CN111890125A - Cutter state online monitoring method and management system - Google Patents

Cutter state online monitoring method and management system Download PDF

Info

Publication number
CN111890125A
CN111890125A CN202010621452.7A CN202010621452A CN111890125A CN 111890125 A CN111890125 A CN 111890125A CN 202010621452 A CN202010621452 A CN 202010621452A CN 111890125 A CN111890125 A CN 111890125A
Authority
CN
China
Prior art keywords
cutter
tool
data
monitoring
module
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010621452.7A
Other languages
Chinese (zh)
Other versions
CN111890125B (en
Inventor
王维龙
郑孟凯
杨开益
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xiamen Runtop Iot Technology Co ltd
Original Assignee
Xiamen Runtop Iot Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xiamen Runtop Iot Technology Co ltd filed Critical Xiamen Runtop Iot Technology Co ltd
Priority to CN202010621452.7A priority Critical patent/CN111890125B/en
Publication of CN111890125A publication Critical patent/CN111890125A/en
Application granted granted Critical
Publication of CN111890125B publication Critical patent/CN111890125B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23Q—DETAILS, COMPONENTS, OR ACCESSORIES FOR MACHINE TOOLS, e.g. ARRANGEMENTS FOR COPYING OR CONTROLLING; MACHINE TOOLS IN GENERAL CHARACTERISED BY THE CONSTRUCTION OF PARTICULAR DETAILS OR COMPONENTS; COMBINATIONS OR ASSOCIATIONS OF METAL-WORKING MACHINES, NOT DIRECTED TO A PARTICULAR RESULT
    • B23Q17/00—Arrangements for observing, indicating or measuring on machine tools
    • B23Q17/09—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23Q—DETAILS, COMPONENTS, OR ACCESSORIES FOR MACHINE TOOLS, e.g. ARRANGEMENTS FOR COPYING OR CONTROLLING; MACHINE TOOLS IN GENERAL CHARACTERISED BY THE CONSTRUCTION OF PARTICULAR DETAILS OR COMPONENTS; COMBINATIONS OR ASSOCIATIONS OF METAL-WORKING MACHINES, NOT DIRECTED TO A PARTICULAR RESULT
    • B23Q17/00—Arrangements for observing, indicating or measuring on machine tools
    • B23Q17/09—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool
    • B23Q17/0952—Arrangements for observing, indicating or measuring on machine tools for indicating or measuring cutting pressure or for determining cutting-tool condition, e.g. cutting ability, load on tool during machining
    • B23Q17/0957—Detection of tool breakage

Landscapes

  • Engineering & Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Machine Tool Sensing Apparatuses (AREA)

Abstract

The invention relates to a cutter state on-line monitoring method and a management system, which comprise a damage monitoring method and a wear monitoring method, wherein the damage monitoring method mainly aims at monitoring the damage condition of a cutter and adopts an improved multi-characteristic time sequence dynamic matching algorithm for monitoring; the wear monitoring sub-method mainly aims at monitoring the wear condition of the cutter and adopts an improved Laser-recurrent neural network model for monitoring; the management system comprises a basic information module, a cutter receiving module, a cutter purchasing module, a cutter using module, a cutter returning module, a cutter monitoring and early warning module, a cutter polishing module, a laser signal processing module and a sensor signal processing module. The invention can practically solve the problems of lack of real-time performance, accuracy and comprehensiveness of the traditional cutter state monitoring and effectively improve the production efficiency and the product quality of the manufacturing industry.

