CN110135241B

CN110135241B - Statistical analysis system of elevator stroke

Info

Publication number: CN110135241B
Application number: CN201910237511.8A
Authority: CN
Inventors: 陈清梁; 王伟; 陈国特; 王超
Original assignee: Zhejiang Xinzailing Technology Co ltd
Current assignee: Zhejiang Xinzailing Technology Co ltd
Priority date: 2019-03-27
Filing date: 2019-03-27
Publication date: 2021-08-31
Anticipated expiration: 2039-03-27
Also published as: CN110135241A

Abstract

The invention discloses a statistical analysis system of elevator travel, which comprises a data acquisition module, a door state analysis module, a floor analysis module, an LSTM travel state analysis module, a travel analysis module and a cloud storage module, wherein the door state analysis module is used for analyzing the door state of a passenger elevator; the invention provides a statistical analysis system of elevator travel, which uses a recurrent neural network to realize the state estimation of the elevator travel, realizes the analysis of the running state of the elevator and provides the data of the elevator riding rule for other subsequent services.

Description

Statistical analysis system of elevator stroke

Technical Field

The invention relates to the field of elevators, in particular to a statistical analysis system for elevator travel.

Background

With the steady and continuous development of the economy of China, particularly the proposal of the residential industry as a new growth point of the economy of China, good opportunities are provided for the development of the elevator industry; in the next few years, China will build 3.5 hundred million square meters of houses in the year, and 1.2 hundred million square meters of public building projects; with the development of large-scale and high-rise cities, the elevator market demand of China is more and more large every year, and the elevator management is more and more intelligent and informationized.

The elevator is required to be controlled in the whole process in the running process and detected in real time, but a general elevator only has a fault detection system and no behavior analysis system, so that the fault analysis is difficult to be carried out at the first time, and the maintenance efficiency of the elevator in the fault process is seriously reduced. Therefore, a statistical analysis system for elevator trips is provided to solve the above problems.

Disclosure of Invention

The invention overcomes the defects of the prior art and provides a statistical analysis system for elevator travel.

In order to solve the technical problems, the technical scheme of the invention is as follows:

a statistical analysis system for elevator travel comprises a data acquisition module, a door state analysis module, a floor analysis module, an LSTM travel state analysis module, a travel analysis module and a cloud storage module;

a data acquisition module: the method comprises the steps that image data are collected through a camera on the top of an elevator car, and data of a gyroscope, an accelerometer, a barometer and a photoelectric tube are collected through a sensor;

a door state analysis module: using image data acquired by a data acquisition module to obtain the real-time door state of the elevator by utilizing image analysis;

floor analysis module: analyzing the data of the gyroscope, the accelerometer, the barometer and the photoelectric tube acquired by the data acquisition module by using Kalman filtering and complementary filtering to obtain the real-time floor number of the elevator;

LSTM journey state analysis module: preprocessing data of a gyroscope, an accelerometer and a barometer, which are acquired by a door state data and data acquisition module, removing noise data, sampling to a data frequency of 25HZ, arranging into a model input vector, sending into an LSTM model, and outputting a stroke state by the LSTM model at intervals;

a journey analysis module: the frequency of floor change data sampling is synchronous with the frequency output by the LSTM stroke state analysis module, the start in the stroke state represents the start of the stroke, the pause represents the transfer of the stroke, and the end represents the end of the stroke; by determining the starting and ending states, the elevator journey, i.e. the floor change, can be analyzed, and the elevator journey without pause state during the starting and ending periods is regarded as an independent journey, and the elevator journey with at least one pause state is regarded as a composite journey;

cloud storage module: and storing the data of the independent elevator journey and the compound elevator journey obtained by analyzing the journey analysis module according to the elevator serial number and the time stamp, and providing elevator riding rule data for other subsequent services.

Further, the total number of the travel states is 5, namely idle, start, running, pause and end.

Compared with the prior art, the invention has the advantages that:

according to the invention, the state estimation of the elevator stroke is realized by using the recurrent neural network, the complex time sequence signal can be modeled, and the current state of the elevator motion can be output in real time, so that the independent stroke and the composite stroke in the elevator stroke can be obtained. The recurrent neural network is composed of 2 layers of LSTM units, each layer of LSTM units adopts 48 hidden units, and the prediction structure of many-to-one is used. And (3) acquiring elevator time sequence data, then marking the time sequence data, and training a network to obtain a model capable of being produced and deployed.

Install the sensor and carry out elevator terminal detection on the elevator, carry out the analysis with the testing result, realize the analysis to elevator running state, provide for follow-up other businesses and take advantage of the ladder regular data, if improve the stability analysis of elevator in the operation process, improve the whole security performance of elevator, maintenance efficiency when improving the elevator trouble knows possible trouble problem fast.

Drawings

FIG. 1 is a diagram of a recurrent neural network of the present invention;

FIG. 2 is a block diagram of an LSTM unit of the present invention;

FIG. 3 is a model structure of the present invention;

FIG. 4 is a system flow diagram of the present invention.

Detailed Description

The invention is further described with reference to the following figures and detailed description.

