CN107203201A - Elevator monitoring method based on CAN - Google Patents

Abstract

The invention discloses the elevator monitoring method based on CAN, including：CAN telecommunication circuit, including at least one CAN controller and CAN transceiver；Master controller, it connects the CAN telecommunication circuit, can monitor the working condition of the elevator；Signal acquisition module, it can carry out the communication function between data acquisition and realization and the CAN telecommunication circuit；Alarm module, it connects the master controller, and the working condition of the elevator can be alarmed extremely.

Description

Elevator Monitoring Method Based on CAN Bus

technical field

The invention relates to an elevator monitoring method, in particular to an elevator monitoring method based on CAN bus.

Background technique

At present, most elevator monitoring systems can only target a few specific elevator models, and are not compatible with early elevators with a high failure rate, resulting in only a part of the coverage.

The interfaces of the existing domestic elevator monitoring systems are generally relatively simple, and many of them are made directly by using configuration software, and the configuration software is not suitable for applications with good user interface.

At present, most elevator monitoring systems in China do not have a database, but only provide basic real-time monitoring functions and alarm functions. Such a system has poor scalability, and it is difficult to perform higher-level monitoring of elevator conditions on it. According to the analysis, even if the elevator monitoring system has a database, it has only developed some simple functions, such as historical record query, fault distribution, etc.

Contents of the invention

The present invention designs and develops the elevator monitoring method based on the CAN bus. One of the invention objectives of the present invention is to use the field bus as a communication carrier to realize the elevator monitoring system, reduce transmission errors, and store and monitor data in real time.

The second invention purpose of the present invention is to adopt the BP neural network algorithm to complete the active alarm of the elevator monitoring system, so as to reduce the accidents caused by the elevator.

The technical scheme provided by the invention is:

The elevator monitoring method based on CAN bus, including:

CAN bus communication circuit, including at least one CAN controller and CAN transceiver;

a main controller, which is connected to the CAN bus communication circuit and can monitor the working state of the elevator;

A signal acquisition module, which can perform data acquisition and realize the communication function with the CAN bus communication circuit;

An alarm module, which is connected to the main controller, can alarm the abnormal working state of the elevator.

Preferably, a dial switch is set in the signal acquisition module, and the signal acquisition module connects the data and the data collected by the signal acquisition module with the CAN bus communication circuit after reading the dial switch. communication function between them.

Preferably, it also includes: a background function module, which is used to monitor the transmission data of the CAN bus communication circuit.

Preferably, the background function module includes:

An active alarm module, which is used to judge whether the elevator has a failure risk;

Main interface module, which is used to refresh the working status of the elevator;

The alarm monitoring module is used for judging whether the elevator works abnormally.

Preferably, a timer is set in the signal acquisition module.

Preferably, the CAN transceiver is a CTM1050T chip.

Preferably, the alarm module is a GSM alarm module.

Preferably, it is characterized in that, adopting BP neural network to carry out early warning to elevator, comprises the steps:

Step 1. According to the sampling period, the collected data includes elevator load, car door opening and closing noise, three currents of the traction machine, three voltages of the traction machine, vertical vibration acceleration, horizontal vibration acceleration, vertical acceleration, vertical deceleration, Elevator host motor temperature, emergency brake coil temperature and machine room temperature;

Step 2: Process the above signals, the elevator load, car door opening and closing noise, three currents of the traction machine, three voltages of the traction machine, vertical vibration acceleration, horizontal vibration acceleration, vertical acceleration and vertical deceleration Carry out data processing by S-type membership function, described elevator host motor temperature, emergency brake coil temperature and machine room temperature carry out data processing by Gaussian-type membership function, determine the input layer node of three-layer BP neural network;

Step 3, the input layer vector is mapped to the hidden layer, the hidden layer vector Y={Y ₁ , Y ₂ ,...,Y _h }; h is the number of nodes in the hidden layer;

Step 4: Get the output layer node T={T ₁ , T ₂ , T ₃ , T ₄ , T ₅ , T ₆ }; among them, T ₁ is the failure coefficient of the elevator traction machine, T ₂ is the circuit failure coefficient, and T ₃ is the failure coefficient of the door machine, T ₄ is the failure coefficient of the elevator controller, T ₅ is the failure coefficient of the three-phase power input unit, and T ₆ is the failure coefficient of the inverter.

Preferably, in said step 3, the number h of said hidden layer nodes satisfies, Among them, n' _i is the number of nodes in the input layer, n' ₀ is the number of nodes in the output layer, and c is a number between 1 and 10.

