Turnout switch machine health assessment method

Abstract

The invention discloses a turnout switch machine health assessment method, and relates to the technical field of turnout switch machine health assessment, and the method comprises the following steps: S1, collecting image data and sensor data of a turnout switch machine; s2, inputting image data and processing the image data based on machine vision to obtain image index data; inputting sensor data to obtain sensor index data; s3, scoring the image index data by adopting a threshold scoring method; scoring the sensor index data by adopting a threshold scoring method and a similarity and standard interval scoring method; and S4, calculating the total score and the health degree of the turnout switch machine. According to the method, all indexes influencing the health of the switch machine are collected to comprehensively evaluate the switch machine, so that one-sided or incomplete evaluation results of a single value are avoided; different schemes for converting different collected data into score values are provided, all indexes can be integrated, and the comprehensive health degree is obtained.

Description

Technical Field

[0001] The invention relates to the technical field of health assessment of turnout machines, and more particularly to a health assessment method for turnout machines. Background Art

[0002] A switch is a switching device for a turnout, which is used to switch turnsouts, lock turnsouts, and reflect the position of the turnout point rail. When a train switches from one track to another or enters a maintenance yard or factory track, a special arrangement must be made on the track to guide the wheel flange to smoothly enter the designated track. This arrangement is called a turnout. The turnout structure mainly includes a set of switch machines, a switch core, two guard rails, and a row of switch sleepers. A set of switch machines consists of two movable switch machines, a set of slide plates, switch rods, and switch machine signs.

[0003] The structure of railway switch machines is complex, the number is large, and the outdoor environment is changeable. The quality of their operating status is not only related to their own status, but also affected by many uncertain factors such as geographical environment and climate. Failures are inevitable during use. If the failure is not repaired in time, it will lead to serious consequences. Therefore, timely and accurate understanding of the operating status of switch machines and research on their health assessment methods are of great significance and practical application value for implementing switch machine maintenance and improving equipment reliability.

[0004] In the prior art, the patent with publication number CN118171164A discloses a method for monitoring the health of turnout switch equipment, including: collecting the action current data of the subway switch, and also including the following steps: constructing a subway switch action current data set; building an MLP-DQN model; training an action current curve model; identifying the action current data and inputting it into the training action current curve model; judging whether the action current data output by the training action current curve model is abnormal. The method for monitoring the health of turnout switch equipment proposed by the present invention can intelligently monitor whether the operation status of the subway switch is normal.

[0005] The health monitoring method for turnout switch equipment disclosed in the above patent only collects the current data of the turnout machine, and uses the model to predict whether the operating current is abnormal to evaluate the health. It has the following defects: In the actual working environment, the health of the turnout switch machine is affected by many factors, including the creep value of the point rail, the opening distance of the point rail, the turnout gap value, the contact gap of the automatic switch, the contact depth of the automatic switch, vibration acceleration, temperature and humidity, conversion force, motor current, etc. The detection data collected by the above patent is only current data, and the detection result is single and incomplete, resulting in inaccurate health assessment. Summary of the invention

[0006] In order to overcome the defects in the above-mentioned prior art, the present invention discloses a health assessment method for a turnout switch machine. The purpose of the present invention is to solve the problem that the detection data collected in the prior art is only current data, and the detection result is single, partial and incomplete. The present invention conducts a comprehensive assessment of the turnout machine by collecting all indicators that affect the health of the turnout machine, avoiding the one-sided or incomplete evaluation result of a single value.

[0007] In order to achieve the above objectives, the technical solution adopted by the present invention is:

[0008] A health assessment method for a turnout machine comprises the following steps:

[0009] 1. Data Collection

[0010] S1, collecting image data and sensor data of the turnout machine;

[0011] In the present invention, a camera can be used to collect image data of the turnout machine, including collecting images of the point rail, the turnout gap, and the automatic switcher. Sensors can be used to collect operating data of the turnout machine, including collecting vibration acceleration, temperature and humidity, conversion force, and motor current.

[0012] 2. Data Input

[0013] S2, input image data and process the image data based on machine vision to obtain image index data; input sensor data to obtain sensor index data;

[0014] Preferably, the image data includes the image data of the point rail, the switch gap and the automatic switch.

[0015] Preferably, the image index data include the point rail creep value, the point rail opening distance, the switch gap value, the automatic switch contact gap, and the automatic switch contact depth.

[0016] Preferably, the sensor index data includes vibration acceleration, temperature and humidity, conversion force, and motor current.