Description

Cutter state online monitoring method and management system
Technical Field
The invention relates to the field of industrial Internet and intelligent manufacturing, in particular to a cutter state online monitoring method and a cutter state online monitoring management system.
Background
With the steady advance of the strategy of intelligent manufacturing in China, the automation and intelligence level of the manufacturing industry is higher and higher, and the application of numerical control machines is wider and wider. The cutter is used as a core component of a numerical control machine tool, and the quality of the state of the cutter directly affects the final quality of a product, so that the monitoring of the state of the cutter in the numerical control machining process is particularly important.
At present, the monitoring of the state of the cutter in the traditional numerical control machining process lacks scientific, comprehensive and effective intelligent monitoring and early warning measures, and has the main problems that: firstly, the tool state monitoring is mostly carried out in a mode of manually detecting by means of auxiliary equipment under the condition of stopping and detaching the tool, the real-time monitoring cannot be carried out, the production time is occupied, the efficiency is low, and the tool state monitoring device is not suitable for large-scale production environment; secondly, the tool is damaged and worn, the requirements of tool state monitoring on timeliness and accuracy are different under the two conditions, and the traditional tool state monitoring system lacks effective and differentiated monitoring measures to comprehensively monitor the two conditions. This leads to the more prominent state monitoring and management problems of the current tools, which further affects the production efficiency, product quality and enterprise image of the enterprise.
Disclosure of Invention
The invention provides a cutter state online monitoring method and a management system, aiming at solving the problems of traditional cutter damage and abnormal wear state monitoring.
The invention adopts the following technical scheme:
a cutter state online monitoring method is characterized in that: and the improved multi-feature time sequence dynamic matching algorithm is adopted to realize damage monitoring, and the improved Laser-recurrent neural network model is adopted to realize wear monitoring.
Preferably, the breakage monitoring comprises the steps of:
A1) obtaining the laser reflection wave time sequence signal Q on the surface of the cutter in the normal state1Acquiring the acoustic emission signal time sequence set Q of the cutter in the normal processing process2And machine tool power time series set Q3Preprocessing the signal data;
A2) respectively give Q1、Q2、Q3Weight value mu1、μ2And mu3Then, the comprehensive time sequence matrix XL is obtained through weight adjustment and is divided into a generation standard sequence set ToAnd computing a set of threshold sequences T1；
A3) According to the generation standard sequence set ToDetermination of the Standard average sequence Xs；
A4) Computing a set of threshold sequences T1And standard average sequence XsThe minimum cumulative distance of (b) as a threshold value β;
A5) collecting laser reflection data, acoustic emission data and machine tool power data in real time during the machining process of the cutter, and calculating the data and a standard average sequence XsMinimum cumulative distance of
Figure BDA0002563157450000021
If the minimum cumulative distance
Figure BDA0002563157450000022
If the threshold value is larger than the beta value, a damage alarm is sent out.
Preferably, the step a3) specifically includes the following steps:
a3.1) from the set of standard sequences ToRandomly selecting a sequence X as an initial average sequence, and calculating ToThe minimum cumulative distance between the other sequences in (b) and the X sequence;
a3.2) determination of the X sequence to ToIf the cumulative sum of squared distances of the sequences is decreased, updating each coordinate in the average sequence X to ToThe mean value of the coordinates matched with the coordinate and returning to the step A3.1); if not, taking the current average sequence as a standard average sequence XsAnd (6) outputting.
4. The tool state on-line monitoring method according to claim 1, characterized in that: the wear monitoring comprises the following steps:
B1) dynamically acquiring and calculating the wear volume of the cutter based on the improved line laser;
B2) utilizing an improved Laser-recurrent neural network model to dynamically track and monitor the worn cutter which does not reach the alarm condition;
B3) and inputting the characteristic data acquired in real time into the trained improved neural network model, and early warning the tool abrasion.
Preferably, the step B1) specifically includes the following steps
B1.1) obtaining a complete tool blade contour coordinate data set X1And worn tool edge profile coordinate dataset X2Respectively fitting the collected blade contour coordinate data sets by adopting polynomial fitting to remove noise points and obtain corresponding blade contour coordinate smooth curves;
b1.2) respectively calculating to obtain the complete blade contour area S according to the blade contour coordinate smooth curve1And worn edge profile area S2Calculating the complete blade profile area S1And the area S of the profile of the worn blade2Obtaining the abrasion area delta S by the difference value, and repeating the steps B1.1) to B1.2) to obtain the abrasion areas of all the scanning sections of the cutter;
b1.3) carrying out accumulation operation on the wear areas of all the scanning sections of the cutter to obtain the wear volume V of the cutting edge of the cutter, and judging the V and the wear threshold value
Figure BDA0002563157450000023
In a relation of (1), if
Figure BDA0002563157450000024
A wear alarm is issued.
6. The tool state on-line monitoring method according to claim 4, wherein: the step B2) specifically comprises the following steps
B2.1) carrying out data preprocessing on the acquired original signal data, forming characteristic data with the tool cutting edge abrasion volume V, and using the characteristic data as training data of an improved Laser-recurrent neural network model, wherein the original signal data comprises cutting speed, stress data, temperature data, feed data and cut time;
b2.2) defining parameters of the improved Laser-recurrent neural network model, including input x of time t sequencetAnd output information htAnd cell state CtSequence output information h at time t-1t-1And cell state Ct-1σ is sigmoid activation function, tanh is hyperbolic tangent function, xtTraining data is obtained;