Preferably, the activation function of the hidden layer is a hyperbolic tangent function, and the activation function of the output layer is a nonlinear logarithmic Sigmoid function.

Compared with the prior art, the present invention has the beneficial effects:

1. The field bus of the present invention realizes the elevator monitoring system managed by the computer as a communication carrier, and different products or products of different manufacturers can be connected together through the field bus. The field bus uses digital signals for transmission, generally with arbitration and error detection, etc. Ability to effectively reduce transmission errors;

2. The present invention stores and monitors the signals collected by the elevator signal acquisition system, completes the real-time monitoring of the elevator, can efficiently complete the fault alarm function, and complete the historical data analysis function;

3. The present invention adopts the data mining algorithm of BP neural network to complete the active alarm function of the elevator monitoring system, and uses data mining to analyze potential development trends and hidden dangers of certain things, which can greatly reduce accidents caused by elevators.

Description of drawings

Fig. 1 is a frame diagram of an elevator monitoring system according to the present invention.

Fig. 2 is a bottom signal diagram of the elevator monitoring system according to the present invention.

Fig. 3 is a background functional module diagram of the elevator monitoring system according to the present invention.

Fig. 4 is a CAN bus network topology diagram according to the present invention.

Fig. 5 is a schematic diagram of a signal screenshot of the elevator monitoring system according to the present invention.

Fig. 6 is a block diagram of the signal acquirer module of the present invention.

FIG. 7 is a partial circuit diagram of the CAN bus according to the present invention.

Fig. 8 is an interface diagram of the present invention.

Fig. 9 is a parameter interface diagram of the present invention.

Fig. 10 is a structural diagram of a three-layer neural network according to the present invention.

Fig. 11 is a flow chart of the BP neural network propagation algorithm of the present invention.

detailed description

The present invention will be further described in detail below in conjunction with the accompanying drawings, so that those skilled in the art can implement it with reference to the description.

As shown in Figure 1, the frame diagram of the elevator monitoring system. The present invention adopts the field bus as the communication carrier to realize the elevator monitoring system managed by the computer, the elevator is the main body to be monitored, the control center runs in the computer, and the operating condition of the elevator is monitored with the help of the computing power of the computer. When the elevator is judged to be abnormal, it can Notify the management personnel or elevator maintenance personnel through certain means to deal with the abnormality. After the elevator signal acquirer obtains the information, it needs to be packaged into a data packet and sent to the CAN bus network through the CAN bus transceiver module. On the computer, there is a CAN bus module connected to the CAN bus network, which will collect data packets and send them to the computer through the serial port.

As shown in Figure 2, the signal acquirer at the bottom of the system is composed. The underlying service of the elevator monitoring system refers to all services on the CAN network and below. It provides the operating data of the elevator for the upper layer of the system. The main implementation part is the signal acquirer. The elevator monitoring system designed in the present invention is mainly aimed at old elevators, because their models are different and the functions they support are also different. Input and output The design of the port is also different. In order to solve this problem, this design adds a dial switch that can be used to configure the elevator model on the signal acquirer. The signal acquirer can read the dial switch and then communicate with the collected data. Upload together, and the upper computer background service will determine the meaning of each pin.

As shown in Figure 3, the block diagram of the background function module, the entrance of the background function module is a serial port monitor, which is responsible for monitoring the data uploaded from the CAN bus network. After the data is uploaded, the configuration table manager will read the configuration from and give specific meaning to the uploaded data. After the data sorting module finishes sorting, it will be stored in the database by the data storage module.

The data sorting module will also deliver the data to the data distribution module, and the data distribution module will deliver the data to the modules that need the data. There are three in total, namely the active alarm module, the main interface and the alarm monitoring. The background of the active alarm module will judge whether the elevator has a risk of failure according to the historical data of the elevator; the main interface is used to refresh the working status of the elevator; the alarm monitoring module will judge whether the elevator is working abnormally according to the parameters of the alarm manager, if abnormal, then according to The contact information provided by the alarm manager generates alarm information, and calls the GSM driver to issue an alarm. GSM is used to drive a GSM alarm. The alarm supports TTS technology, which can convert text into voice and play it on the phone.

The main interface can create three background sub-modules, which are the user authority manager, which is used to handle the background logic of user management; the data query result generator, which is used to provide data support for data query; the alarm manager, which is used to manage users The thresholds set and the contact information of the management personnel.

As shown in Figure 4, the multi-CAN bus network topology structure, a computer can accept data from multiple CAN bus networks, and each CAN bus network is a star network. This network topology can improve the CAN bus network to a certain extent. The stability of the network and the avoidance of network congestion.