[0017] (1) Point rail creep value

[0018] In the present invention, the point rail creep refers to the abnormal displacement of the turnout point rail (the point rail is the movable part of the turnout, used to guide the train to switch from one track to another) along the rail direction due to various reasons. The specific value of the displacement is called the point rail creep value. This displacement may cause the position of the point rail to deviate from its designed position, thereby affecting the normal function of the turnout and may even cause train driving safety problems. Its impact on the health of the turnout switch is:

[0019] 1. Reduced positioning accuracy: The creep of the point rail causes the switch machine to be unable to accurately position the point rail, which may cause the turnout to be switched incompletely and affect the normal function of the turnout.

[0020] 2. Increased mechanical load: The creep of the point rail will increase the resistance of the switch machine during the switch conversion, causing the mechanical parts to bear additional loads, increase wear and shorten the service life of the equipment.

[0021] 3. Increased failure rate: As the point rail deviates from its normal position, the switch machine may frequently experience failures such as slow movement and jamming, and the failure rate increases significantly.

[0022] 4. Misjudgment of the control system: The creep of the point rail may cause the switch machine indication circuit to feedback incorrect point rail position information, affecting the judgment of the interlocking system and posing a safety hazard.

[0023] 5. Increased maintenance requirements: Point rail creep causes the switch machine to require frequent adjustment and maintenance, which increases maintenance workload and costs and affects the long-term stable operation of the equipment.

[0024] (2) Opening distance of point rail

[0025] In the present invention, the distance of the switch rail refers to the distance that the switch rail moves from the position close to the base rail to the position completely separated from the base rail during the switch switching process. This distance is usually an important parameter for switch switching, which is used to ensure that the train wheels can pass through the switch smoothly without collision or jamming. Its impact on the health of the switch machine is:

[0026] 1. Increased mechanical wear: Failure to completely separate the point rail will cause the wheels to rub against the point rail when passing, increasing the wear of mechanical parts such as the point rail and the switch connecting rod, and accelerating their aging.

[0027] 2. Increased failure rate: Frequent mechanical friction and impact will increase the movement resistance of the switch machine, leading to jamming, incomplete conversion and other failures, and increasing the failure rate.

[0028] 3. Misjudgment of the interlocking system: Since the point rail has not reached the designed opening distance, the circuit may not be able to correctly identify the position of the point rail, resulting in the interlocking system being unable to make accurate judgments, increasing the potential safety hazard of driving.

[0029] 4. Increased maintenance requirements: Failure to meet the opening standards of the point rail will lead to frequent failures of the switch machine, increase maintenance requirements, and affect the long-term stable operation of the equipment.

[0030] 5. Safety hazards: Failure to completely separate the point rail may cause interference between the wheels and the point rail, which may lead to major accidents such as train derailment in serious cases.

[0031] (3) Turnout gap value

[0032] In the present invention, the turnout gap value refers to the gap between the inspection post and the gap wall of the indicating rod after the inspection post of the automatic switch is driven into the gap of the indicating rod inside the switch machine. The indicating gap reflects the gap size between the point rail and the base rail when locked. If the indicating gap is too large or too small, it may affect the accurate indication of the point rail, and in serious cases, it may cause safety accidents such as train derailment. Its impact on the health of the turnout switch machine is:

[0033] 1. Delayed or stuck action: If the indication gap is too small, the indication circuit will malfunction when the point rail is not fully in place. The switch machine may therefore lock or switch when the point rail is not fully in place, increasing the possibility of delayed or stuck action.

[0034] 2. Misjudgment and safety hazards: A bad indication gap may lead to misjudgment of the indication circuit, making it impossible for the interlocking system to correctly determine whether the turnout is in a safe state, increasing driving safety hazards.

[0035] 3. Frequent failures: Repeated malfunctions or misjudgments caused by poor indication notches will accelerate the wear of the internal mechanical and electrical components of the switch machine and increase the failure rate.

[0036] 4. Increased maintenance workload: Frequent failures caused by poor notches will increase maintenance workload and reduce the long-term reliability and stability of the equipment.

[0037] (4) Automatic switch contact gap

[0038] In the present invention, the contact gap of the automatic switch refers to the gap existing after the moving contact and the static contact of the automatic switch come into contact, and its influence on the health of the turnout switch machine is:

[0039] 1. Poor electrical contact: The gap causes unstable electrical contact, which may cause poor or intermittent signal transmission, resulting in the switch machine being unable to operate or the loss of the position status information of the turnout.

[0040] 2. Maloperation or refusal to operate: Poor contact will cause abnormal action signals of the switch machine, which may cause the switch machine to malfunction or refuse to operate, affecting the reliability of turnout switching.

[0041] 3. Arcing and overheating risks: Gaps may cause arcing, which may cause the switch to overheat, increase the risk of equipment failure, and may even cause an electrical fire.

[0042] 4. Frequent failures and increased maintenance requirements: Due to poor contact, the switch machine may frequently fail, resulting in increased maintenance requirements and reduced operating stability and life of the equipment.