b2.3) dividing h at the time of tt-1And training data xtInputting a forgetting gate function to calculate the forgetting cell state ftLet ht at time tt-1And xtInput neural network model calculation input gate information itAnd C*；
B2.4) use of the amnestic cell State ftAnd input gate information itAnd C*Calculating the cell renewal state C at time ttAccording to the cell renewal state CtCalculating the output information h at the current time ttRepeating the steps B2.3) to B2.4) and calculating all output values in the time sequence;
b2.5) reversely calculating total error value items on each time point of all output values on the time sequence, and solving the sum of first order partial derivatives of the weight matrix according to the total error value items to obtain the gradient of the weight matrix and finish the reverse propagation calculation of a time step;
b2.6) repeating the step B2.5) until the optimal weight matrix is solved, namely the gradient of the weight matrix is optimal, and finishing the model training.
The utility model provides a cutter state on-line management system which characterized in that: comprises that
The laser signal processing module is used for preprocessing laser signal data and transmitting the preprocessed data to the cutter monitoring and early warning module;
the sensor signal processing module is used for preprocessing data acquired by the sensor and transmitting the preprocessed data to the cutter monitoring and early warning module;
the cutter monitoring and early warning module comprises a damage monitoring unit and a wear monitoring unit; the breakage monitoring unit is used for realizing breakage monitoring on the received data by adopting an improved multi-feature time sequence dynamic matching algorithm; and the wear monitoring unit is used for realizing wear monitoring on the received data by adopting an improved Laser-recurrent neural network model.
Preferably, the system also comprises a basic information module, a cutter getting module, a cutter purchasing module, a cutter using module, a cutter returning module and a cutter polishing module; the basic information module is used for providing basic information management functions of organization architecture, user management, authority management, role management, operation logs and system parameters; the tool receiving module provides tool receiving data information management and tool receiving process approval and tracking; the tool purchasing module realizes the approval, tracking and purchasing supplier evaluation of the tool purchasing process; the tool using module is used for realizing information management of a tool using workshop, a using machine tool, a production work order, a user, tool setting parameters and a using state; the tool returning module realizes approval and tracking of the tool returning process; the cutter grinding module realizes the management of cutter grinding and maintenance information and process approval and tracking.
As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following advantages:
(1) the invention divides the cutter state monitoring into damage monitoring and wear monitoring, adopts different algorithm models for monitoring according to different characteristics of cutter damage and wear, can comprehensively and differentially cover various cutter abnormal conditions, and greatly improves the accuracy and comprehensiveness of cutter state monitoring.
(2) The damage monitoring method provided by the invention adopts laser combined with acoustic emission and power sensing data to perform multi-feature time sequence matching calculation, has high monitoring sensitivity and high calculation speed block, can find damage abnormity in time, and effectively solves the problem of high defective rate of processed products caused by cutter damage.
(3) The wear monitoring method adopts the laser to scan the cutter, carries out the comparison and calculation of the blade profile, accurately obtains the wear area and volume data of the blade, and then combines the improved recurrent neural network model to carry out multi-characteristic dynamic prediction on the wear data, thereby further improving the early warning capability of the cutter wear. The method can timely find and predict the abrasion degree of the small cutter abrasion change, has high accuracy, and effectively helps enterprises to improve production efficiency and reduce cost.
Drawings
Fig. 1 is a schematic flow chart of a breakage monitoring algorithm according to an embodiment of the present invention.
FIG. 2 is a schematic flow chart of a wear monitoring algorithm in an embodiment of the present invention.
Fig. 3 is a schematic diagram of a system structure according to an embodiment of the present invention.
The invention is described in further detail below with reference to the figures and specific examples.
Detailed Description
The invention is further described below by means of specific embodiments.
A cutter state online monitoring method comprises damage monitoring and wear monitoring; the damage monitoring mainly monitors the damage condition of the cutter, has high requirement on effectiveness, and adopts an improved multi-feature time sequence dynamic matching algorithm for monitoring; the wear monitoring mainly monitors the wear condition of the cutter, has high requirement on the wear accuracy of the cutter and belongs to predictive monitoring, and an improved Laser-recurrent neural network model is adopted for monitoring.
The RT-ITMS intelligent tool management system is taken as a prototype, and the embodiment of the invention is described in detail. The RT-ITMS intelligent tool management system is developed based on the Internet of things, big data and artificial intelligence technology, has perfect tool management function and database management function, and can carry out full life cycle management on tools of different manufacturers and different types. On the basis, the cutter state is monitored and early warned on line by adopting an improved multi-feature time sequence matching algorithm and an improved Laser-neural network algorithm, so that the problem of cutter state monitoring is effectively solved, and the intelligent level of cutter management and the production efficiency of enterprises are improved.
The breakage monitoring comprises the following steps:
A1) obtaining the laser reflection wave time sequence signal Q on the surface of the cutter in the normal state1Acquiring the acoustic emission signal time sequence set Q of the cutter in the normal processing process2And machine tool power time series set Q3And preprocessing the signal data.