As shown in Figure 5, the signal interception diagram of the elevator monitoring system, the elevator monitoring system of the present invention is based on the system above the existing elevator system, so first abstract the operating mode of most elevators, and on this basis, find A suitable insertion point is to insert the elevator monitoring system into the normal elevator system to monitor the operation of the elevator.

As shown in Figure 6, the structure diagram of the signal acquirer module, the function of the signal acquirer is to collect the signal input from the outside world, the collected signal may be an analog signal or a digital signal, and the signal acquirer performs different functions for different input ports. processing. Use a DIP switch to determine the configuration, and then send the collected information and configuration number to the signal converter.

There is a timer inside the signal collector, which is used to scan each input port in a regular cycle. After scanning once, it will be timed again and wait for the next scan. In addition, there is a DIP switch manager. The main function of this manager is to determine what type of information the collected information is. There is a table maintained in the system, which divides the configuration items used for different configuration items. , scan the information and add configuration items to get the specific meaning of the information.

The only function of a data converter is to complete the operation of converting information into standard format data packets. The format design of the data packet is shown in Table 1;

Table 1 Format design of data packet

The configuration number is used to inform the server of the configuration of the signal acquirer, and the data description is used to save the occupied space of the data collected by the A/D conversion channel.

The CAN bus interface module is a module that provides external communication for the microcontroller. This module is only responsible for initializing the CAN bus controller, receiving data packets sent by the data converter, and sending them to the CAN bus node whose address is 0x00. CAN bus interface module is composed of two parts: the first part is: CAN bus maintenance. This part is responsible for the maintenance of the CAN bus, and initializes the CAN bus controller at the beginning of the startup, including the setting of each register and the filling of the sending buffer. When required to send data, fill in the relevant information according to the pre-designed template and complete the sending. The second part is: CAN bus data splitter. Since the data frame used by the CAN bus used in this design can transmit an effective byte size of 8Byte per frame, when processing a large data packet, the data packet needs to be disassembled and sent, and the data splitter will divide the data packet Split into several data frames according to the size, and then number each data frame, send the numbered data packet to the CAN bus maintenance module and send it out.

In the present invention, the CTM1050T chip used by the CAN bus transceiver integrates all CAN bus transceiver devices in the chip, and the total size does not exceed 3cm ² , which conforms to the ISO 11898 standard. The chip model used by the CAN bus controller is Yes SJA1000T. The PeliCAN mode is used in the present invention. In this mode, SJA1000T can support all CAN2.0B specifications.

As shown in Figure 7, part of the CAN bus circuit diagram, CTM1050T is controlled by SJA1000T, only use two pins RX and TX, respectively connected to TX0 and RX0 of SJA1000T, and need to connect a resistance between CANH and CANL 120 ohm terminal resistor, but only one resistor is needed in the entire CAN bus network, not all nodes need a terminal resistor, so the terminal resistor will be directly connected to the bus. SJA100T0 is connected to the single chip microcomputer, in which AD0 to AD7 are used as data lines, ALE/AS, CS, RD, WR are used as control lines, in addition, SJA needs to connect a crystal oscillator as a signal source, and provides an interrupt port for convenience SCM controls the working state of SJA1000T.

The single-chip microcomputer adopts STC15F2K60S2 in the STC single-chip series. This single-chip microcomputer supports eight A/D channels, and the maximum operating frequency can reach 33MHz. It supports power-down wake-up and hardware watchdog. In the hardware circuit, the role of the single chip microcomputer is signal acquisition and control of the CAN bus controller. In the circuit, P1 is the A/D conversion channel. Since the voltage to be detected is higher than 5V, the resistive row is used to divide the voltage, which is converted into a voltage lower than 5V. 5V analog voltage, the resistance is too small. It varies according to the information to be obtained by the signal acquirer. When obtaining the input and output signals of the elevator, use 5K. When obtaining high voltage, you need to use about 100k resistance division. . CD4067 is used for digital switching value. CD4067 is a 16-channel analog switch. Although there is interference, it does not affect the digital value. Resistors are also used for voltage division at the output port. Since a single single-chip microcomputer cannot monitor all signals of an elevator, it is necessary to add a dial switch to determine the configuration. Different configurations indicate different content to be collected. The network AD is connected to the CAN bus network part as a data bus, and the others are used to control the CAN bus network part.

The signal acquirer in the present invention uses a dual power supply switching circuit. In the case of a 12V power supply, the 12V power supply is used, and the battery power supply is cut off at the same time. When the 12V power supply is missing, the power supply can be switched to the backup battery for power supply. The design is to be able to transmit a power loss signal to the computer in case of power failure, and notify the management personnel to check.