[0043] 5. Safety hazards: The gap between the dynamic and static contacts may cause the interlocking system to misjudge the state of the turnout, increase the safety hazard of driving, and may cause train accidents.

[0044] (5) Automatic opener and closeer contact depth

[0045] In the present invention, the contact depth of the automatic switch refers to the distance between the contact point between the moving contact ring and the static contact piece and the outermost edge of the static contact piece after the moving contact of the automatic switch is driven into the static contact. Its influence on the health of the turnout switch machine is:

[0046] 1. Unstable electrical contact: Insufficient contact depth leads to reduced contact area and unstable electrical contact, which may cause poor contact or poor signal transmission and affect the normal operation of the switch machine.

[0047] 2. Arc generation and component wear: Insufficient contact depth can easily cause arcing at the contact point, accelerating the erosion and wear of the contact surface, reducing the service life of the contact and increasing the risk of failure.

[0048] 3. Maloperation or refusal to move: Due to insufficient contact depth, the switch machine may not be able to reliably execute the action command, resulting in maloperation or refusal to move, affecting the switching accuracy and safety of the turnout.

[0049] 4. Increased maintenance frequency: Failures caused by poor contact depth will cause the switch machine to require more frequent maintenance and repairs, increasing maintenance costs and workload and affecting the long-term stability of the equipment.

[0050] 5. Increased safety hazards: Poor contact depth may cause the interlocking system to misjudge the turnout status, increase safety hazards in railway operations, and may cause accidents.

[0051] (6) Vibration acceleration

[0052] In the present invention, vibration acceleration refers to the installation of a vibration sensor inside the turnout switch machine to monitor the vibration data of the turnout switch machine in real time online when the turnout is pulled and the vehicle passes. Its impact on the health of the turnout switch machine is:

[0053] 1. Accelerated wear of mechanical parts: Excessive vibration will cause abnormal stress on the internal mechanical parts of the switch machine (such as gears, connecting rods, bearings, etc.), increase the wear rate, and shorten the service life of the parts.

[0054] 2. Loose or damaged connections: Excessive vibration can cause the internal connections (such as bolts and fasteners) of the switch machine to loosen or become damaged, which may cause the switch machine to operate, become unstable or fail.

[0055] 3. Turnout loss: Frequent or severe vibrations may affect the operation of the switch machine, making it impossible for the turnout to be accurately positioned, increasing the safety risk of train operation.

[0056] 4. Sensor false alarm or failure: Excessive vibration may cause the vibration sensor to false alarm or be damaged due to exceeding the designed vibration range, thus reducing the reliability of the monitoring system and affecting the fault warning function.

[0057] 5. Increased failure rate: Due to various problems caused by vibration (such as component wear, loose connections, and reduced accuracy), the failure rate of the switch machine will increase significantly, affecting the overall reliability and safety of the system.

[0058] (7) Temperature and humidity

[0059] In the present invention, temperature and humidity refer to adding temperature and humidity sensors inside the switch machine to monitor the temperature and humidity inside the switch machine in real time online. Its impact on the health of the turnout switch machine is:

[0060] 1. Accelerated aging of electrical components: High temperature will accelerate the aging of electrical components inside the switch machine (such as motors, cables, etc.), shorten their service life, and increase the occurrence rate of failures.

[0061] 2. Deterioration of insulation performance: A high humidity environment will cause the insulation performance of electrical equipment to deteriorate, increase the risk of short circuit and leakage, and affect the normal operation and safety of the switch machine.

[0062] 3. Corrosion and rust: Excessive humidity will cause corrosion or rust of the internal metal parts of the switch machine (such as connectors and mechanical transmission parts), accelerate the deterioration of mechanical parts, and affect the mechanical properties of the switch machine.

[0063] 4. Mechanical lubrication failure: High temperature and high humidity may cause the internal lubricant of the switch machine to deteriorate or fail, increase the friction and wear of mechanical parts, and affect the smoothness and reliability of the switch switching.

[0064] 5. Reduced sensor accuracy: Excessively high temperature and humidity may affect the accuracy of internal sensors (including newly added temperature and humidity sensors, vibration sensors, power sensors, etc.) and other monitoring equipment, resulting in false alarms or inability to accurately monitor the health status of the switch machine.

[0065] (8) Conversion power

[0066] In the present invention, conversion force refers to the mechanical resistance that the switch machine needs to overcome when the switch is converted from one position to another (i.e., when the point rail is pressed against one side to the other side). This resistance includes the friction force when the point rail moves, the mechanical resistance of the switch components, and the additional resistance caused by environmental factors (such as temperature, humidity, debris, etc.). By adding sensors to monitor the conversion resistance in real time and display the conversion force curve in real time, its impact on the health of the switch machine is:

[0067] 1. Excessive wear of mechanical parts: Excessive conversion resistance will increase the load on the internal mechanical parts of the switch machine (such as motors, gears, and connecting rods), accelerate the wear and aging of these parts, and shorten their service life.

