A door crane controller with sub-health pre-diagnosis and fault type clustering functions

Through cerebellar neural network and stray coding technology, the switching process feature vectors of rail vehicle doors are extracted, and sub-health prediagnosis and fault type clustering are used to solve the lack of online real-time evaluation of the health of the door machine system in the existing technology. The problem of the situation improves the safety and reliability of the rail vehicle door system.

CN112348078BActive Publication Date: 2024-11-01SHENZHEN POSTMAN TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011222663.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-11-09
Publication Date
2024-11-01
Estimated Expiration
2040-11-09

AI Technical Summary

Technical Problem

The existing rail vehicle door machine controller lacks online real-time sub-health prediagnosis and fault type clustering functions, and cannot effectively evaluate the health status of the door machine system, affecting vehicle safety.

Method used

The cerebellar neural network is used to perform sub-health prediagnosis and fault type clustering. By dividing the switching process of rail vehicle doors into 4 segments, time and displacement parameters are extracted to form input samples, and feature vectors are calculated using stray encoding and Minkovsky distance. Realize detection of the sub-health status and fault type of the door machine system.

Benefits of technology

It realizes online real-time sub-health prediagnosis and fault type clustering of rail vehicle door systems, improves the safety and reliability of vehicle door control, and can promptly alarm and transfer to emergency treatment procedures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112348078B_ABST
    Figure CN112348078B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for judging the sub-healthy state and clustering fault types of a rail vehicle door machine system and a corresponding door machine controller. The data sample of the operating state is extracted from the door opening and closing motion process curve of the door machine system, and is input into the trained CMAC neural network in the rail vehicle door machine controller to obtain the characteristic vector of the operating state. The characteristic vectors corresponding to the normal operating state and the main typical fault state are stored in the door machine controller. By calculating the Minkowski distance between the current operating state characteristic vector and the normal operating state characteristic positive vector, it can be judged whether the door machine system is in a sub-healthy state. Furthermore, by calculating the Minkowski distance between the current operating characteristic vector and the characteristic vectors of each typical state, fault type clustering can be achieved. The present invention also provides the hardware and software structure of the door machine controller with the above functions. The beneficial effect of the present invention is that the safety of rail vehicle operation can be improved by performing sub-health judgment and fault type clustering analysis through the characteristic vector of the door opening and closing curve automatically extracted by the door controller.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of rail vehicle door fault detection, and in particular to a door machine controller with sub-health pre-diagnosis and fault type clustering functions. Background Art

[0002] Rail vehicle doors work in a complex and harsh real-world environment. The safety of rail vehicle doors is the most important part of the overall vehicle safety and is directly related to public safety. At present, general rail vehicle door machine controllers do not have the function of online real-time pre-diagnosis of sub-health and clustering of fault types for rail vehicle door machine systems. In today's developed Internet of Things, online real-time evaluation of the health status of door machine systems is an important means to improve the safety of rail vehicle door systems. Rail vehicle door machine controllers include door machine parts and door controllers. Door machines operate under the control and drive of door controllers. The sub-health pre-diagnosis and fault type clustering functions of door machine systems can be reflected through the characteristic vectors of the door opening and closing curves automatically extracted by the door controller. Summary of the invention

[0003] The object of the present invention is to provide a door machine controller with sub-health pre-diagnosis and fault type clustering functions to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method with sub-health pre-diagnosis and fault type clustering functions, characterized in that: sub-health pre-diagnosis and fault type clustering are realized by using a cerebellar neural network, the switch process of a door body controlled by a door machine system is segmented to form input samples of the cerebellar neural network, the cerebellar neural network includes a virtual memory and a stray code, the stray code value is used as a feature vector, and sub-health pre-diagnosis and fault type clustering are realized according to the Minkowski distance of each feature vector in the vector space, and the process includes the following steps:

[0006] Step 1: Divide the operating parameters of the rail vehicle door in the current door opening and closing process into 4 sections: rising section, constant speed section, deceleration section and pressure speed section, extract the time and displacement in these 4 sections, so as to form the input samples of the cerebellar neural network, the input samples include: rail vehicle door opening and closing speed rising time, rail vehicle door opening and closing speed rising displacement, rail vehicle door opening and closing uniform speed time, rail vehicle door opening and closing uniform speed displacement, rail vehicle door opening and closing speed falling time, rail vehicle door opening and closing speed falling displacement, rail vehicle door opening and closing pressure speed time, rail vehicle door opening and closing pressure speed displacement;

[0007] Step 2: Based on the input samples obtained in step 1, the feature vector of the door opening and closing state of the rail vehicle is obtained after concept mapping and spurious transformation of the cerebellar neural network. The specific process is:

[0008] The input sample S after the switch gate is (k) Input to the door controller, after concept mapping, activate C storage units in the concept memory AC, and then after spurious coding, find the weights of the actual memory for the activated C storage units, read the weights, and output vector W i , vector W i That is the characteristic vector of each door opening and closing state;

[0009] Step 3: Calculate the Minkowski distance between the current running feature vector and the typical feature vector obtained in advance to perform sub-health pre-diagnosis and fault type clustering:

[0010] S1: Calculate the Minkowski distance between the operating feature vector of the fault-free state and the operating feature vector of the typical fault state in the vector space:

[0011]

[0012]

[0013]

[0014]

[0015]

[0016]

[0017]

[0018]

[0019] Find the minimum value of the above Minkowski distance:

[0020] l0=min{D(W1, W2), D(W1, W3),..., D(W1, W9)};

[0021] S2: Calculate the Minkowski distance between the operating feature vector of the rail vehicle door's current door opening and closing state and the operating feature vector of the fault-free state in the vector space:

[0022]

[0023] S3: Compare the shortest Minkowski distance l0 between the running feature vector of the fault-free state obtained by S1 and the running feature vector of the typical fault state in the vector space with the Minkowski distance l between the running feature vector of the current door opening and closing state of the rail vehicle door obtained by S2 and the running feature vector of the fault-free state in the vector space, and perform a sub-health pre-diagnosis of the current door opening and closing state:

[0024] when When the door machine system is in healthy operation;

[0025] when When the door machine system is in sub-healthy operation, the door machine controller will alarm and switch to normal control state;

[0026] when When the door machine system fails in operation, the fault type is clustered, the door machine controller alarms and enters the emergency processing procedure;

[0027] When the rail vehicle door has a door opening and closing fault, the door machine controller clusters the fault types. The clustering conditions are as follows:

[0028] Where W is the operating characteristic vector of the door opening and closing state, W i and W j is the operating characteristic vector of each typical fault state.

[0029] The output feature vector W of the operating state i and W j Including: output vector W1 for no fault, output vector W2 for V-type abnormal fault, output vector W3 for loose motor component fault, output vector W4 for small centering dimension change fault, output vector W5 for large centering dimension change fault, output vector W6 for upper slide outward movement fault, output vector W7 for lower stop pin lateral interference fault, output vector W8 for lower stop pin longitudinal interference fault, and output vector W9 for pressure wheel overpressure fault.

[0030] The C of the cerebellar neural network is the generalization width, which is taken as 6.

[0031] The hash code is a compression mapping, that is, the space of hash values ​​is much smaller than the space of inputs, and different inputs may hash to the same output. Simply put, it is a function that compresses a message of any length to a message digest of a fixed length. Here, the hash code function uses the division and remainder method. The remainder obtained after the keyword is divided by a number p that is not greater than the length of the hash table m is taken as the hash address. That is, H(key) = key MOD p, p≤m. Among them, the divisor p is called the modulus. The key to this method lies in the selection of the modulus p. Make the probability of each keyword in the data element set being mapped to any address of the memory unit through the hash function equal, thereby minimizing the possibility of hash conflicts. One experience of using the division and remainder method is that if the length of the hash table is m, p is usually the smallest prime number less than or equal to the length (preferably close to m) or a composite number that does not contain less than 20 prime factors.