The software design of the signal acquirer in the present invention is mainly based on the single-chip microcomputer STC15F2K60S2, and the software should realize the input information of each input port of timing cycle scanning, determine the significance of the input port information according to the dial switch, generate standard data packets, and can operate the CAN bus control device, including initializing the CAN bus controller and sending data. The software design of the signal acquirer is divided into four parts, which are CAN bus transceiver, timing scan, standard data packet generation and hardware watchdog.

The CAN bus controller of this design uses SJA1000T, so the main work content of the transceiver part of CAN bus is to operate SJA1000T. The operation of SJA1000T is essentially the operation of its various registers. The CAN bus driver code refers to the SJA1000T register address defined in the defined SJA1000T header file and a function named SJA1000T_Write().

The timing scanning procedure of this design is divided into two parts. First is the timed loop section, followed by the scan section. The timing cycle part starts to run after the system is started. The timing scan program uses TIMER0 as a timer, the timing time is 20ms, and the acquisition frequency is 50HZ, which is the same frequency as the mains 220v. The program first modifies the TH0 register and the TL0 register, sets the loop count, then opens the total interrupt, the program exits, and after the timer expires, it will enter the interrupt. A global variable will be set in the interrupt function, and it will be set to True. , and then clear the interrupt flag, and enter the loop here. The scanning part is the core of the regular scanning program. The code entry mark of this part is set in the interrupt function of interrupt 1, and the code is located in the main thread. When the scan part detects that the scan flag is allowed to be True, start the scan, and finally change the scan flag back to False. The scan function first scans the eight A/D conversion channels one by one, and calculates the original voltage according to the resistance value of the voltage divider circuit, and finally forms an array for storage. Then control the CD4067 to scan the digital quantities one by one and store them in an array, and finally send the configuration and the array to the data packet generator to complete the entire scanning process.

The data packet generator of this design is responsible for reading the position of the DIP switch in the design, and then determining the configuration. It is known from the hardware design that the signal acquirer uses two I/O ports and can only use four configurations at most. In the memory of the single-chip microcomputer, four tables are stored, indicating four different configurations. The regular scanning program will send the array generated by scanning to the data packet generator, and the data packet generator needs to generate a data packet in a standard format from the array and send it to the CAN bus module. The data description in the data packet describes the data of the A/D conversion channel. It uses one byte for description. The input of the current channel is 1, and the input of no input is 0. The A/D scan array plus the digital signal scan array of two bytes together form the data content.

The system stability maintenance program design of the present invention consists of:

1. Use the single-chip microcomputer to support the hardware watchdog function. The function of the watchdog is that after the system starts the watchdog, the watchdog starts timing. Within the limited time, that is, before the watchdog timer overflows, the system must execute The "feed the dog" operation is to clear the watchdog timer and let it start counting again. If the system still does not "feed the dog" when the time expires, the watchdog will think that the system program has run away and will directly reset the MCU to let the system Start working again to prevent the system from falling into an infinite loop or program confusion. In this design, considering the reason of CAN bus transmission and reception, the watchdog overflow time is about 1s. After each system scans and sends data through the CAN bus, the code control right will be returned to the main loop program, which is controlled by the main loop program. Do the work of "feeding the dog" and avoid spills. If there is an error in the CAN bus node, the system will fall into the CAN bus sending code, that is, it will trigger the watchdog to perform a reset operation and reset the entire signal collector;

2. Use dual power supplies for power supply. If the main power supply fails, the battery power supply will be activated. When the battery power supply is enabled, the system will restart. After startup, the system will start collecting data according to the established procedures, but because the main power supply has been cut off power, so the collected signals are all 0, in this case, the system will send a power-down alarm through the CAN bus. After the monitoring software in the calculation receives this data packet, it can directly trigger the elevator power-off alarm according to the address in the data packet. The power-down error detection code is located in the main loop of the system like the watchdog, written after the data packet generator, and before sending to the CAN bus transceiver program.