[0068] 2. Motor overload and failure: Excessive switching resistance may cause the switch motor to overload, increase current consumption, cause the motor to overheat or even burn, and cause failure.

[0069] 3. Slow movement or jamming: Greater resistance can cause the switch machine to move slowly or jam during the switching process, affecting the normal switching of the turnout and increasing driving safety hazards.

[0070] 4. Interlocking system misjudgment: Excessive resistance may cause the point rail to fail to fully reach its position, indicating that the circuit cannot correctly identify the position of the point rail, and the turnout loses its position indication, affecting driving safety.

[0071] 5. Increased maintenance requirements: Due to frequent failures and component wear caused by excessive conversion resistance, transfer machines require more frequent maintenance and repairs, increasing operating costs.

[0072] (10) Motor current

[0073] In the present invention, the motor current refers to adding a current sensor on the lead-out line side of the switch machine motor to monitor the current value in the turnout switching process in real time online, and generating the action current curve of the turnout at the back end to analyze the running state of the turnout. Its impact on the health of the turnout machine is:

[0074] 1. Motor overload or damage: Excessive current may cause the switch machine motor to overload. Long-term operation in an overload state will accelerate the aging of the motor or cause motor damage.

[0075] 2. Unstable operation: Abnormal current fluctuations may cause unstable operation of the switch machine, resulting in slow movement, jamming or failure to complete the switch conversion, affecting the normal function of the switch.

[0076] 3. Increased mechanical wear: Abnormal current reflects that there may be abnormal resistance or friction in the mechanical system, which leads to increased wear of mechanical parts and shortens the life of the equipment.

[0077] 4. Increased maintenance frequency: Abnormal current usually indicates that there may be problems inside the equipment, resulting in an increase in the failure rate, requiring increased maintenance frequency, and affecting the long-term stability of the equipment.

[0078] In the present invention, the evaluation indicators include the point rail creep value, the point rail opening distance, the turnout gap value, the automatic switch contact gap, the automatic switch contact depth, the vibration acceleration, the temperature and humidity, the conversion force, and the motor current. Through the above evaluation indicators, a comprehensive health assessment of the turnout switch machine can be carried out.

[0079] III. Index Scoring

[0080] S3. Use the threshold scoring method to score the image index data; use the threshold scoring method, similarity and standard interval scoring method to score the sensor index data;

[0081] 1. Threshold score

[0082] Preferably, in step S3, the point rail creep value, point rail opening distance, switch gap value, automatic switch contact gap, automatic switch contact depth in the image index data, and the vibration acceleration, temperature and humidity in the sensor index data are scored using the threshold scoring method.

[0083] Preferably, in step S3, the threshold score includes:

[0084] For the point rail creep value, the threshold score is: full score for relative displacement ≤19 mm, linear score from 100 to 60 points between 19 mm and 21 mm, and 0 points for the rest;

[0085] For the open track, the threshold score is: full score within 180±2mm, linear score from 100 to 60 points between 180±2mm and 180±4mm, and 0 points for the rest;

[0086] For the turnout notch value, the threshold score is: full score within 2±0.4mm, linearly from 100 to 60 points between 2±0.4mm and 2±0.6mm, and 0 points for the rest;

[0087] For the contact gap of the automatic opener and closer, the threshold score is: full score for no gap, linear score from 100 to 60 points between 0 and 0.2 mm, and 0 points for the rest;

[0088] For the contact depth of the automatic opener and closer, the threshold score is: >4 mm full score, 100 to 60 points linearly between 4 and 3.8 mm, and 0 points for the rest;

[0089] For vibration acceleration, the threshold score is: full score within ±5g, otherwise 0 points;

[0090] For temperature and humidity, the threshold scores are: full marks when the temperature is between -38℃ and 68℃ and the humidity is ≤93%; the temperature increases linearly from 50 to 30 points from -38℃ to -42℃ or 68℃ to 72℃, and the humidity increases linearly from 50 to 30 points from 93% to 97%; the rest are 0 points.

[0091] 2. Similarity and standard interval scoring

[0092] Preferably, in step S3, the conversion force and motor current in the sensor indicator data are scored using a similarity and standard interval scoring method, and the scores are calculated using Pearson similarity and standard range.

[0093] The following is an example of current flow:

[0094] Preferably, in step S3, the calculation score of the motor current in the sensor index data includes: the primary action current of the turnout machine is divided into three stages: the starting stage, the conversion stage, and the inspection stage, and the three stages are scored separately.

[0095] In the present invention, during the operation of the switch machine, the changes in the values ​​of each stage are very important. In order to avoid the influence of uneven number of samples in each stage, the primary action current of the turnout switch machine is divided into three stages: starting stage, conversion stage, and inspection stage according to the working condition of the switch machine, and evaluated separately. Then, the average score of the three stages is used to obtain the final current score.