Specifically, a laser sensor is used for emitting laser beams to irradiate the surface of the cutter at an angle theta, and the light flux of reflected light is collected through a photosensitive element; and carrying out differential amplification signal processing on the luminous flux of the reflected light to obtain the luminous flux variation delta rho.
Defining the light intensity U of the laser beam0Beam radius R, reflected light lens focal length f, tool surface wave number x, surface wave angular frequency w, surface wave amplitude u, photoelectric conversion coefficient
Figure BDA0002563157450000041
At time t, the amount of change in luminous flux is expressed as:
Figure BDA0002563157450000051
repeatedly collecting multiple groups of luminous flux variable quantities as a reflected wave time sequence set Q1。
Acquisition of acoustic emission signal time sequence set Q of tool in normal machining process by using acoustic emission sensor and power sensor2And machine tool power time series set Q3And preprocessing the signal data.
Examples are: irradiating the surface of the tool at an angle theta of 45 DEG, and setting the light intensity U of the laser beam010mW, the beam radius R is 10, the focal length f of the reflecting lens is 8cm, the surface wave number x of the cutter is 4, the angular frequency w of the surface wave is 135Hz, the amplitude u of the surface wave is 0.6,
Figure BDA0002563157450000052
reflected wave time series set Q1The following were used:
Figure BDA0002563157450000053
acoustic emission signal time series set Q2And machine tool power time series set Q3After data preprocessing, the method comprises the following steps:
Figure BDA0002563157450000054
and
Figure BDA0002563157450000055
A2) are respectively endowed withTo Q1、Q2、Q3Weight value mu1、μ2And mu3Then, the comprehensive time sequence matrix XL is obtained through weight adjustment,
XL＝[μ1·Q1，μ2·Q2，μ3·Q3]
the time sequence matrix XL is divided into a set of generation standard sequences ToAnd computing a set of threshold sequences T1。
Examples are: respectively give Q1、Q2、Q3The weights 0.5, 0.3, and 0.2, the integrated time series matrix XL after weight adjustment is:
XL＝[0.5Q1，0.3Q2，0.2Q3]
A3) according to the generation standard sequence set ToDetermination of the Standard average sequence XsThe method specifically comprises the following steps:
a3.1) from the set of standard sequences ToRandomly selecting a sequence X as an initial average sequence, and calculating ToThe minimum cumulative distance between the other sequences and the X sequence is expressed as:
Lk(i，j)＝l(i，j)+min{Lk-1(i-1，j)，Lk-1(i，j-1)，Lk-1(i-1，j-1)}
wherein i represents ToThe ith element in a sequence, j represents the jth element in the X sequence, and k represents ToMiddle k sequence, Lk(i, j) represents ToThe minimum cumulative distance between one sequence and the X sequence, l (i, j) represents ToThe distance between the ith element of the sequence and the jth element of the X sequence.
Examples are: from ToRandomly selecting a sequence X in the set as [0.2,0.5 … 0.4.4 ]]As the initial average sequence.
A3.2) determination of the X sequence to ToIf the cumulative sum of squared distances of the sequences is decreased, updating each coordinate in the average sequence X to ToThe mean value of the coordinates matched with the coordinate and returning to the step A3.1); if not, taking the current average sequence as a standard average sequence XsOutputs, for example: xs＝[0.5,0.3…0.5]。
A4) Computing a set of threshold sequences T1And standard average sequence XsIs taken as a threshold value β, for example β is 0.5.
A5) Collecting laser reflection data, acoustic emission data and machine tool power data in real time during the machining process of the cutter, and calculating the data and a standard average sequence XsMinimum cumulative distance of
Figure BDA0002563157450000061
If the minimum cumulative distance
Figure BDA0002563157450000062
If the threshold value is larger than the beta value, a damage alarm is sent out.
The wear monitoring comprises the following steps:
B1) and dynamically acquiring and calculating the wear volume of the tool based on the improved line laser. Specifically, the method comprises the following steps
B1.1) emitting laser to the vertical direction of the cutting edge of the complete cutter and the worn cutter by using an industrial laser, and acquiring a complete cutter edge contour coordinate data set X by a reflected light triangulation imaging method1And worn tool edge profile coordinate dataset X2And respectively fitting the collected blade contour coordinate data sets by adopting polynomial fitting to remove noise points and obtain corresponding blade contour coordinate smooth curves. E.g. complete tool edge profile coordinate dataset X1＝{[1.2,1.4],[2.3,3.1],[4.3,4.1]…[9.1,1.3]And worn edge profile coordinate dataset X2＝{[1.4,2.1],[3.2,4.3],[4.6,5.6]…[11.2,2.1]}。
B1.2) transforming and translating the X through a coordinate system1And X2The edge contour coordinate smooth curves are unified to the same coordinate system; calculating X in a two-dimensional coordinate system1Area S enclosed by coordinate set and x axis1,X2Area S enclosed by coordinate set and x axis2(ii) a Calculating the complete blade profile area S1And the area S of the profile of the worn blade2To obtain the wear area Δ S, Δ S ═ S1-S2(ii) a Laser deviceScanning is carried out along the edge line direction at a spacing distance d, for example d is 2mm, and steps B1.1) -B1.2) are repeated to obtain all the scanning section wear areas of the cutter.
For example: s1＝9.4(mm2)，S2＝7.8(mm2)，ΔS＝1.6(mm2)
B1.3) carrying out accumulation operation on the wear areas of all the scanning sections of the cutter to obtain the wear volume V of the cutting edge of the cutter. Wherein the wear area within the separation distance d is averaged
Figure BDA0002563157450000063
The number of laser scanning sections l, for example, is:
Figure BDA0002563157450000064
Figure BDA0002563157450000065
wherein: d represents the unit spacing distance of the worn area, Δ SdAnd Δ Sd+1Representing the area of wear on both sides at a separation distance d, dtRepresenting the differential of t.
For example: l is 50, then V is 160 (mm)3)。
Judging V and wear threshold
Figure BDA0002563157450000066
In a relation of (1), if
Figure BDA0002563157450000067
A wear alarm is issued. E.g. threshold value
Figure BDA0002563157450000068
B2) Utilizing an improved Laser-recurrent neural network model to dynamically track and monitor the worn cutter which does not reach the alarm condition; specifically, the method comprises the following steps