The main reason for choosing MySQL in this design is that the elevator monitoring system is a real-time system, and the data storage speed of MySQL is relatively fast, so it is suitable for use as the database of the elevator monitoring system. Three parts of data tables are designed in the database, which are used for user management, elevator management, and alarm parameter management respectively. User management only uses one user management table, as shown in Table 2;

Table 2 User Management Table

serial number name type of data meaning 1 UserName text username 2 Password text User password (MD5 encryption) 3 Auth integer User rights 4 Last Login datetime Last Login Time

Elevator management is a combination of tables used to manage elevators, including the elevator list, configuration table, and elevator data table. The elevator management table is shown in Table 3;

Table 3 Elevator Management Table

serial number name type of data meaning 1 LiftID integer elevator master key 2 Desc text Description of the elevator 3 floor integer Number of elevator floors 4 position text elevator location

The configuration table describes the configuration of the elevator. An elevator may have multiple configurations, because there may be multiple signal acquirers attached to an elevator. It is impossible for different signal acquirers on an elevator to use one configuration. The configuration table is shown in Table 4 shown;

Table 4 configuration table

The elevator data table is used to store the data uploaded by the signal acquirer, as shown in Table 5;

Table 5 Elevator Data Sheet

serial number name type of data meaning 1 DataID integer Data table primary key 2 SettingID integer Foreign key, pointing to configuration table 3 CollectTime DataTime record collection date 4 LiftID integer foreign key, pointing to elevator table 5 AD0 integer Analog signal channel 0 data 6 ... ... ... 7 AD7 integer Analog signal channel 0 data 8 Dig0 integer Digital channel 0 data 9 ... ... ... 10 Dig15 integer Digital channel 1 data

The alarm management table is used to store alarm configuration information. Two data tables are used, namely the alarm contact table and the alarm parameter table. The alarm contact table is shown in Table 6;

Table 6 Alarm Contact Form

The alarm parameter is used to allow the user to set the parameter alarm threshold of a certain elevator. When the data uploaded by the elevator is not within the range of the alarm parameter, an alarm will be triggered. The parameter can be set directly by the user. The alarm parameter table is shown in Table 7;

Table 7 Alarm parameter table

serial number name type of data meaning 1 AlertSettingID integer Alarm parameter table primary key 2 LiftID integer Foreign key, pointing to the elevator number 3 SettingID text Foreign key, pointing to configuration table 4 AD0 text Analog signal channel 0 data range 5 ... ... ... 6 AD7 text Analog signal channel 7 data range 7 Dig0 text Digital channel 0 data range 8 ... ... ... 9 Dig15 text Digital channel 15 data range

In this design, the data management service module is used to receive the data uploaded by the signal acquirer, and read the configuration table in the database to give the data meaning, then store it in the elevator data table in the database, and finally throw the data to the required module. According to its function, the data management service module needs a serial port driver module and a data processing module. Among them, the serial port driver module is responsible for completing the access to the CAN bus network, writing a specific buffer for use by the upper program; the data processing module is divided into two parts, the first part completes the meaning of the data, and generates a data instance class, And store the data, the second part is to hand over the data to the modules that need to be used. The two parts can be considered to exist independently, and the serial port driver class is defined as follows:

Among them, ComDrive is the constructor, which needs to be passed in to the serial port number used and the transmission speed. By default, the data bit is 8 bits, the stop bit is 1, and there is no checksum method. The owner of the class can use the OpenCom and CloseCom functions to operate the switch of the serial port. When uploading data, the ReceiveData function will be triggered. This function judges the validity of the data. If the verification is passed, it will check whether the event ReceiveDataHandle has been set. If it has been set, the event registrant will be notified through the event. If it has not been called SetReceiveDataHandle, the data will be discarded.

After the data passes through ComDrive, the data management service will generate a data instance class according to the configuration ID and data in the data.

In this design. The main interface program is divided into foreground interface and background function. The foreground interface is used to present the elevator running status, parameters and entry points for various functions, and the background function is responsible for providing support for the foreground interface.

As shown in Figure 8, the front desk interface is divided into two areas, which are the menu function area and the elevator operation status indication area. Double-click the elevator to display the detailed operation parameters of the elevator. The upper menu bar is the entrance of various functions, including system settings for completing some settings, historical record query for querying historical elevator operation data, user management for managing user information, and alarm settings for setting alarm-related information. There are two buttons below, which can be used to turn pages when there are too many elevators. For elevators with more than 9 floors, two more control buttons will be added to view different floors. If you click on the elevator, in the case of a high floor, the location of the current elevator will be located, and there will be no response on the low floor. If you double-click the elevator, you will enter the detailed parameter interface of the elevator.

As shown in Figure 9, on the detailed parameter interface, the detailed parameters of the selected elevator will be given, generally including call status, running status, door status, limit switch status, inverter status, etc., if the elevator has some unique The parameters can also be displayed, but the interface can automatically generate display items only after pre-configuration in the database.