[0032] The typical characteristic vectors include: an operation characteristic vector of a fault-free state, an operation characteristic vector of a V-type abnormal fault state, an operation characteristic vector of a motor component loose fault state, an operation characteristic vector of a centering dimension change small fault state, an operation characteristic vector of a centering dimension change large fault state, an operation characteristic vector of an upper slideway outward movement fault state, an operation characteristic vector of a lower stop pin lateral interference fault state, an operation characteristic vector of a lower stop pin longitudinal interference fault state, and an operation characteristic vector of a pressure wheel overpressure fault state. The characteristic vectors of these nine typical operating states are stored in the rail vehicle door machine controller in advance.

[0033] The Minkowski distance calculation formula is:

[0034] Experiments have shown that when the p value is 1.5, the Minkowski distance formula is most effective in sub-health pre-diagnosis and fault type clustering.

[0035] The present invention provides a control method with sub-health pre-diagnosis and fault type clustering functions and a door machine controller. A control method with sub-health pre-diagnosis and fault type clustering functions comprises the following steps:

[0036] Step 1: H output by the rotor position sensor in the control system with sub-health pre-diagnosis and fault type clustering function a , H b , H c Detection is performed to form a speed feedback signal ω * ;

[0037] Step 2: Output current i of the control system with sub-health pre-diagnosis and fault type clustering function A 、i BDetection is performed and compared with the torque direction signal DIR output by the absolute value circuit and the H output by the rotor position sensor. a , H b , H c The current feedback signal i is synthesized by the current feedback signal f ;

[0038] Step 3: Subtract the speed feedback signal ω from the speed given signal ω * , the speed error is obtained, and the speed error is adjusted by the proportional integral operation to obtain the signal u gi ;

[0039] Step 4: Signal u gi The absolute value circuit outputs the torque control signal MCMD and the aforementioned current feedback signal i f By comparison, the current error is obtained, and this current error is output as PWM through the proportional integral operation with anti-integral saturation.

[0040] Step 5: Output PWM and the H output by the aforementioned rotor position sensor a , H b , H c Together, the three-phase inverter is controlled through the commutation logic to drive the brushless DC motor.

[0041] A door machine controller with sub-health pre-diagnosis and fault type clustering functions uses a brushless DC motor as a driving motor of the door controller, uses a power MOS three-phase bridge circuit to feed the brushless DC motor, and uses an STM32F429 chip as its built-in controller to realize the control of the position, speed and current of the brushless DC motor. Two LEM Hall current sensors are used to detect the current, and the current value of the third phase can be calculated according to the Kirchhoff current criterion, and the high-speed A / D channel of the STM32F429 is used to complete the sampling and conversion of the motor circuit signal. The control software of the door machine controller also includes a position control module, a speed control module, a D-axis current control module, a Q-axis current control module, a rotation transformation module, an inverse rotation transformation module, a speed position calculation software module, a speed position storage module, a torque calculation software module, a torque storage module, a CMAC sample input module, a spurious code mapping module, a feature vector extraction module, a sub-health pre-diagnosis and alarm module, and the automatic door controller also includes a hardware module, which uses a Hall current detector to form a two-phase current detection module, and uses an intelligent power module IPM to form an inverter to realize the drive of the motor.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. The present invention detects and records the operating parameters (time, displacement) of the rail vehicle door opening and closing, and inputs them into the cerebellar neural network as input samples to obtain the operating characteristic vector of the current door opening and closing state, calculates the Minkowski distance in the vector space between the operating characteristic vector of the current door opening and closing state and 9 typical operating characteristic vectors pre-stored in the rail vehicle door machine controller, and performs sub-health pre-diagnosis of the current door opening and closing state and fault type clustering according to the mathematical relationship between the Minkowski distances;

[0044] 2. It can be applied to automatic door control systems in various important occasions. When applied to rail vehicle automatic door systems, it can effectively improve the safety and reliability of door control, thereby improving the safety of the entire train operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The figure is a software and hardware structure diagram of the door machine controller in an embodiment of the present invention.

[0046] Figure 2 1 is a control principle block diagram of a door machine controller in an embodiment of the present invention.

[0047] Figure 3 The schematic diagram is a principle diagram of a door machine controller with a function of automatically extracting characteristic vectors of door opening and closing curves according to an embodiment of the present invention.