[0096] Preferably, in step S3, scoring the three stages separately includes: collecting the normal current action curve when the switch machine is working, calculating the number of samples with the highest probability in each stage as the standard number, and performing linear interpolation or sampling on the remaining non-standard data to keep the sample length consistent.

[0097] The number of samples with the highest probability in each of the above stages refers to the mode of the number of sampling points of different samples. It is set as the standard number because the time of each switch machine action is different, and the different sampling numbers under the same collection frequency cannot be compared. The standard number is used for the subsequent calculation of Pearson similarity. The remaining non-standard data includes samples with less or more sampling points than the standard number.

[0098] Preferably, in step S3, for the startup phase and the inspection phase, the median of each sample point is taken to form a new curve as the standard curve, and the Pearson similarity is used as the score value to calculate the current score of the startup phase and the inspection phase.

[0099] In the present invention, for the startup phase and the inspection phase, the waveform changes greatly and there is an obvious linear law. The median of each sample point is taken to form a new curve as the standard curve, and the Pearson similarity is used as the score value. The greater the similarity, the higher the score. The Pearson coefficient is used to measure the strength of the linear correlation between two variables X and Y, and is defined as the quotient of the covariance of the two variables and the product of their standard deviations.

[0100] In the present invention, the variable X of the Pearson coefficient represents the sampling point data of the standard current curve, and the variable Y represents the sampling point data of the current curve to be evaluated.

[0101] Preferably, in the Pearson similarity, the Pearson coefficient is used to measure the strength of the linear correlation between two variables x and y, and is defined as the quotient of the covariance of the two variables and the product of their standard deviations:

[0102]

[0103]

[0104] cov(x,y)=E(xy)-E(x)E(y)

[0105] Among them, ρ xy is the overall correlation coefficient; cov(x,y) is the covariance of x and y, σ x is the standard deviation of x, σ y is the standard deviation of y, E is the expected value, and x and y are variables.

[0106] Preferably, in step S3, for the conversion stage, the interquartile range is calculated for each sample point, the upper and lower limits of the data distribution are determined, and then the ratio of samples falling between the upper and lower limits of the data to the total samples is calculated as the score value, including:

[0107] IQR = Q3-Q1;

[0108] Lower limit = Q1-1.5*IQR;

[0109] Upper limit=Q3+1.5*IQR;

[0110] Among them, IQR is the interquartile range, Q3 is the third quartile, and Q1 is the first quartile.

[0111] In the present invention, for the conversion stage, the value threshold changes very little and the fluctuation is irregular. The interquartile range is calculated for each sample point to determine the upper and lower limits of the data distribution, and the ratio of samples falling between the upper and lower limits to the total samples is calculated as the score value.

[0112] IV. Calculation of all indicator scores

[0113] S4, using the score of the image index data, the score of the sensor index data and the weight of each index, to obtain the total score and health of the turnout machine;

[0114] In the present invention, the creep value of the point rail, the opening distance of the point rail, the gap value of the turnout, the contact gap of the automatic switch, the contact depth of the automatic switch, and the temperature, humidity, and vibration acceleration obtained by the sensor are scored in combination with technical standards and expert experience. The conversion force and motor current obtained by the sensor are used to calculate the score through Pearson similarity and standard range.

[0115] At this point, all the data that affect the health of the turnout machine have been converted into standard scores. According to expert experience, each indicator is weighted and scored to obtain the final health of the turnout machine.

[0116] Preferably, in step S4, the total score of the turnout machine is:

[0117]

[0118] Among them, Score 转辙机 is the total score of the turnout machine, W i is the weight of the i-th indicator, Score i is the score of the i-th indicator, and i represents the i-th indicator.

[0119] Preferably, the weights of the indicators include:

[0120] The weight of the point rail creep value is 0.08, the weight of the point rail opening distance is 0.08, the weight of the turnout gap value is 0.15, the weight of the automatic switch contact gap is 0.15, the weight of the automatic switch contact depth is 0.1, the weight of the vibration acceleration is 0.07, the weight of the temperature and humidity is 0.05, the weight of the conversion force is 0.2, and the weight of the motor current is 0.12.

[0121] Preferably, the total score and health of the turnout machine include:

[0122] When 90 ≤ total score ≤ 100, the health level is health level 4, long-term health;

[0123] When 80≤total score<90, the health level is health level 3, healthy;

[0124] When 60≤total score<80, the health level is healthy level 2, sub-healthy;

[0125] When the total score is less than 60, the health level is healthy level 1 and faulty.

[0126] 5. Health prediction

[0127] S5. Use the long short-term memory network to make the next prediction for the image index data and sensor index data, and return to S3 and S4 steps to score and predict the next health of the turnout machine.

