energy efficiency of the track traffic rotating equipment according to claim 1, wherein the specific steps of generating the working condition action index table are as follows: Based on the trigger list, splicing the grade level, the curve radius level and the inter-station distance level into working condition index keys, enumerating a traction power gear, a regenerative braking proportion gear and a torque limiting gear according to the working condition index keys, writing the working condition index keys into a row, and generating a working condition action candidate list; and based on the working condition action candidate list, converting the unit distance energy consumption value with the speed curve and the traction power, adding the multi-unit degradation posterior interval set slope penalty item and the energy consumption value to obtain an action score, deleting the out-of-range action line, and outputting an entry according to the working condition index key to obtain the working condition action index list.

Description

Full-chain health prediction and energy efficiency optimization method for track traffic rotating equipment Technical Field The invention relates to the technical field of equipment health prediction, in particular to a full-chain health prediction and energy efficiency optimization method for rail transit rotating equipment. Background The technical field of equipment health prediction aims at constructing a computable model for the running state of equipment, and by constructing deterministic functions for a plurality of running parameters, calculating performance degradation tracks and calculating residual life indexes, the running maintenance activities of the equipment are provided with predictability, a correlation structure is constructed for the running state quantity of the equipment, and pattern recognition is performed for abnormal changes, so that potential failure patterns are recognized before actual failure, and decision amounts for scheduling maintenance plans, reducing sudden shutdown events and stabilizing running efficiency are generated. The full-chain health prediction and energy efficiency optimization method of the rail transit rotating equipment aims at establishing a degradation model, a health index model and an energy consumption function model for a plurality of rotating units comprising a motor rotor, a bearing assembly, a gear transmission group and a brake executing mechanism, enabling maintenance planning, operation scheduling and energy consumption adjustment to have predictability and executable performance by calculating equipment performance attenuation tracks and energy usage distribution, reducing unplanned shutdown times in an equipment operation period, improving the quantifiable degree of a health state, reducing energy consumption of a unit operation distance, and enabling the availability of equipment groups to be maintained above a set threshold. In the prior art, a state quantity association structure is split and implemented with abnormal change pattern recognition, when cross-unit symbol consistency constraint is absent, phenomena such as traction load rising, braking power consumption fluctuation, meshing impact phase drift and the like can be respectively interpreted by different functions, degradation direction judgment contradiction occurs under the same working condition, maintenance plan trigger point drift is caused, a residual life index is provided with a single point value rather than an interval, after working condition switching and noise disturbance enter a calculation link, repeated boundary crossing easily occurs near a threshold value, maintenance scheduling is forced to be frequently adjusted, when pattern recognition depends on discrete features and single channel abnormality, a potential fault pattern overlapping scene is difficult to separate, for example, when bearing clearance slowly expands and rotor unbalance synchronously rises, vibration energy and current fluctuation simultaneously change, function certainty is easy to cause single component change, spare part preparation and maintenance window selection deviate units, energy consumption adjustment is often separated from health prediction, constraint expression occurs under the condition adjustment lack of power gear and regeneration proportion adjustment, and after energy consumption parameter adjustment is difficult to trace to source. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a full-chain health prediction and energy efficiency optimization method for track traffic rotating equipment. In order to achieve the aim, the invention adopts the following technical scheme that the full-chain health prediction and energy efficiency optimization method of the track traffic rotating equipment comprises the following steps: S1, addressing a table for an axle box bearing, a gear pair, a traction motor rotor and a brake disc based on a full chain unit list of rail transit rotary equipment, pairing state components and observed quantity item by item, solidifying symbol constraint, and establishing a constraint set; s2, based on the constraint set, sequentially sorting key performance index sequences along mileage and cycle times, calculating neighboring point change rate summarizing attenuation rate parameters, marking degradation phase numbers according to change rate segmentation rules, marking conversion ratios by stages and pushing a stage duty ratio sequence by a Markov chain, fusing life occupation ratios and stage penalty coefficients, and summarizing to obtain a health index track; s3, based on the health index track, updating the multi-unit state according to time steps and calculating a residual sequence, performing consistent direction, increment homodromous, phase difference threshold value and power consumption boundary cutting and writing back item by item, adopting Kalman filtering, and selecting an updating path according to a