The background function of the main interface is mainly responsible for refreshing the interface. After the initialization of the interface, it registers the system's data management service event notification. When the elevator running status changes, the natural background function will be notified, and then refresh the interface to achieve for real-time display purposes. In addition, the background function is also responsible for the action processing of each button and control. The main interface is a loop. Whether there is data upload depends on whether the data management service module is notified. When there is no data notification, the main interface will exchange control rights to Operating system, when the data management service module notifies that there is a new data upload, the main interface takes back the control right and continues to execute. According to the content in the obtained data instance, the status of the elevator is updated and the interface is modified, and finally the control right is restored Hand over to the operating system to allow the user to operate.

When the user selects user management on the main interface, the following three buttons are provided, corresponding to different user management module parts, namely adding users, deleting users, and modifying user passwords. User management is also divided into front-end interface and background implementation. The interface includes three interfaces, each function provides an interface. The background implementation is data checking, password encryption and operations on the database.

Adding a user will provide an interface that allows the user to add a user. The user needs to enter a user name and a password. After clicking the OK button, if the check passes, a user will be added to the database. If not, the reason for the failure will be prompted. In addition to adding users, there is also the function of deleting users. To delete users, refer to the account management function of Windows. You can only delete other users with lower permissions than your own, but you cannot delete your own account. The delete user interface provides a list. After selecting all accounts that can be deleted, click Delete to delete the selected account. The modify user password interface provides a function that allows users to modify their own passwords. The difference from adding users is that in modifying user passwords, the user name is If it cannot be modified, it is displayed in the interface as an unmodifiable state. The background function of the user management module is to connect to the database and provide user information encryption, which is implemented by a user management class.

The function of the design information checking module is to check whether the running condition of the elevator is normal. The signal of the elevator is mostly digital, which cannot be solved by setting a threshold. Therefore, it is necessary to judge whether the elevator is working normally by the information inspection module. The inspection content includes abnormal door switch, abnormal floor stop, abnormal operation, and voltage and current values. When the data is sent from the data management service module to the information inspection module, the information inspection module will go through the above four situations, and as long as an exception occurs, an abnormal message will be generated and an alarm will be issued.

In this design, the GSM alarm module provides alarm support. The user can set the alarm number through the setting interface. When the information checking module throws an alarm test, the GSM alarm module will alarm according to the alarm number. It can also directly generate an alarm through an external call. The GSM alarm uses an alarm named FB200-B. This alarm uses the RS232 communication protocol and can communicate directly with the computer. It only needs to use the ATL command to send TTS voice and the ATM command to send SMS.

The present invention can inquire about the historical records of any elevator, mainly the historical alarm records can be viewed, and two types of graphs can be viewed: graphs and lists.

The present invention adopts the database, data warehouse and server provided by SQL Server, and the data mining engine and above needs to be designed and realized by oneself. There is no visual user interface in this design, and the background output is used directly.

The present invention adopts the data mining algorithm of the neural network. First, the data cleaning work is carried out. The elevator signal acquirer is constantly uploading the data. In order to reduce the burden on the model, the data must be cleaned. This process is performed before the data is put into the data warehouse. To proceed, first check the difference between the data collected this time and the data collected last time. If there is no difference, it will be discarded directly, and if there is a difference, it will be put into the data warehouse. The second is the preprocessing of data. For the elevator monitoring system designed in this paper, what needs to be completed is the scale transformation and numerical data conversion. Since the receiving range of the neural network model is between [-1.0, 0] and [0.0, 1.0], and the range of the elevator monitoring data is far beyond this limit, it is necessary to perform a scale transformation. The neural network model is used in the elevator monitoring system. What is predicted is voltage and current, the voltage range is 0-311V, and the current range varies according to different circuits, all of which need to be mapped to the acceptable range of the neural network model. In addition, data conversion is also required. The elevator monitoring system uses a configuration table to describe the meaning of elevator data, and the neural network does not accept non-digital data processing, so it is necessary to expand the data in the database and create a hash Function to generate a unique identification number for each collected data.

The neural network algorithm that the present invention adopts comprises the steps:

Step 1: Establish a BP god general network model.

A BP network is generally composed of an input layer, a hidden layer, and an output layer. There can be multiple hidden layers. Different layers are connected in a one-to-many interconnection mode. There is no connection between nodes in the same layer. This design uses The three-layer BP neural network, the three-layer network depth can already solve the general function fitting problem. The backpropagation algorithm adopts the gradient method to modify the weight value, and requires the output function to be differentiable, and uses the Sigmoid function as the output function. Figure 10 is the structure diagram of the three-layer BP neural network.