[0048] Figure 4 This is a door closing time speed curve of a rail vehicle according to an embodiment of the present invention.

[0049] Figure 5 This is a schematic diagram of the algorithm of the cerebellar neural network software system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] See also Figure 1-5 , the present invention provides a technical solution:

[0052] A method with sub-health pre-diagnosis and fault type clustering functions, characterized in that: sub-health pre-diagnosis and fault type clustering are realized by using a cerebellar neural network, the switch process of a door body controlled by a door machine system is segmented to form input samples of the cerebellar neural network, the cerebellar neural network includes a virtual memory and a stray code, the stray code value is used as a feature vector, and sub-health pre-diagnosis and fault type clustering are realized according to the Minkowski distance of each feature vector in the vector space, and the process includes the following steps:

[0053] Step 1: Divide the operating parameters of the rail vehicle door in the current door opening and closing process into 4 sections: rising section, constant speed section, deceleration section and pressure speed section, extract the time and displacement in these 4 sections, so as to form the input samples of the cerebellar neural network, the input samples include: rail vehicle door opening and closing speed rising time, rail vehicle door opening and closing speed rising displacement, rail vehicle door opening and closing uniform speed time, rail vehicle door opening and closing uniform speed displacement, rail vehicle door opening and closing speed falling time, rail vehicle door opening and closing speed falling displacement, rail vehicle door opening and closing pressure speed time, rail vehicle door opening and closing pressure speed displacement;

[0054] Step 2: Based on the input samples obtained in step 1, the feature vector of the door opening and closing state of the rail vehicle is obtained after concept mapping and spurious transformation of the cerebellar neural network. The specific process is:

[0055] The input sample S after the switch gate is (k) Input to the door controller, after concept mapping, activate C storage units in the concept memory AC, and then after spurious coding, find the weights of the actual memory for the activated C storage units, read the weights, and output vector W i , vector W i That is the characteristic vector of each door opening and closing state;

[0056] Step 3: Calculate the Minkowski distance between the current running feature vector and the typical feature vector obtained in advance to perform sub-health pre-diagnosis and fault type clustering:

[0057] S1: Calculate the Minkowski distance between the operating feature vector of the fault-free state and the operating feature vector of the typical fault state in the vector space:

[0058]

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066] Find the minimum value of the above Minkowski distance:

[0067] l0=min{D(W1, W2), D(W1, W3),..., D(W1, W9)};

[0068] S2: Calculate the Minkowski distance between the operating feature vector of the rail vehicle door's current door opening and closing state and the operating feature vector of the fault-free state in the vector space:

[0069]

[0070] S3: Compare the shortest Minkowski distance l0 between the running feature vector of the fault-free state obtained by S1 and the running feature vector of the typical fault state in the vector space with the Minkowski distance l between the running feature vector of the current door opening and closing state of the rail vehicle door obtained by S2 and the running feature vector of the fault-free state in the vector space, and perform a sub-health pre-diagnosis of the current door opening and closing state:

[0071] when When the door machine system is in healthy operation;

[0072] when When the door machine system is in sub-healthy operation, the door machine controller will alarm and switch to normal control state;

[0073] when When the door machine system fails in operation, the fault type is clustered, the door machine controller alarms and enters the emergency processing procedure;

[0074] When the rail vehicle door has a door opening and closing fault, the door machine controller clusters the fault types. The clustering conditions are as follows:

[0075] Where W is the operating characteristic vector of the door opening and closing state, W i and W i is the operating characteristic vector of each typical fault state.

[0076] The output feature vector W of the operating state i and W iIncluding: output vector W1 for no fault, output vector W2 for V-type abnormal fault, output vector W3 for loose motor component fault, output vector W4 for small centering dimension change fault, output vector W5 for large centering dimension change fault, output vector W6 for upper slide outward movement fault, output vector W7 for lower stop pin lateral interference fault, output vector W8 for lower stop pin longitudinal interference fault, and output vector W9 for pressure wheel overpressure fault.

[0077] The C of the cerebellar neural network is the generalization width, which is taken as 6.