[0128] In the present invention, the health assessment of the turnout machine is a post-evaluation of the completion of the turnout machine action. For the rail transit industry, preventive detection is essential to avoid major failures and impacts. All the above indicators affecting the health will change with the turnout action. Here, the long short-term memory network LSTM is introduced to make the next step of prediction for all the above collected indicators, and the predicted value is scored using the aforementioned weights to obtain the predicted health.

[0129] Beneficial effects of the present invention:

[0130] The health assessment method for turnout machines provided by the present invention collects all indicators that affect the health of turnout machines to conduct a comprehensive assessment of the turnout machines, thereby avoiding one-sided or incomplete assessment results of a single value; different schemes for converting different collected data into scoring values ​​are provided, so that all indicators can be integrated to obtain a comprehensive health degree. BRIEF DESCRIPTION OF THE DRAWINGS

[0131] Figure 1 A schematic diagram of a health assessment method for a turnout machine according to the present invention;

[0132] Figure 2 The present invention is a schematic diagram of a switch machine in which the primary operating current is divided into three stages: a starting stage, a conversion stage, and an inspection stage. DETAILED DESCRIPTION

[0133] The concept, specific structure and technical effects of the present invention will be clearly and completely described below in conjunction with the embodiments and drawings to fully understand the purpose, characteristics and effects of the present invention.

[0134] Example 1

[0135] A health assessment method for turnout machines, such as Figure 1 As shown, the following steps are included:

[0136] S1, collecting image data and sensor data of the turnout machine;

[0137] S2, input image data and process the image data based on machine vision to obtain image index data; input sensor data to obtain sensor index data;

[0138] S3. Use the threshold scoring method to score the image index data; use the threshold scoring method, similarity and standard interval scoring method to score the sensor index data;

[0139] S4, using the score of the image index data, the score of the sensor index data and the weight of each index, to obtain the total score and health of the turnout machine;

[0140] S5. Use the long short-term memory network to make the next prediction for the image index data and sensor index data, and return to S3 and S4 steps to score and predict the next health of the turnout machine.

[0141] Example 2

[0142] This embodiment further describes step S1 based on embodiment 1.

[0143] In step S1, a camera is used to collect image data of the turnout machine, including the image of the point rail, the turnout gap, and the automatic switch. Sensors are used to collect operation data of the turnout machine, including vibration acceleration, temperature and humidity, conversion force, and motor current.

[0144] Example 3

[0145] This embodiment further describes step S2 based on embodiment 2.

[0146] In step S2, the image data includes the image data of the point rail, the switch gap, and the automatic switch. The image index data includes the creep value of the point rail, the opening distance of the point rail, the switch gap value, the contact gap of the automatic switch, and the contact depth of the automatic switch. The sensor index data includes vibration acceleration, temperature and humidity, conversion force, and motor current.

[0147] In this embodiment, the evaluation indicators include the point rail creep value, the point rail opening distance, the turnout gap value, the automatic switch contact gap, the automatic switch contact depth, the vibration acceleration, the temperature and humidity, the conversion force, and the motor current. Through the above evaluation indicators, a comprehensive health assessment of the turnout machine can be carried out subsequently.

[0148] Example 4

[0149] This embodiment further describes step S3 based on embodiment 3.

[0150] 1. Threshold score

[0151] In step S3, the point rail creep value, point rail opening distance, turnout gap value, automatic switch contact gap, automatic switch contact depth in the image index data, and the vibration acceleration, temperature and humidity in the sensor index data are all scored using the threshold scoring method.

[0152] Threshold scores include:

[0153] For the point rail creep value, the threshold score is: full score for relative displacement ≤19 mm, linear score from 100 to 60 points between 19 mm and 21 mm, and 0 points for the rest;

[0154] For the point rail opening, taking the 9# turnout as an example, the threshold score is: full score within 180±2mm, linearly from 100 to 60 points between 180±2mm and 180±4mm, and 0 points for the rest;

[0155] For the turnout notch value, the threshold score is: full score within 2±0.4mm, linearly from 100 to 60 points between 2±0.4mm and 2±0.6mm, and 0 points for the rest;

[0156] For the contact gap of the automatic opener and closer, the threshold score is: full score for no gap, linear score from 100 to 60 points between 0 and 0.2 mm, and 0 points for the rest;

[0157] For the contact depth of the automatic opener and closer, the threshold score is: >4 mm full score, 100 to 60 points linearly between 4 and 3.8 mm, and 0 points for the rest;

[0158] For vibration acceleration, the threshold score is: full score within ±5g, otherwise 0 points;

[0159] For temperature and humidity, the threshold scores are: full marks when the temperature is between -38℃ and 68℃ and the humidity is ≤93%; the temperature increases linearly from 50 to 30 points from -38℃ to -42℃ or 68℃ to 72℃, and the humidity increases linearly from 50 to 30 points from 93% to 97%; the rest are 0 points.