As shown in fig. 1 to 4, a statistical analysis system for elevator trips comprises a data acquisition module, a door state analysis module, a floor analysis module, an LSTM trip state analysis module, a trip analysis module, and a cloud storage module.

A data acquisition module: the method comprises the steps of collecting image data through a camera on the top of an elevator car, and collecting data of a gyroscope, an accelerometer, a barometer and a photoelectric tube through a sensor.

A door state analysis module: and the real-time door state of the elevator is obtained by using the image data acquired by the data acquisition module and utilizing image analysis. Four states (door open, door close) are provided.

Floor analysis module: the method comprises the steps that data of a gyroscope, an accelerometer, a barometer and a photoelectric tube collected by a data collection module are analyzed by Kalman filtering and complementary filtering to obtain the number of real-time floors of an elevator, wherein the photoelectric tube is installed on a floor 1 reference floor and used for calibrating the elevator to reach the reference floor, the Kalman filtering is used for establishing a motion model for observation data of the gyroscope and the accelerometer to estimate a running distance, sampling frequencies of sensors are often inconsistent, and the complementary filtering is used for integrating the estimated distance and height data estimated by the barometer to finally obtain the floor to which the elevator moves.

LSTM journey state analysis module: the data of the door state data and the data of the gyroscope, the accelerometer and the barometer collected by the data collection module are preprocessed, noise data are removed, the data frequency of 25HZ is sampled, model input vectors are arranged and sent to the LSTM model, and the LSTM model outputs a stroke state at intervals. The LSTM model predicts a state by adopting a many-to-one prediction mode, as shown in FIG. 1, namely, inputting a plurality of time segments, and improves the modeling capability of a long time span. The LSTM model incorporates a structure for determining whether information is useful or not, and this structure is called a cell. Three doors, namely an input door, a forgetting door and an output door, are placed in the cell. One message goes into the LSTM model and can be used to determine if it is useful based on three doors. Corresponding model parameters can be generated through training of the LSTM, the three doors control the opening and closing degree of the doors according to calculation of the model parameters and input information, and information is processed through the output door or discarded through the forgetting door under the action of the three doors.

The structure and input/output of the LSTM model are as follows. The structure of the LSTM model adopts a two-layer LSTM stacking mode, as shown in FIG. 3, which is helpful for carrying out deeper nonlinear prediction and improving the expression capability of the model. The lower row of dots is used as an input node, the upper single dot is used as an output node, each input dot is represented as a vector, the vector contains input parameters of the current moment, and the vector consists of 8 variables in total, namely a gate state (1 variable), a gyroscope (3 variables), an accelerometer (3 variables) and a barometer (1 variable). I.e. all variables need only be superimposed to form a vector. The collected vector data is preprocessed to remove noise and remove high-frequency interference, and 64 times of sampling is used as a group of input, namely 64 input vectors are sent into a model through a fixed sampling interval, namely 0.04 second, and a stroke state is output. Meanwhile, there is 50% coincidence between the input of the current batch and the input of the next batch, that is, 32 vectors in 64 input vectors are last input data, that is, 1.28 seconds are output once travel states.

A journey analysis module: the frequency of floor change data sampling is synchronous with the frequency output by the LSTM journey state analysis module, wherein the start in the journey state represents the beginning of the journey, the pause represents the transfer of the journey, and the end represents the end of the journey. By determining the starting and ending states, the elevator journey, i.e. the change in floor, can be analyzed, and the journey without a pause state during the starting and ending periods can be regarded as an independent journey, and the journey with at least one pause state can be regarded as a compound journey.

The elevator trip refers to a process of continuously running the elevator upward or downward, including an independent trip and a combined trip, wherein the independent trip of the elevator is defined as a process triggered from the state of opening the door to the state of closing the door, then moving, and then moving from the state of closing the door to the state of opening the door, and the elevator is in an idle state before and after the process or the moving direction of the elevator is opposite to the running direction before the elevator. For example, the elevator idles from floor 1, then moves to floor 10 and then idles, or the elevator last moves from floor 5 to floor 1, does not idle and moves from floor 1 to floor 10, then idles, and then defines the moving process from floor 1 to floor 10 as an independent journey. Therefore, the independent travel is standing at the angle of the elevator taking person, and the elevator taking process of the elevator taking person with the same travel is described. The compound travel of the elevator is defined as the elevator taking target with different travel, and the elevator taking process with overlapped independent travel exists, which is often combined by several independent travels and simultaneously requires the consistent motion direction of the sub-travel. As can be seen from the above definitions, the compound travel of the elevator is distinguished from the independent travel, for example, if people go from floor 1 to floor 9 and people go from floor 1 to floor 12, and the two groups of people go from floor 1 at the same time, then there is a person going out of floor 9 and then going out of floor 12 again, i.e., a travel of 1- >9- > 12.

The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and decorations can be made without departing from the spirit of the present invention, and these modifications and decorations should also be regarded as being within the scope of the present invention.

Claims

1. A statistical analysis system for elevator travel is characterized by comprising a data acquisition module, a door state analysis module, a floor analysis module, an LSTM travel state analysis module, a travel analysis module and a cloud storage module;

2. A system according to claim 1, characterized in that the total number of travel states is 5, idle, start, run, pause and end.