[0078] The hash code is a compression mapping, that is, the space of hash values ​​is much smaller than the space of inputs, and different inputs may hash to the same output. Simply put, it is a function that compresses a message of any length to a message digest of a fixed length. Here, the hash code function uses the division and remainder method. The remainder obtained after the keyword is divided by a number p that is not greater than the length of the hash table m is taken as the hash address. That is, H(key) = key MOD p, p≤m. Among them, the divisor p is called the modulus. The key to this method lies in the selection of the modulus p. Make the probability of each keyword in the data element set being mapped to any address of the memory unit through the hash function equal, thereby minimizing the possibility of hash conflicts. One experience of using the division and remainder method is that if the length of the hash table is m, p is usually the smallest prime number less than or equal to the length (preferably close to m) or a composite number that does not contain less than 20 prime factors.

[0079] The typical characteristic vectors include: an operation characteristic vector of a fault-free state, an operation characteristic vector of a V-type abnormal fault state, an operation characteristic vector of a motor component loose fault state, an operation characteristic vector of a centering dimension change small fault state, an operation characteristic vector of a centering dimension change large fault state, an operation characteristic vector of an upper slideway outward movement fault state, an operation characteristic vector of a lower stop pin lateral interference fault state, an operation characteristic vector of a lower stop pin longitudinal interference fault state, and an operation characteristic vector of a pressure wheel overpressure fault state. The characteristic vectors of these nine typical operating states are stored in the rail vehicle door machine controller in advance.

[0080] The Minkowski distance calculation formula is:

[0081] Experiments have shown that when the p value is 1.5, the Minkowski distance formula is most effective in sub-health pre-diagnosis and fault type clustering.

[0082] The present invention provides a control method with sub-health pre-diagnosis and fault type clustering functions and a door machine controller. A control method with sub-health pre-diagnosis and fault type clustering functions comprises the following steps:

[0083] Step 1: H output by the rotor position sensor in the control system with sub-health pre-diagnosis and fault type clustering function a , H b , H c Detection is performed to form a speed feedback signal ω * ;

[0084] Step 2: Output current i of the control system with sub-health pre-diagnosis and fault type clustering function A 、i B Detection is performed and compared with the torque direction signal DIR output by the absolute value circuit and the H output by the rotor position sensor. a , H b , H c The current feedback signal i is synthesized by the current feedback signal f ;

[0085] Step 3: Subtract the speed feedback signal ω from the speed given signal ω * , the speed error is obtained, and the speed error is adjusted by the proportional integral operation to obtain the signal u gi ;

[0086] Step 4: Signal u gi The absolute value circuit outputs the torque control signal MCMD and the aforementioned current feedback signal i f By comparison, the current error is obtained, and this current error is output as PWM through the proportional integral operation with anti-integral saturation.

[0087] Step 5: Output PWM and the H output by the aforementioned rotor position sensor a , H b , H c Together, the three-phase inverter is controlled through the commutation logic to drive the brushless DC motor.

[0088] A door machine controller with sub-health pre-diagnosis and fault type clustering functions uses a brushless DC motor as a driving motor of the door controller, uses a power MOS three-phase bridge circuit to feed the brushless DC motor, and uses an STM32F429 chip as its built-in controller to realize the control of the position, speed and current of the brushless DC motor. Two LEM Hall current sensors are used to detect the current, and the current value of the third phase can be calculated according to the Kirchhoff current criterion, and the high-speed A / D channel of the STM32F429 is used to complete the sampling and conversion of the motor circuit signal. The control software of the door machine controller also includes a position control module, a speed control module, a D-axis current control module, a Q-axis current control module, a rotation transformation module, an inverse rotation transformation module, a speed position calculation software module, a speed position storage module, a torque calculation software module, a torque storage module, a CMAC sample input module, a spurious code mapping module, a feature vector extraction module, a sub-health pre-diagnosis and alarm module, and the automatic door controller also includes a hardware module, which uses a Hall current detector to form a two-phase current detection module, and uses an intelligent power module IPM to form an inverter to realize the drive of the motor.