B2.1) using a sensor to carry out data acquisition on the working condition of the cutter, and acquiring original signal dataAnd (4) preprocessing line data, forming characteristic data with the normalized cutter edge wear volume V, and taking the characteristic data as training data I of an improved neural network modelx1，Ix2，...，Ix6]TWhere x represents the number of rows in the matrix, Ix1Representing the first column of the data set. The preprocessing comprises signal feature dimension reduction, screening, feature extraction and the like based on time-frequency domain analysis, and the feature data obtained by the sensor comprises cutting speed, stress data, temperature data, feed data, cut time and the like.
For example, the model input data is as follows:
Figure BDA0002563157450000071
b2.2) defining neural network model parameters including input x of t time sequencetAnd output information htAnd cell state CtSequence output information h at time t-1t-1(initial state ht-10) and cell state Ct-1(in the initial state, Ct-10); sigma is sigmoid activation function and takes the value of [0, 1%]To (c) to (d); tanh is a hyperbolic tangent function with a value of [ -1,1 [)]In the meantime.
B2.3) dividing h at the time of tt-1And xtInputting a forgetting gate function to calculate the forgetting cell state ft，
ft＝σ(Wf·[ht-1，xt]+bf)
Wherein, WfTo forget the weight matrix of the gate, bfTo forget the gate bias term. For example: setting t to h at time 10＝0,x1＝0.44，bfInputting a forgetting gate function to calculate the forgetting cell state f1The expression is as follows:
f1＝σ(Wf·[0，0.44]+0.7)。
h at tt-1And xtInput neural network model calculation input gate information itAnd C*；
it＝σ(Wi·[ht-1，xt]+bi)
C*＝tanh(Wc·[ht-1，xt]+bc)
Wherein, WiAnd WcAs input gate weight matrix, biAnd bcThe gate bias term is entered.
For example: bi＝0.6，bcWhen the value is equal to 0.9, then i1＝σ(Wi·[0，0.44]+0.6)，C*＝tanh(Wc·[0，0.44]+0.9)。
B2.4) use of the amnestic cell State ftAnd inputting gate information to calculate cell update state C at time tt：
Ct＝Ct-1⊙ft+it⊙C*
According to cell renewal state CtCalculating the output information h at the current time tt：
ht＝σ(Wo·[ht-1，xt]+bo)⊙tanh(Ct)
Wherein, WoAs a weight matrix of output gates, boThe gate bias term is output.
Examples are: calculation of cell renewal State C1The expression is as follows: c1＝C0⊙f1+i1⊙C*According to the cell renewal state C1Calculating the output information h of the current time1The expression is as follows:
h1＝σ(Wo·[0，0.44]+0.5)⊙tanh(C1)
wherein, WoAs a weight matrix of output gates, bo0.5 is the output gate offset term.
Repeating steps B2.3) -B2.4), calculating all output values in time series.
B2.5) reversely calculating the total error value item at each time point according to all the output values in the time sequence, and according to the total error value item, obtaining the weight matrix Wf、Wi、WcAnd WoSumming the first order partial derivatives to obtain the weightsValue matrix gradient, finishing the back propagation calculation of a time step;
b2.6) repeating the step B2.5) until the optimal weight matrix is solved, namely the gradient of the weight matrix is optimal, and finishing the model training.
B3) And inputting the characteristic data acquired in real time into the trained improved neural network model, and early warning the tool abrasion.
The invention also provides an on-line tool state management system, which comprises a laser signal processing module 800, a sensor signal processing module 900 and a tool monitoring and early warning module 600. The laser signal processing module 800 is connected to the laser collector hardware outside the system, and is configured to preprocess the laser signal data and transmit the preprocessed data to the tool monitoring and early warning module 600. The sensor signal processing module 900 is connected to sensor hardware outside the system, and is configured to preprocess data collected by the sensor and transmit the preprocessed data to the tool monitoring and warning module 600.
The cutter monitoring and early warning module 600 comprises a damage monitoring unit and a wear monitoring unit; the damage monitoring unit is used for realizing damage monitoring on the received data by adopting an improved multi-feature time sequence dynamic matching algorithm and giving an alarm on abnormal conditions; the wear monitoring unit is used for monitoring wear of the received data by adopting an improved Laser-recurrent neural network algorithm and giving an alarm for abnormal conditions.
The system of the present invention further comprises a basic information module 100, a tool receiving module 200, a tool purchasing module 300, a tool using module 400, a tool returning module 500 and a tool sharpening module 700. The basic information module 100 is used for providing basic information management functions such as organization architecture, user management, authority management, role management, running logs, system parameters and the like; the tool acceptance module 200 provides tool acceptance data information management and tool acceptance process approval and tracking; the tool procurement module 300 realizes tool procurement process approval, tracking and procurement supplier evaluation; the tool use module 400 realizes information management of a tool use workshop, a machine tool, a production work order, a user, tool setting parameters, a use state and the like; the tool return module 500 realizes approval and tracking of the tool return process; the tool sharpening module 700 implements management and process approval and tracking of tool sharpening and maintenance information.
All management modules of the system are deployed on the application server, and the system further comprises a data server used for managing the full life cycle data of the cutter. The invention creatively combines the technologies of laser measurement, sensor signal acquisition and processing, an improved time sequence matching algorithm, an improved recurrent neural network algorithm and the like, establishes a corresponding management system, is comprehensively applied to the monitoring and management of the cutter state under complex working conditions, and realizes the dynamic monitoring in the cutter processing process. By adopting the technical scheme, the problem of online monitoring of the cutter state of the manufacturing enterprise can be effectively solved, and the production efficiency and the product quality of the enterprise are improved.
The above description is only an embodiment of the present invention, but the design concept of the present invention is not limited thereto, and any insubstantial modifications made by using the design concept should fall within the scope of infringing the present invention.