[0160] 2. Similarity and standard interval scoring

[0161] In step S3, the conversion force and motor current in the sensor indicator data are scored using the similarity and standard interval scoring method, and the scores are calculated using the Pearson similarity and standard range.

[0162] The following is an example of current flow:

[0163] During the operation of the switch machine, the changes in the values ​​of each stage are very important. In order to avoid the influence of uneven sample numbers in each stage, the primary action current of the turnout switch machine is divided into three stages: starting stage, conversion stage, and inspection stage according to the working conditions of the switch machine, and evaluated separately. Figure 2 shown.

[0164] Collect a large number of normal action curves, calculate the number of samples with the highest probability in each stage as the standard number, and perform linear interpolation or sampling on the remaining non-standard data to keep the sample length consistent.

[0165] For the startup and inspection phases, the waveform changes greatly and has an obvious linear pattern. The median of each sample point is taken to form a new curve as the standard curve, and the Pearson similarity is used as the score value. The greater the similarity, the higher the score. The Pearson coefficient is used to measure the strength of the linear correlation between two variables X and Y. It is defined as the quotient of the covariance of the two variables and the product of their standard deviations, that is:

[0166]

[0167] cov(x,y)=E(xy)-E(x)E(y)

[0168] Among them, ρ xy is the overall correlation coefficient; cov(x,y) is the covariance of x and y, σ x is the standard deviation of x, σ y is the standard deviation of y, E is the expected value, and x and y are variables.

[0169] For the conversion stage, the value threshold changes very little and the fluctuation is irregular. By calculating the interquartile range for each sample point, the upper and lower limits of the data distribution can be determined, and the ratio of samples falling between the upper and lower limits to the total samples is calculated as the score value.

[0170] IQR = Q3-Q1;

[0171] Lower limit = Q1-1.5*IQR;

[0172] Upper limit=Q3+1.5*IQR;

[0173] Among them, IQR is the interquartile range, Q3 is the third quartile, and Q1 is the first quartile.

[0174] Example 5

[0175] This embodiment further describes step S4 based on embodiment 4.

[0176] The point rail creep value, point rail opening distance, turnout gap value, automatic switch contact gap, automatic switch contact depth obtained through machine vision image processing, and the temperature, humidity, and vibration acceleration obtained through sensors are scored in combination with technical standards and expert experience. The conversion force and motor current obtained by the sensor are used to calculate the score through Pearson similarity and standard range.

[0177] At this point, all the data that affect the health of the turnout machine have been converted into standard scores. According to expert experience, each indicator is weighted and scored to obtain the final health of the turnout machine.

[0178]

[0179] Among them, Score 转辙机 is the total score of the turnout machine, W i is the weight of the i-th indicator, Score i is the score of the i-th indicator, and i represents the i-th indicator.

[0180] Specifically, the weight of each indicator is:

[0181] Table 1 Index weights

[0182] index Weight index Weight Point rail creep value 0.08 Vibration acceleration 0.07 Turnout gap value 0.15 Temperature and humidity 0.05 Automatic opener and closeer contact gap 0.15 Conversion Power 0.2 Automatic opener and closeer contact depth 0.1 Motor current 0.12 Point rail opening 0.08

[0183] The total score and health level are as follows:

[0184] Table 2 Total score and health level

[0185] Total score range Health 90≤X≤100， Long-term health, health level 4 80≤X＜90 Health, Health Level 3 60≤X＜80 Sub-health, healthy level 2 X＜60 Fault, Health Level 1

[0186] Example 6

[0187] This embodiment further describes step S5 based on embodiment 5.

[0188] The health assessment of the turnout machine is a post-evaluation of the turnout machine's operation. For the rail transit industry, preventive testing is essential to avoid major failures and impacts. All of the above indicators that affect the health will change with each turnout operation. Here, the long short-term memory network LSTM is introduced to make the next step of prediction for all the above collected indicators, and the predicted values ​​are scored using the aforementioned weights to obtain the predicted health.

[0189] The above is a specific description of the implementation mode of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention, and these equivalents or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A method for health assessment of a turnout machine, characterized in that: The following steps are involved: S1, collecting image data and sensor data of the turnout machine; S2, input image data and process the image data based on machine vision to obtain image index data; input sensor data to obtain sensor index data; S3. Use the threshold scoring method to score the image index data; use the threshold scoring method, similarity and standard interval scoring method to score the sensor index data; S4, using the score of the image index data, the score of the sensor index data and the weight of each index, to obtain the total score and health of the turnout machine; S5. Use the long short-term memory network to make the next prediction for the image index data and sensor index data, and return to S3 and S4 steps to score and predict the next health of the turnout machine.