X _i represents the signal of a certain node in the input layer; Y _h represents the output signal of a hidden node; T _j represents the target signal; _{ω ih} represents the connection weight of the hidden layer node h of the input layer node i; The connection weight of the output layer node j containing the layer node h; Δ represents the error between the target signal and the output signal; N ₁ represents the number of input layer nodes; N ₂ represents the number of hidden layer nodes; N ₃ represents the number of output layer nodes .

The output of the hidden layer node is: y _h = f(net _h );

The output of the output layer node is: z _j = f(net _j );

The error function is:

Define the local gradient as:

Consider the influence of the weight ω _ih on the error:

The weight correction should reduce the error as soon as possible, and the correction amount is: ω _ih (t+1) = ω _ih (t) + Δω _ih (t);

For the sigmoid function:

In order to speed up the convergence speed of the network, add the previous weight correction amount to the weight correction amount, Δω _ih (t)=-η·δ _h ·y _h +α·Δω _ih (t-1)

Among them, θ _h is the offset between the input layer and the hidden layer; Υ _j is the offset between the hidden layer and the output layer; η is the learning step; α is the coefficient of the inertia term; Figure 11 is the BP neural network Propagation algorithm flow chart; for the number of network nodes, there is currently no clear theoretical relationship between the number of nodes and the results. Through the experimental results of training samples and test samples, this paper finally uses the following formula to determine the number of nodes in the hidden layer: Among them, n' is the number of nodes in the hidden layer, n' _i is the number of nodes in the input layer, n' ₀ is the number of nodes in the output layer, and c is a number between 1 and 10.

In the present invention, there are 11 input signals, including:

(1) Elevator load, (2) Car door opening and closing noise, (3) Traction machine three currents, (4) Traction machine three voltages, (5) Vertical vibration acceleration, (6) Horizontal vibration acceleration, ( 7) Vertical acceleration, (8) Vertical deceleration, (9) Main engine motor temperature, (10) Emergency brake coil temperature, (11) Machine room temperature; the above signals are used as the input vector of the neural network, where (1)-( 8) and other 8 signals adopt S-type membership function; 3 signals such as (9), (10), and (11) adopt Gaussian membership function to form three languages of "low", "normal" and "high". variable; therefore, the number of input layer nodes is 3×3+8=17;

By analyzing the causes of elevator failures and consulting related literature, we can get six common output failure signals of elevators: (1) Elevator traction machine failure, (2) Circuit failure, (3) Door machine failure, (4) Elevator controller failure , (5) Three-phase power input unit failure, (6) Frequency converter failure;

From the formula mentioned above:

It can be seen that the number of hidden layer nodes is between 6 and 20, and the system is optimized when the number of hidden layer layers is 20, so the number of hidden layer nodes is 20; therefore, the network layer is 3 layers. The number of nodes in the input layer is 17, and the fault information of 11 elevators in this layer is used as the input signal; the number of nodes in the output layer is 6, and this layer represents that the fault information obtained from the input layer has 6 outputs after the fault processing after adjusting the weight coefficient Fault signal; the number of nodes in the hidden layer is 20, and the BP neural network algorithm is used in this layer to optimize the training of samples. Before training, the initial offset and weight take a number between (-5,5), and optimize these ten numbers to obtain an optimized solution, that is, the offset is fixed at 0.5, and then network training is performed. In this design, in order to improve the accuracy of the input vector, reduce the error, and limit the range of the output vector to [-1, 1], the activation function of the hidden layer is selected as a hyperbolic tangent function, and the activation function of the output layer is selected as a nonlinear logarithm Sigmoid function; as shown in Table 8, a set of training samples and the values of each node in the training process are given.

Table 8 Training samples, the value of each node in the training process

In order to improve the learning speed of the algorithm, the trainlm function corresponding to the Levenberg Marquardt algorithm is used for measurement and calculation. The weight of the algorithm is corrected according to dX=-(jX ^T *jX+I*mu) ^-1 *E, where jX is the error pair The Jacobian matrix of weight differential, E is the error vector, and mu is the adjustment amount. The training error is 10 ^-4 .

Assuming that there are P pairs of training samples, the training error of the p-th pair of training samples: Use the mean square error as the total output error:

Among them, n represents the number of nodes in the output layer, Indicates the actual output of the tth node of the pth group sample output layer, p=1,2,3,...,p, t=1,2,3,...,n, Represents the expected output of the tth node of the pth group sample output layer, p=1,2,3,...,p, t=1,2,3,...,n, (E ^f ) ² represents all sample output errors; As shown in Table 9, it is a training sample table;

Table 9 training sample table

In order to improve the accuracy of training, the data is preprocessed during data training, as follows:

data preprocessing

After data processing, trigger with training function traindm. If the number of training times exceeds epochs, the performance function is lower than the goal, the gradient value is lower than mingrad, or the training time exceeds time and the training will end. The program code is as follows:

The training results show that the algorithm trains the network very quickly, the number of iterations is significantly reduced, and the convergence speed is accelerated. Through sample training and simulation, the data mining algorithm of BP neural network is used to complete the active alarm function of the elevator monitoring system. Analyzing the potential development trends and hidden dangers of certain things has a significant effect in reducing accidents caused by elevators; after processing, the data is shown in Table 3.

Table 3 Training samples and the values of each node during the training process

Although the embodiment of the present invention has been disclosed as above, it is not limited to the use listed in the specification and implementation, it can be applied to various fields suitable for the present invention, and it can be easily understood by those skilled in the art Therefore, the invention is not limited to the specific details and examples shown and described herein without departing from the general concept defined by the claims and their equivalents.

Claims (10)

1. the elevator monitoring method based on CAN, it is characterised in that including：

CAN telecommunication circuit, including at least one CAN controller and CAN transceiver；

Master controller, it connects the CAN telecommunication circuit, can monitor the working condition of the elevator；

Signal acquisition module, it can carry out the communication function between data acquisition and realization and the CAN telecommunication circuit；

Alarm module, it connects the master controller, and the working condition of the elevator can be alarmed extremely.

2. the elevator monitoring method as claimed in claim 1 based on CAN, it is characterised in that the signal acquisition module Middle setting toggle switch, the signal acquisition module by being read out to the toggle switch after data and the signal are adopted Collect the communication function between the data and the CAN telecommunication circuit of module collection.

3. the elevator monitoring method as claimed in claim 1 or 2 based on CAN, it is characterised in that also include：Backstage work( Energy module, it is used for the transmission data for monitoring the CAN telecommunication circuit.

4. the elevator monitoring method as claimed in claim 3 based on CAN, it is characterised in that the background function module Including：

Active fire alarm module, it is used to judge whether elevator has failure risk；

Main interface module, it is used for the working condition for refreshing elevator；

Alarm monitoring module, it is used to judge whether elevator work is abnormal.

5. the elevator monitoring method as claimed in claim 4 based on CAN, it is characterised in that the signal acquisition module Middle setting timer.

6. the elevator monitoring method based on CAN as claimed in claim 5, it is characterised in that the CAN transceiver is CTM1050T chips.

7. the elevator monitoring method as claimed in claim 6 based on CAN, it is characterised in that the alarm module is GSM Alarm module.

8. the elevator monitoring method based on CAN as any one of claim 1,2,4-7, it is characterised in that adopt Early warning is carried out to elevator with BP neural network, comprised the following steps：

Step 1: according to the sampling period, gathered data includes elevator loading, car switch door noise, three electric currents of traction machine, draged Draw three voltages of machine, vertical vibration acceleration, horizontal vibration acceleration, vertical direction acceleration, vertical direction deceleration, elevator Main frame motor temperature, accident brake coil temperature and computer room temperature；

Step 2: above-mentioned signal is handled, the elevator loading, car switch door noise, traction machine three electric currents, tractions Three voltages of machine, vertical vibration acceleration, horizontal vibration acceleration, vertical direction acceleration and vertical direction deceleration pass through S π membership functions carry out data processing, the elevator main motor temperature, accident brake coil temperature and computer room temperature Data processing is carried out by Gauss π membership function, the input layer of three layers of BP neural network is determined；

Step 3: the input layer DUAL PROBLEMS OF VECTOR MAPPING is to hidden layer, the hidden layer vector Y={ Y₁,Y₂,…,Y_h}；H is hidden layer Node number；

Step 4: obtaining output node layer T={ T₁,T₂,T₃,T₄,T₅,T₆}；Wherein, T₁For elevator traction machine failure coefficient, T₂For Loop fault coefficient, T₃For door machine failure coefficient, T₄For electric life controller failure coefficient, T₅For three phase mains input block failure Coefficient, T₆For frequency converter failure coefficient.

9. the elevator monitoring method as claimed in claim 8 based on CAN, it is characterised in that in the step 3, institute Hidden layer node number h satisfactions are stated,Wherein, n '_iFor input layer number, n '₀For output layer nodes, C is the number between 1~10.

10. the elevator monitoring method as claimed in claim 9 based on CAN, it is characterised in that the hidden layer activates letter Number is hyperbolic tangent function, and the output layer activation primitive is non-linear logistic S type functions.