Claims (8)

1. A cutter state online monitoring method is characterized in that: and the improved multi-feature time sequence dynamic matching algorithm is adopted to realize damage monitoring, and the improved Laser-recurrent neural network model is adopted to realize wear monitoring.
2. The tool state on-line monitoring method according to claim 1, characterized in that: the breakage monitoring comprises the following steps:
A1) obtaining the laser reflection wave time sequence signal Q on the surface of the cutter in the normal state1Acquiring the acoustic emission signal time sequence set Q of the cutter in the normal processing process2And machine tool power time series set Q3Preprocessing the signal data;
A2) respectively give Q1、Q2、Q3Weight value mu1、μ2And mu3Then, the comprehensive time sequence matrix XL is obtained through weight adjustment and is divided into a generation standard sequence set ToAnd computing a set of threshold sequences T1；
A3) According to the generation standard sequence set ToDetermination of the Standard average sequence Xs；
A4) Computing a set of threshold sequences T1And standard average sequence XsThe minimum cumulative distance of (b) as a threshold value β;
A5) collecting laser reflection data, acoustic emission data and machine tool power data in real time during the machining process of the cutter, and calculating the data and a standard average sequence XsMinimum cumulative distance of
Figure FDA0002563157440000011
If the minimum cumulative distance
Figure FDA0002563157440000012
If the threshold value is larger than the beta value, a damage alarm is sent out.
3. The tool state on-line monitoring method according to claim 2, characterized in that: the step a3) specifically includes the following steps:
a3.1) from the set of standard sequences ToRandomly selecting a sequence X as an initial average sequence, and calculating ToThe minimum cumulative distance between the other sequences in (b) and the X sequence;
a3.2) determination of the X sequence to ToIf the cumulative sum of squared distances of the sequences is decreased, updating each coordinate in the average sequence X to ToThe mean value of the coordinates matched with the coordinate and returning to the step A3.1); if not, taking the current average sequence as a standard average sequence XsAnd (6) outputting.
4. The tool state on-line monitoring method according to claim 1, characterized in that: the wear monitoring comprises the following steps:
B1) dynamically acquiring and calculating the wear volume of the cutter based on the improved line laser;
B2) utilizing an improved Laser-recurrent neural network model to dynamically track and monitor the worn cutter which does not reach the alarm condition;
B3) and inputting the characteristic data acquired in real time into the trained improved neural network model, and early warning the tool abrasion.
5. The tool state on-line monitoring method according to claim 4, wherein: the step B1) specifically comprises the following steps
B1.1) obtaining a complete tool blade contour coordinate data set X1And worn tool edge profile coordinate dataset X2Respectively fitting the collected blade contour coordinate data sets by adopting polynomial fitting to remove noise points and obtain corresponding blade contour coordinate smooth curves;
b1.2) respectively calculating to obtain the complete blade contour area S according to the blade contour coordinate smooth curve1And worn edge profile area S2Calculating the complete blade profile area S1And the area S of the profile of the worn blade2Obtaining the abrasion area delta S by the difference value, and repeating the steps B1.1) to B1.2) to obtain the abrasion areas of all the scanning sections of the cutter;
b1.3) carrying out accumulation operation on the wear areas of all the scanning sections of the cutter to obtain the wear volume V of the cutting edge of the cutter, and judging the V and the wear threshold value
Figure FDA0002563157440000021
In a relation of (1), if
Figure FDA0002563157440000022
A wear alarm is issued.
6. The tool state on-line monitoring method according to claim 4, wherein: the step B2) specifically comprises the following steps
B2.1) carrying out data preprocessing on the acquired original signal data, forming characteristic data with the tool cutting edge abrasion volume V, and using the characteristic data as training data of an improved Laser-recurrent neural network model, wherein the original signal data comprises cutting speed, stress data, temperature data, feed data and cut time;
b2.2) defining improved Laser-cyclesParameters of a neural network model, including input x of a sequence of times ttAnd output information htAnd cell state CtSequence output information h at time t-1t-1And cell state Ct-1σ is sigmoid activation function, tanh is hyperbolic tangent function, xtTraining data is obtained;
b2.3) dividing h at the time of tt-1And training data xtInputting a forgetting gate function to calculate the forgetting cell state ftH at time tt-1And xtInput neural network model calculation input gate information itAnd C*；
B2.4) use of the amnestic cell State ftAnd input gate information itAnd C*Calculating the cell renewal state C at time ttAccording to the cell renewal state CtCalculating the output information h at the current time ttRepeating the steps B2.3) to B2.4) and calculating all output values in the time sequence;
b2.5) reversely calculating total error value items on each time point of all output values on the time sequence, and solving the sum of first order partial derivatives of the weight matrix according to the total error value items to obtain the gradient of the weight matrix and finish the reverse propagation calculation of a time step;
b2.6) repeating the step B2.5) until the optimal weight matrix is solved, namely the gradient of the weight matrix is optimal, and finishing the model training.
7. The utility model provides a cutter state on-line management system which characterized in that: comprises that
The laser signal processing module is used for preprocessing laser signal data and transmitting the preprocessed data to the cutter monitoring and early warning module;
the sensor signal processing module is used for preprocessing data acquired by the sensor and transmitting the preprocessed data to the cutter monitoring and early warning module;
the cutter monitoring and early warning module comprises a damage monitoring unit and a wear monitoring unit; the breakage monitoring unit is used for realizing breakage monitoring on the received data by adopting an improved multi-feature time sequence dynamic matching algorithm; and the wear monitoring unit is used for realizing wear monitoring on the received data by adopting an improved Laser-recurrent neural network model.
8. The tool state online management system of claim 7, wherein: the device also comprises a basic information module, a cutter receiving module, a cutter purchasing module, a cutter using module, a cutter returning module and a cutter polishing module; the basic information module is used for providing basic information management functions of organization architecture, user management, authority management, role management, operation logs and system parameters; the tool receiving module provides tool receiving data information management and tool receiving process approval and tracking; the tool purchasing module realizes the approval, tracking and purchasing supplier evaluation of the tool purchasing process; the tool using module is used for realizing information management of a tool using workshop, a using machine tool, a production work order, a user, tool setting parameters and a using state; the tool returning module realizes approval and tracking of the tool returning process; the cutter grinding module realizes the management of cutter grinding and maintenance information and process approval and tracking.
CN202010621452.7A 2020-06-30 2020-06-30 Cutter state online monitoring method and management system Active CN111890125B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010621452.7A CN111890125B (en) 2020-06-30 2020-06-30 Cutter state online monitoring method and management system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010621452.7A CN111890125B (en) 2020-06-30 2020-06-30 Cutter state online monitoring method and management system