[0089] The present invention is a door machine controller with sub-health pre-diagnosis and fault type clustering functions, which includes online detection of the opening and closing actions of rail vehicle doors, automatically providing the characteristic value of the opening and closing curve to the monitoring system, pre-diagnosing whether the rail vehicle door is healthy during the closing process by calculating the output characteristic vector of the operating state, and performing fault type clustering analysis on the opening and closing operating state of the faulty rail vehicle door.

[0090] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, all of which fall within the scope of protection of the present invention.

Claims

1. A method for pre-diagnosis of sub-health of a door crane system and clustering of fault types, characterized by: The cerebellar neural network is used to realize sub-health pre-diagnosis and fault type clustering. The switch process of the door body controlled by the door machine system is segmented to form the input sample of the cerebellar neural network. The cerebellar neural network includes virtual memory and stray code, and the stray code value is used as the feature vector. The typical feature vectors include: the operation feature vector of the fault-free state, the operation feature vector of the V-type abnormal fault state, the operation feature vector of the motor component loose fault state, the operation feature vector of the centering size change small fault state, the operation feature vector of the centering size change large fault state, the operation feature vector of the upper slideway outward displacement fault state, the operation feature vector of the lower stop pin lateral interference fault state, the operation feature vector of the lower stop pin longitudinal interference fault state, and the operation feature vector of the pressure wheel overpressure fault state. The feature vectors of these 9 typical operating states are stored in the rail vehicle door machine controller in advance, and the sub-health pre-diagnosis and fault type clustering are realized according to the Minkowski distance of each feature vector in the vector space. The process includes the following steps: Step 1: The door machine controller divides the operating parameters of the current door opening and closing process into four sections: rising section, constant speed section, speed reduction section and pressure speed section, and extracts the time and displacement of these four sections to form the input samples of the cerebellar neural network; Step 2: Based on the input samples obtained in step 1, the feature vector of the door opening and closing state of the rail vehicle is obtained after concept mapping and spurious transformation of the cerebellar neural network; Step 3: Calculate the Minkowski distance between the current running feature vector and the typical feature vector obtained in advance to perform sub-health pre-diagnosis and fault type clustering.

2. The method for pre-diagnosis of sub-health of a door crane system and clustering of fault types according to claim 1 is characterized by: In the step 1, the door opening and closing process is decomposed into: the rail vehicle door opening and closing speed increase time, the rail vehicle door opening and closing speed increase displacement, the rail vehicle door opening and closing uniform speed time, the rail vehicle door opening and closing uniform speed displacement, the rail vehicle door opening and closing speed reduction time, the rail vehicle door opening and closing speed reduction displacement, the rail vehicle door opening and closing pressure speed time, and the rail vehicle door opening and closing pressure speed displacement.

3. The method for pre-diagnosis of sub-health of a door crane system and clustering of fault types according to claim 1 is characterized by: The generalization width C in the cerebellar neural network is 6, and the spurious coding function adopts the division remainder method.

4. The method for pre-diagnosis of sub-health of a door crane system and clustering of fault types according to claim 1, characterized in that: The Minkowski distance includes: The Minkowski distance between the operating feature vector of the fault-free state and the operating feature vector of the typical fault state in the vector space: The shortest Minkowski distance between the operating feature vector of the fault-free state and the operating feature vector of the typical fault state in the vector space: l0=min{D(W1, W2), D(W1, W3),..., D(W1, W9)}; The Minkowski distance between the operating feature vector of the door opening and closing state and the operating feature vector of the fault-free state in the vector space is:

5. The method for pre-diagnosis of sub-health of a door crane system and clustering of fault types according to claim 4 is characterized by: The basis for pre-diagnosis of sub-health based on Minkowski distance is: when When the door machine system is in healthy operation; when When the door machine system is in sub-healthy operation state; when , the door machine system operating status fails.

6. The method for pre-diagnosis of sub-health of a door crane system and clustering of fault types according to claim 4 is characterized by: The basis for clustering fault types according to Minkowski distance is: Where W is the operating characteristic vector of the door opening and closing state, W i and W j is the operating characteristic vector of each typical fault state.