2. A turnout machine health assessment method as claimed in claim 1, characterized in that: The image index data include the point rail creep value, the point rail opening distance, the switch gap value, the automatic switch contact gap, and the automatic switch contact depth; the sensor index data include vibration acceleration, temperature and humidity, conversion force, and motor current.

3. A turnout machine health assessment method as claimed in claim 2, characterized in that: In step S3, the point rail creep value, point rail opening distance, turnout gap value, automatic switch contact gap, automatic switch contact depth in the image index data, and the vibration acceleration, temperature and humidity in the sensor index data are all scored using the threshold scoring method.

4. A method for health assessment of a turnout machine as claimed in claim 3, characterized in that: In step S3, the threshold score includes: For the point rail creep value, the threshold score is: full score for relative displacement ≤19 mm, linear score from 100 to 60 points between 19 mm and 21 mm, and 0 points for the rest; For the open track, the threshold score is: full score within 180±2mm, linear score from 100 to 60 points between 180±2mm and 180±4mm, and 0 points for the rest; For the turnout notch value, the threshold score is: full score within 2±0.4mm, linearly from 100 to 60 points between 2±0.4mm and 2±0.6mm, and 0 points for the rest; For the contact gap of the automatic opener and closer, the threshold score is: full score for no gap, linear score from 100 to 60 points between 0 and 0.2 mm, and 0 points for the rest; For the contact depth of the automatic opener and closer, the threshold score is: >4 mm full score, 100 to 60 points linearly between 4 and 3.8 mm, and 0 points for the rest; For vibration acceleration, the threshold score is: full score within ±5g, otherwise 0 points; For temperature and humidity, the threshold scores are: full marks when the temperature is between -38℃ and 68℃ and the humidity is ≤93%; the temperature increases linearly from 50 to 30 points from -38℃ to -42℃ or 68℃ to 72℃, and the humidity increases linearly from 50 to 30 points from 93% to 97%; the rest are 0 points.

5. A method for health assessment of a turnout machine as claimed in claim 2, characterized in that: In step S3, the conversion force and motor current in the sensor indicator data are scored using the similarity and standard interval scoring method, and the scores are calculated using the Pearson similarity and standard range.

6. A method for health assessment of a turnout machine as claimed in claim 5, characterized in that: In step S3, the calculation score of the motor current in the sensor index data includes: the primary action current of the turnout machine is divided into three stages: the starting stage, the conversion stage, and the inspection stage, and the three stages are scored separately.

7. A method for health assessment of a turnout machine as claimed in claim 6, characterized in that: For the startup and inspection phases, the median of each sample point is taken to form a new curve as the standard curve, and the Pearson similarity is used as the score value to calculate the current score of the startup and inspection phases; In the Pearson similarity, the Pearson coefficient is used to measure the strength of the linear correlation between two variables x and y, and is defined as the quotient of the covariance of the two variables and the product of their standard deviations: cov(x,y)=E(xy)-E(x)E(y) Among them, ρ xy is the overall correlation coefficient; cov(x,y) is the covariance of x and y, σ x is the standard deviation of x, σ y is the standard deviation of y, E is the expected value, and x and y are variables.

8. A method for health assessment of a turnout machine as claimed in claim 6, characterized in that: In step S3, for the conversion stage, the interquartile range is calculated for each sample point to determine the upper and lower limits of the data distribution, and then the ratio of samples falling between the upper and lower limits of the data to the total samples is calculated as the score value, including: IQR = Q3-Q1; Lower limit = Q1-1.5*IQR; Upper limit=Q3+1.5*IQR; Among them, IQR is the interquartile range, Q3 is the third quartile, and Q1 is the first quartile.

9. A method for health assessment of a turnout machine according to claim 1, characterized in that: In step S4, the total score of the turnout machine is: Among them, Score 转辙机 is the total score of the turnout machine, W i is the weight of the i-th indicator, Score i is the score of the i-th indicator, and i represents the i-th indicator.

10. A health assessment method for a turnout machine according to claim 9, characterized in that: In step S4, the weights of each indicator include: The weight of the point rail creep value is 0.08, the weight of the point rail opening distance is 0.08, the weight of the switch gap value is 0.15, the weight of the automatic switch contact gap is 0.15, the weight of the automatic switch contact depth is 0.1, the weight of the vibration acceleration is 0.07, the weight of the temperature and humidity is 0.05, the weight of the conversion force is 0.2, and the weight of the motor current is 0.12; The overall score and health of the turnout machine include: When 90 ≤ total score ≤ 100, the health level is health level 4, long-term health; When 80≤total score<90, the health level is health level 3, healthy; When 60≤total score<80, the health level is healthy level 2, sub-healthy; When the total score is less than 60, the health level is healthy level 1 and faulty.

Images