Publications (2)

Publication Number Publication Date
CN111890125A true CN111890125A (en) 2020-11-06
CN111890125B CN111890125B (en) 2021-06-22

Family

ID=73191344

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010621452.7A Active CN111890125B (en) 2020-06-30 2020-06-30 Cutter state online monitoring method and management system

Country Status (1)

Country Link
CN (1) CN111890125B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112496860A (en) * 2020-11-27 2021-03-16 张海强 Method for monitoring service life of turning tool in real time
CN115905886A (en) * 2023-01-09 2023-04-04 北京和利时系统集成有限公司 Fault diagnosis method and device for platform screen door and storage medium
CN116126836A (en) * 2023-04-12 2023-05-16 山东越海通信科技股份有限公司 Digital intelligent monitoring system and method for optical fiber device processing equipment

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1398598A1 (en) * 2002-09-16 2004-03-17 WENDT GmbH Method and device to measure small cutting edge geometries
CN105150030A (en) * 2015-08-31 2015-12-16 厦门金鹭特种合金有限公司 Self-adaption intelligent system for monitoring abnormal conditions of turning tool and method thereof
DE102016002995B3 (en) * 2016-03-14 2017-03-02 Gebr. Heller Maschinenfabrik Gmbh Method for monitoring a drive system of a machine tool
CN107553219A (en) * 2017-10-23 2018-01-09 沈阳百祥机械加工有限公司 A kind of Tool Wear Monitoring method based on multiple types sensor composite signal
CN206883328U (en) * 2017-07-13 2018-01-16 厦门嵘拓物联科技有限公司 A kind of cutting tool state detection means based on voice signal
KR20180048218A (en) * 2016-10-31 2018-05-10 현대위아 주식회사 Machine tool breakdown diagnosis system based on Machine Learning, and method thereof
CN109434564A (en) * 2018-12-21 2019-03-08 哈尔滨理工大学 A kind of cutter wear state monitoring method based on deep neural network
CN109934424A (en) * 2019-04-30 2019-06-25 厦门嵘拓物联科技有限公司 A kind of cutter life prediction technique and intelligent tool management system

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1398598A1 (en) * 2002-09-16 2004-03-17 WENDT GmbH Method and device to measure small cutting edge geometries
CN105150030A (en) * 2015-08-31 2015-12-16 厦门金鹭特种合金有限公司 Self-adaption intelligent system for monitoring abnormal conditions of turning tool and method thereof
DE102016002995B3 (en) * 2016-03-14 2017-03-02 Gebr. Heller Maschinenfabrik Gmbh Method for monitoring a drive system of a machine tool
KR20180048218A (en) * 2016-10-31 2018-05-10 현대위아 주식회사 Machine tool breakdown diagnosis system based on Machine Learning, and method thereof
CN206883328U (en) * 2017-07-13 2018-01-16 厦门嵘拓物联科技有限公司 A kind of cutting tool state detection means based on voice signal
CN107553219A (en) * 2017-10-23 2018-01-09 沈阳百祥机械加工有限公司 A kind of Tool Wear Monitoring method based on multiple types sensor composite signal
CN109434564A (en) * 2018-12-21 2019-03-08 哈尔滨理工大学 A kind of cutter wear state monitoring method based on deep neural network
CN109934424A (en) * 2019-04-30 2019-06-25 厦门嵘拓物联科技有限公司 A kind of cutter life prediction technique and intelligent tool management system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
傅致远，王仲逸，吴一凡，张铫，周贺，路云浩: "基于DTW和DBA算法的CNC刀具破损在线检测", 《组合机床与自动化加工技术》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112496860A (en) * 2020-11-27 2021-03-16 张海强 Method for monitoring service life of turning tool in real time
CN115905886A (en) * 2023-01-09 2023-04-04 北京和利时系统集成有限公司 Fault diagnosis method and device for platform screen door and storage medium
CN116126836A (en) * 2023-04-12 2023-05-16 山东越海通信科技股份有限公司 Digital intelligent monitoring system and method for optical fiber device processing equipment
CN116126836B (en) * 2023-04-12 2023-06-13 山东越海通信科技股份有限公司 Digital intelligent monitoring system and method for optical fiber device processing equipment

Also Published As

Publication number Publication date
CN111890125B (en) 2021-06-22

Similar Documents

Publication Publication Date Title
CN111890125B (en) Cutter state online monitoring method and management system
US20210287098A1 (en) On-line prediction method of surface roughness of parts based on sdae-dbn algorithm
CN103056759B (en) Robot grinding system based on feedback of sensor
CN110059442B (en) Turning tool changing method based on part surface roughness and power information
CN108490880B (en) Method for monitoring wear state of cutting tool of numerical control machine tool in real time
Lu Study on prediction of surface quality in machining process
Palani et al. Prediction of surface roughness in CNC end milling by machine vision system using artificial neural network based on 2D Fourier transform
US20230288920A1 (en) In-process digital twinning
Song et al. A method for grinding removal control of a robot belt grinding system
TWI481978B (en) Method for predicting machining quality of machine tool
CN108907896B (en) A kind of cutter remaining life on-line prediction method and system
CN112518425B (en) Intelligent machining cutter wear prediction method based on multi-source sample migration reinforcement learning
CN113496312A (en) Cutter residual life prediction method and equipment based on multi-dimensional feature extraction fusion and long-term and short-term memory network and storage medium
CN111126255A (en) Numerical control machine tool cutter wear value prediction method based on deep learning regression algorithm
Pereverzev et al. Designing optimal automatic cycles of round grinding based on the synthesis of digital twin technologies and dynamic programming method
Cheng et al. Tool wear intelligent monitoring techniques in cutting: a review
Wang et al. Towards smart remanufacturing and maintenance of machinery-review of automated inspection, condition monitoring and production optimisation
Brecher et al. Evaluation of toolpath quality: User-assisted CAM for complex milling processes
EP3736648B1 (en) Method for autonomous optimization of a grinding process
CN114169131A (en) Mold manufacturing system based on digital multi-mode
Guo Surface roughness prediction by combining static and dynamic features in cylindrical traverse grinding
Taga et al. Determination of cutting parameters in end milling operation based on the optical surface roughness measurement
CN112946072A (en) Abrasive belt wear state monitoring method based on machine learning
WO2021253689A1 (en) Multiple regression model-based method and system for predicting price of product processing
Reser et al. Grinding burn prediction with artificial neural networks based on grinding parameters

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant