CN110489861A - A kind of elastic implementation method of AI boundary operation chip design - Google Patents
A kind of elastic implementation method of AI boundary operation chip design Download PDFInfo
- Publication number
- CN110489861A CN110489861A CN201910763925.4A CN201910763925A CN110489861A CN 110489861 A CN110489861 A CN 110489861A CN 201910763925 A CN201910763925 A CN 201910763925A CN 110489861 A CN110489861 A CN 110489861A
- Authority
- CN
- China
- Prior art keywords
- chip
- design
- waste
- algorithm
- function
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Educational Administration (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to the design fields for having boundary operation ability chip in enterprise production management, disclose a kind of elastic implementation method of AI boundary operation chip design, comprising the following steps: the waste confirmation greatly of production line scene eight；Essence, which is grown directly from seeds, produces improvement plan foundation；Internet of Things perceptional function, algorithm, Decision-making project evaluation；Edge calculations degree determines and chip functions design；Self-criticism and simulation after chip design, chip structure layout, design；Actual circuit simulation run test；Flow production and test.Chip design scheme that the present invention provides complete set, having Technology Correlation, successively confirm that chip should have function, should be applicable in the information such as algorithm, circuit layout according to the waste actual conditions greatly of scene eight, be conducive to the design that elasticity realizes chip, specification chip design cycle, relevance is higher, but also chip specific aim with higher, reduces the place of finished chip using difficulty, improves chip desin speed and reliability.
Description
Technical field
The present invention relates to the design fields more particularly to a kind of AI that have boundary operation ability chip in enterprise production management
The elastic implementation method of boundary operation chip design.
Background technique
The design of current chip mostly uses a large amount of manpowers, and carries out personal or group's decentralized design, mutual connection
Property it is inadequate, in addition, chip designer, can't be grown directly from seeds according to the essence that actual production line needs produces regulatory requirement and be directed to
Property chip design, therefore be often only capable of producing the chip of general-using type, this also leads to the volume of subsequent design edge calculations function
Cheng personnel's difficulty is promoted, and desin speed is slack-off, and integration is not unified in advance for design scheme, also can not functional perfect design
Expert database is examined for automatic checking and self-criticism, chip design is significantly increased and using difficulty.
Summary of the invention
The technical problem to be solved by the present invention is in view of the above shortcomings of the prior art, provide a kind of AI boundary operation chip
The elastic implementation method of design.
To achieve the goals above, the technical solution adopted by the present invention are as follows:
A kind of elastic implementation method of AI boundary operation chip design, comprising the following steps:
The waste confirmation greatly of production line scene eight: verification assessment is carried out according to expert database data, confirms the pipe at production line scene
Excessively waste, the superfluous waste of processing, carrying waste, inventory excessively waste, wait waste for reason waste, bad amendment waste, manufacture
And it acts the case where any one or more waste in wasting and reports grade separation for repairment；
Essence, which is grown directly from seeds, produces improvement plan foundation: the case where to waste with report expert database data in conjunction with grade separation for repairment and check
With after legitimacy assessment using PDCA method, man-machine material method is around-France, appointing in DEMAC method, management program method and risk assessment method
One or more methods of anticipating, which establish essence and grow directly from seeds, produces improvement plan；
Internet of Things perceptional function, algorithm, Decision-making project evaluation: grown directly from seeds according to essence and produce the sensing that improvement plan decision chip should have
Function and sensing standard value carry out being applicable in algorithm assessment and determine should have according to chip according to sensing function and sensing standard value
Standby sensing function completes the design of chip decision making function with sensing standard value, applicable algorithm；
Edge calculations degree determines and chip functions design: according to chip decision making function and reporting grade separation for repairment, it is resolved that edge meter
Calculation degree, and complete chip functions design；
Chip design, chip structure are laid out, examine after design and simulation: it is designed according to chip functions and utilizes software, progress algorithm,
Circuit design and emulation, and complete each layout layer of circuit；
Actual circuit simulation run test: actual circuit simulation run test is carried out using synthesis tool or MCU chip；
Flow production and test: each layout layer of combined circuit and actual circuit simulation run test result carry out flow production, complete
It is tested at the operating of flow.
Chip design scheme that above scheme provides complete set, having Technology Correlation, it is real according to the waste greatly of scene eight
Border situation successively confirms that chip should have function, should be applicable in algorithm, circuit layout etc. information, is conducive to elastic realization
The design of chip, specification chip design cycle, so that more relevant property of each design phase, but also chip needle with higher
To property, the place of finished product AI boundary operation chip is reduced using difficulty, it is important that reduce edge fortune to a certain extent
The design programming difficulty for calculating programming personnel improves the desin speed of subsequent edges operation design；Expert database is provided to be used for
The design self-criticism and verification of each link, the data feasibility of links, further mentions in the chip design scheme that is conducive to check on
High desin speed and the reliability for increasing the finished product AI boundary operation chip that final design goes out, it is ensured that final product quality.
Further, further includes: result of design, which restores, to be updated: the design specification for the flow that operating is successfully tested is stored in special
Family's database, the design specification include at least the sensing function and sensing that corresponding production line scene eight waste situation greatly is determined
Standard value is applicable in algorithm and each layout layer of circuit.Result of design is saved into expert database, is conducive to subsequent chip design
In the process, dimension applies realization, improves chip desin speed and serviceability.
Further, self-criticism utilizes expert's number with simulation steps after the chip design, chip structure layout, design
It checks and evades with IP to complete each layout layer of circuit according to the critical size that library data carry out circuit layout.
Further, self-criticism is utilized with simulation steps after the chip design, chip structure layout, design
Microwind, Verilog, Debussy, C++, MatLab, Python software carry out algorithm, circuit design and emulation.
Further, the edge calculations degree determine and chip functions design procedure in, chip functions design includes setting
Meter chip structure bottom is CMOS matrix arrangement, designs algorithm used in edge calculations and reaches using metal is online.
Further, in the actual circuit simulation run testing procedure, Design compiler, FPGA are utilized
Complier, FPGA or other MCU chips carry out actual circuit simulation run test.
It further, further include carrying out electricity using expert database data in the actual circuit simulation run testing procedure
Gas characteristic, mechanical property, the assessment judgement and confirmation of functional characteristic and regulation characteristic.It is advantageously ensured that the chip of final finished
Practicability.
Further, the Internet of Things perceptional function, algorithm, in Decision-making project evaluation step, according to expert database number
Accordingly and sensing function carries out being applicable in algorithm assessment and decision with sensing standard value.
Further, in the waste verification step greatly of production line scene eight further include: according to expert database data and
What is confirmed reports grade separation for repairment, generates the correction strategy for corresponding to and each reporting grade for repairment.
Further, the expert database data include:
Raw material and product is physical and chemically data；
Physical security data table data；
Relevant laws and regulations prescribed requirement, at present the circuit layout critical size limitation of available IP；
The technical characteristic of raw material and product in this procedure for producing；
Order status and default rational inventory scheme and quantity；
PDCA method, man-machine material method is around-France, DEMAC method, management program method, risk assessment method decision principle and program；
Affine transformation evolves, is fuzzy, gene, annealing, heredity, population, ant colony and trigonometric function algorithm application rule；
The information input that line management needs.
After adopting the above technical scheme, the beneficial effects of the present invention are: core providing complete set, having Technology Correlation
Piece design scheme successively confirms that chip should have function, should be applicable in algorithm, electricity according to the waste actual conditions greatly of scene eight
Road layout etc. information is conducive to the design that elasticity realizes chip, specification chip design cycle, so that each design phase more has
Relevance, but also chip specific aim with higher, difficulty is applied in the place for reducing finished product AI boundary operation chip, important
, the design programming difficulty of boundary operation programming personnel is reduced to a certain extent, improves subsequent edges operation design
Desin speed；Design self-criticism and verification of the expert database for each link are provided, in the chip design scheme that is conducive to check on
The data feasibility of links further increases desin speed and increases the finished product AI boundary operation chip that final design goes out
Reliability, it is ensured that final product quality is reached essence and is grown directly from seeds the edge calculations of production really by effectively grasp production line eight big wastes
Chip design object.
Detailed description of the invention
In order to illustrate the embodiments of the present invention more clearly or the technical solution of the prior art, attached drawing are as follows:
Fig. 1 is a kind of elastic implementation method flow chart of AI boundary operation chip design provided by the invention；
Fig. 2 is the elastic implementation method flow chart of preferred AI boundary operation chip design provided by the invention；
Fig. 3 is AI boundary operation chip structure bottom schematic diagram provided by the invention；
Fig. 4 is the encapsulating structure sectional view of AI boundary operation chip provided by the invention.
Specific embodiment
Following is a specific embodiment of the present invention in conjunction with the accompanying drawings, technical scheme of the present invention will be further described,
However, the present invention is not limited to these examples.
As shown in Figure 1, the present embodiment provides a kind of elastic implementation method of AI boundary operation chip design, including following step
It is rapid:
S10: verification assessment, confirmation production line scene the waste confirmation greatly of production line scene eight: are carried out according to expert database data
Management waste, bad amendment waste, the excessive waste of manufacture, the superfluous waste of processing, carry waste, inventory excessively wastes, waits
Waste and movement waste in any one or more waste the case where with report grade separation for repairment；
Data in expert database are the data standard library for being needed and being established according to line management in advance, are designed as chip
Self-criticism and verification in the process；It include to identify, report grade separation for repairment etc. data eight big the case where wasting.
Confirmation waste the case where institute by means be generally scene expert or scene be arranged similar expert
Robot or other sensory perceptual systems, such as image recognition system；
Detailed eight include: the case where waste greatly
Line management waste: including personnel amount/type, equipment usage quantity/type, materials'use quantity/type, method
With correctness, whether meet situations such as such as production line environment moving-wire arrangement；
Bad amendment waste: situations such as including shape/quantity/weight/order degree of conformity/color of raw material and product；
The excessive waste of manufacture: the quantity degree of conformity situation comprising raw material and product；
Situations such as processing is superfluous to be wasted: the usage quantity comprising raw material and product, accessory/equipment storage capacity, amount of waste material；
It carries waste: comprising transporting moving-wire reasonability, carrying quantity, carry number, carrying personnel's number, the feelings such as handbarrow number
Condition；
Situations such as inventory excessively wastes: the stack height comprising raw material and product, width, quantity, weight；
It waits waste: including production line movement speed, transporting velocity, personnel's movement speed, the feelings such as raw material and product movement speed
Condition；
Movement waste: include personnel's movement speed/posture, mechanical equipment speed/state, tool equipment speed/state, raw material and
Situations such as speed/state of product.
According to scene situation by expert database verification of data and assess the scene need to confirm which kind of or which kind
The case where waste；After confirmation waste situation, each classification waste situation is carried out reporting tier definition for repairment；
The detailed grade separation that reports for repairment includes: to be obtained from expert database data corresponding first according to different waste situations
Beginning tier definition can will generally report grade for repairment and be divided into 3 grades of even more grades, and 3 grades include alerting immediately and correcting, warn immediately
Show and can arrange time complexity curve and do not warn only to record；
It is preferred that expert database have update calculation ability, can first set grade above, then via do not stop operation with
Calculation does not stop to correct to set out and most reasonably reports grade for repairment for chip design use, eight big wastes of production line is allowed to happen
To minimum state.
S20: essence grow directly from seeds produces improvement plan foundation: to waste the case where with report expert database data in conjunction with grade separation for repairment
Around-France PDCA method, man-machine material method, DEMAC method, management program method and risk assessment are utilized after carrying out verification and legitimacy assessment
Any one or more method in method, which establishes essence and grows directly from seeds, produces improvement plan；
Specifically, different waste situations is had nothing in common with each other with the prioritization scheme that grade is applicable in is reported for repairment, and only choosing is to correct
With the good prioritization scheme of property, certain improvement of production could be ensured；By taking inventory excessively wastes as an example, when the raw material judged or
When product can sufficiently obtain the technology evaluation information of quantization, can directly using preceding four kinds of methods obtain optimization raw material and product it
Stack height, width, quantity, the chips design value such as weight recycle risk assessment method to obtain when can not obtain optimization.
S30: Internet of Things perceptional function, algorithm, Decision-making project evaluation: grown directly from seeds according to essence produce improvement plan determine chip should have
Standby sensing function and sensing standard value is carried out being applicable in algorithm assessment and be determined, foundation according to sensing function and sensing standard value
The sensing function and sense standard value, be applicable in algorithm completion chip decision making function design that chip should have；
Specifically, Internet of Things perceptional function is assessed: being grown directly from seeds according to essence and is produced the optimization core of corresponding corresponding waste situation in improvement plan
Piece design value determines sensing function and sense standard value that chip should have, and each inducing function has the sensing standard of oneself
Value；Such as inventory excessively wastes in situation, needs the sensing function comprising obtaining to raw material and shape of product, corresponding sensing mark
Quasi- value is grown directly from seeds for essence produces raw material and shape of product optimization design value in improvement plan；It is right also comprising the sensing function for determining quantity
The sensing standard value answered is that smart grow directly from seeds produces the optimization design value of quantity in improvement plan, and so on；
Algorithm evaluation: it carries out being applicable in algorithm assessment and determine, mesh according to sensing function and sensing standard value in conjunction with expert database
Preceding every kind of algorithm all more or less will appear the restricted of application, can not a kind of algorithm whole patience, for different applicabilities,
Also there is specific achievement to find at present, such as fuzzy algorithmic approach is useful in the high control of robustness, such as Motor Control, and loses
Propagation algorithm then has effect of optimization very much, but time-consuming very big, so not being suitble to a large amount of operations very much if system is too big；
Decision-making project evaluation: the sensing function that should have according to chip completes chip decision with sensing standard value, applicable algorithm
Functional Design establishes the sensing function that should have, sensing standard value and the one-to-one decision scheme of applicable algorithm；Due to
The data that the realization of some sensing functions needs that multiple functional units is combined directly to sense acquisition can just obtain after carrying out comprehensive operation
, it is preferred that further include being assembled corresponding to the functional unit that should have sensing function item in decision scheme, such as production line
The sensing of environment moving-wire arrangement, needs by two functional units of GPS and magnetic induction, the sensing of production line movement speed needs
By three speedometer, vibrometer and accelerometer functional units, so as to the guidance designed as chip functions, certain function
Unit allocation plan can also row considers and designs again in chip functions design procedure.
S40: edge calculations degree determines and chip functions design: according to chip decision making function and reporting grade separation for repairment, certainly
Determine edge calculations degree, and completes chip functions design；
Specifically, by taking inventory excessively wastes as an example, it is assumed that sensing function is the length, width and height for sensing storehouse, sense in chip decision making function
Survey standard value is 2M*1M*1M, reports grade for repairment and be divided into following 5 grades: when alerting and correct immediately, alerting and can arrange immediately
Between amendment, warn and can arrange slower time complexity curve immediately, warning can be postponed and interim can arrange amendment and be not necessary to warn
Only record；
Then edge calculations degree can design as follows:
When being > 4M*2M*2M using being applicable in algorithm to obtain field condition value, that is, when being more than 2 times of conventional criteria volume, by it
Be classified as immediately warning and it is modified report grade for repairment, needing alert management, person is reduced immediately；
It is > 2M*2M*2M when obtaining field condition value, that is, when being more than 1.6 times of conventional criteria volume, is classified as warning immediately
Time complexity curve is accused and can be arranged, needs alert management person immediately, but time complexity curve can be arranged to be reduced；
It is > 2M*1M*2M when obtaining field condition value, that is, when being more than 1.3 times of conventional criteria volume, is classified as warning immediately
Show and can arrange slower time complexity curve, manager need to be warned immediately, but slower time complexity curve can be arranged to be reduced；Generally, it warns
Grade lower than warning, represent warning Xiang Kehou in warning item processing；
It is > 2M*1M*1.2M when obtaining field condition value, that is, when being more than 1.1 times of conventional criteria volume, is classified as to prolong
Warning late and interim amendment can be arranged, can postpone to warn manager, and interim can arrange amendment decrement；
Remaining case values is then classified as and is not necessary to warning only record, i.e., warns without issuing, only record field condition value
；
Finished chip after the completion of design also can go out optimize solution according to the carry out judgment value optimization of sensing data, iteration, thus real
The real production management of the essence of existing producing line；
I.e. by the acceptable value of 1 ~ 5 any one grade of manager's setting, furthermore expert database can also sum up one in advance
If initial value, that is, standard value is sensed, starts gradually to obtain optimization solution using algorithm according to this initial value, then do not stop later to repair
Positive 1 ~ 5 setting judgment value, such manager just can be carried out management judgment or control measure.Such as: when certain production line, inventory is set
When being set to 2M*1M*1M, but can only store 1M*1M*1M just according to fire code requirements fire carrying capacity will not can not control when fire
The intensity of a fire processed, at this time prioritization scheme sensing standard value just will become 1M*1M*1M storage amount, but if needed because of business,
Manager increases automation fire-fighting equipment, and pyric factor reduces, and perhaps can at this time be relaxed according to assessment as 2M*2M*2M work
For the sensing standard value of optimization, the rest may be inferred for the setting of remaining waste situation edge calculations degree；
Chip functions design: the algorithm according to used by boundary operation and edge calculations degree determine the function list of chip
Member assembles, such as: determine that quantity/type etc. applies GPS, approach application correctness, which differentiates, applies magnetic induction, and production line environment is dynamic
Line arrangement applies GPS and magnetic induction, and the shape sensing of raw material and product, limits of application switch, GPS and poidometer etc. are former
The usage quantity of material and product sensing, application traffic meter, counter and timer, production line movement speed sensing, using speed
Meter, vibrometer, accelerometer etc. are spent, in addition, image identification is then that can all apply, but action response speed is relatively
Slowly, it should be noted that the content of chip functions design is not only limited to above content, may also include its for realizing chip functions
He pays attention to item.
S50: self-criticism and simulation after chip design, chip structure layout, design: utilizing software according to chip functions design,
Algorithm, circuit design and emulation are carried out, and completes each layout layer of circuit；
By taking quasi- affine transformation evolution algorithm as an example, the design of software emulation subprogram is as follows:
S60: actual circuit simulation run test actual circuit simulation run test: is carried out using synthesis tool or MCU chip；
S70: flow production and test: each layout layer of combined circuit and actual circuit simulation run test result carry out flow system
Make, completes the operating test of flow.Flow production herein is to refer to sample making, rather than produce in batches, in this step, according to
Each layout layer of circuit completes circuit design and burning enters used corresponding algorithm when emulating and forms integrated chip sample, and
Chip sample is arranged and carries out operating test at production line scene, between ideal effect when checking actual measurement situation and emulation
Difference size.
Chip design scheme that above scheme provides complete set, having Technology Correlation, it is real according to the waste greatly of scene eight
Border situation successively confirms that chip should have function, should be applicable in algorithm, circuit layout etc. information, is conducive to elastic realization
The design of chip, specification chip design cycle, so that more relevant property of each design phase, but also chip needle with higher
To property, the place of finished product AI boundary operation chip is reduced using difficulty；
Importantly, reducing the design programming difficulty of boundary operation programming personnel to a certain extent, subsequent edges are improved
The desin speed of operation design；Design self-criticism and verification of the expert database for each link are provided, the chip that is conducive to check on is set
The data feasibility of links in meter scheme further increases desin speed and increases the edge finished product AI that final design goes out
The reliability of operation chip, it is ensured that final product quality.
Further, as shown in Figure 2, further includes:
S80: result of design, which restores, to be updated: the design specification for the flow that operating is successfully tested is stored in expert database, described to set
Meter specification includes at least the sensing function that corresponding production line scene eight waste situation greatly is determined and sensing standard value, is applicable in calculation
Method and each layout layer of circuit.Result of design is saved into expert database, is conducive in subsequent chip design process, dimension
Apply realization, improve chip desin speed and serviceability.
After chip design, chip structure layout, design self-criticism in simulation steps, using expert database data into
The critical size verification of row circuit layout is evaded with IP to complete each layout layer of circuit.I.e. expert database has pre-saved at present
It can get the circuit layout critical size limitation of IP, to carry out dodging for layout dimension.
After chip design, chip structure layout, design self-criticism in simulation steps, using Microwind,
Verilog, Debussy, C++, MatLab, Python software carry out algorithm, circuit design and emulation.
In the edge calculations degree decision and chip functions design procedure, chip functions design includes design chips structure
Bottom is CMOS matrix arrangement, and chip structure fabric figure is as shown in figure 3, algorithm used in design edge calculations utilizes
Metal is online to be reached, specifically the encapsulating structure sectional view of chip, as shown in Figure 4.
In the actual circuit simulation run testing procedure, Design compiler, FPGA complier, FPGA are utilized
Or other MCU chips carry out actual circuit simulation run test.
It further include carrying out electrical characteristic, machine using expert database data in the actual circuit simulation run testing procedure
Tool characteristic, the assessment judgement and confirmation of functional characteristic and regulation characteristic.It is advantageously ensured that the practicability of the chip of final finished.
The Internet of Things perceptional function, algorithm, in Decision-making project evaluation step, according to expert database data and sensing
Function and sensing standard value are carried out being applicable in algorithm assessment and be determined.
Generally, quasi- affine transformation evolution algorithm can be used for optimization discontinuous, noise, as the time changes and ask
Topic；Fuzzy algorithmic approach is suitble to industrial production control, household appliances control；Genetic algorithm is suitable for solving Global Optimal Problem, such as
Solve the problems, such as timetable Arrangement Problem or Other Engineering；Annealing algorithm is suitable for the random fluctuation by a solution in search space
The method for finding Best Point；Particle swarm algorithm is abbreviated as PSO, is a kind of very promising neural network algorithm, PSO speed ratio
Very fast and available relatively good result, suitable for such as using the data set analysis of classification problem；Neural network algorithm is suitable
Together in the process made inferences according to logic rules, information is first melted into concept, and symbolically by it, then according to symbol
Operation carries out reasoning from logic in a serial mode, this process can be write as serial instruction, and computer is allowed to execute, however straight
The thinking for the property seen is by the informix of distributed storage, the result is that the method had an idea or solved the problems, such as suddenly；
Ant group algorithm is suitable for real variable dynamics problem, stochastic problem, multiple target realization arranged side by side；Trigonometric function algorithm is suitable for
The continuous problem of plane or three-dimensional space shift in position optimization solve, therefore according to sensing function with sense standard value carry out it is applicable
The judgement of algorithm be very it is necessary to, improve chip elasticity design.
Preferably, the grade of reporting for repairment is divided into 5 grades, comprising: when alerting and correct immediately, alerting and can arrange immediately
Between amendment, warn and can arrange slower time complexity curve immediately, warning can be postponed and interim arrange amendment and be not necessary to warning only
Record.That is the AI boundary operation chip of finished product finally only has warning/warning effect, and final producing line adjustment and control are also
It is to need manager works；
In order to further increase the excellent performance of this programme final finished AI boundary operation chip, in addition to confirm in step S10
It the case where waste and reports for repairment outside grade, further includes according to expert database data and what is confirm report grade separation, generation pair for repairment
It should be controlled by the reasonability that AI boundary operation chip carries out production line scene, i.e. root in each correction strategy for reporting grade for repairment
According to expert database data carry out data calculation, after calculation generate control signal give by control object (such as motor adjust revolving speed,
And then conveyer belt will change speed, wait waste that will lower, but on condition that chip needs can find out operation still just
It under normal state, and is not only one and tunes up speed with hiding)；
Such as: by taking production line intermediate semi-finished products inventory as an example, when after station there is operating speed and reduce situation, if setting out in advance to make arrangements still constant speed
Production, that intermediate semi-finished products buffer pool storage will continue to increase, and speed of standing after generally utilizing goes control next stop speed, but
Be the scheme optimized should be determined with intermediate semi-finished products buffer pool storage, be best suitable for actual management demand in this way, no
Increase operation puzzlement, because, if inventory space can not increase, just necessarily there is additional carrying waste once increasing in inventory thus,
Side can only first be moved to put, moved back again when station will use after waiting.
Further, the expert database data include:
Eight big waste resolution principles；
Raw material and product is physical and chemically data；
Physical security data table data；
Relevant laws and regulations prescribed requirement, at present the circuit layout critical size limitation of available IP；
The technical characteristic of raw material and product in this procedure for producing；
Order status and default rational inventory scheme and quantity；
PDCA method, man-machine material method is around-France, DEMAC method, management program method, risk assessment method decision principle and program；
Affine transformation evolves, is fuzzy, gene, annealing, heredity, population, ant colony and trigonometric function algorithm application rule；
The information input etc. that other line managements need.
Current eight big waste resolution principles, essence grow directly from seeds produce improvement plan, Internet of Things framework, quasi- affine transformation evolution algorithm/
Fuzzy/gene/annealing/heredity/population/ant colony/trigonometric function (distance d and angle, θ) closed model scheduling algorithm,
Microwind/Verilog/Debussy/Design compiler/FPGA complier/C++/ MatLab/ Python、
Chip stream blade technolgy etc. is very mature and is used for the edge AI fortune using frequently, these technologies are integrated and become above scheme
The design for calculating chip has no technical difficulty.
Specific embodiment described herein is only an example for the spirit of the invention.The neck of technology belonging to the present invention
The technical staff in domain can make various modifications or additions to the described embodiments or replace by a similar method
In generation, however, it does not deviate from the spirit of the invention or beyond the scope of the appended claims.
Claims (10)
1. a kind of elastic implementation method of AI boundary operation chip design, which comprises the following steps:
The waste confirmation greatly of production line scene eight: verification assessment is carried out according to expert database data, confirms the pipe at production line scene
Excessively waste, the superfluous waste of processing, carrying waste, inventory excessively waste, wait waste for reason waste, bad amendment waste, manufacture
And it acts the case where any one or more waste in wasting and reports grade separation for repairment；
Essence, which is grown directly from seeds, produces improvement plan foundation: the case where to waste with report expert database data in conjunction with grade separation for repairment and check
With after legitimacy assessment using PDCA method, man-machine material method is around-France, appointing in DEMAC method, management program method and risk assessment method
One or more methods of anticipating, which establish essence and grow directly from seeds, produces improvement plan；
Internet of Things perceptional function, algorithm, Decision-making project evaluation: grown directly from seeds according to essence and produce the sensing that improvement plan decision chip should have
Function and sensing standard value carry out being applicable in algorithm assessment and determine should have according to chip according to sensing function and sensing standard value
Standby sensing function completes the design of chip decision making function with sensing standard value, applicable algorithm；
Edge calculations degree determines and chip functions design: according to chip decision making function and reporting grade separation for repairment, it is resolved that edge meter
Calculation degree, and complete chip functions design；
Chip design, chip structure are laid out, examine after design and simulation: it is designed according to chip functions and utilizes software, progress algorithm,
Circuit design and emulation, and complete each layout layer of circuit；
Actual circuit simulation run test: actual circuit simulation run test is carried out using synthesis tool or MCU chip；
Flow production and test: each layout layer of combined circuit and actual circuit simulation run test result carry out flow production, complete
It is tested at the operating of flow.
2. a kind of elastic implementation method of AI boundary operation chip design according to claim 1, which is characterized in that also wrap
Include: result of design, which restores, to be updated: the design specification for the flow that operating is successfully tested is stored in expert database, the design specification
The sensing function and sensing standard value that are determined including at least corresponding production line scene eight waste situation greatly, be applicable in algorithm and
Each layout layer of circuit.
3. a kind of elastic implementation method of AI boundary operation chip design according to claim 1, which is characterized in that described
Chip design, chip structure layout, design after self-criticism with simulation steps in, using expert database data carry out circuit layout it
Critical size verification is evaded with IP to complete each layout layer of circuit.
4. a kind of elastic implementation method of AI boundary operation chip design according to claim 1, which is characterized in that described
Self-criticism utilizes Microwind, Verilog, Debussy, C+ with simulation steps after chip design, chip structure layout, design
+, MatLab, Python software carries out algorithm, circuit design and emulation.
5. a kind of elastic implementation method of AI boundary operation chip design according to claim 1, which is characterized in that described
In the decision of edge calculations degree and chip functions design procedure, chip functions design includes that design chips structure bottom is CMOS square
Configuration arrangement designs algorithm used in edge calculations and reaches using metal is online.
6. a kind of elastic implementation method of AI boundary operation chip design according to claim 1, which is characterized in that described
In actual circuit simulation run testing procedure, using Design compiler, FPGA complier, FPGA or other MCU cores
Piece carries out actual circuit simulation run test.
7. a kind of elastic implementation method of AI boundary operation chip design according to claim 6, which is characterized in that described
It further include carrying out electrical characteristic, mechanical property, function using expert database data in actual circuit simulation run testing procedure
The assessment judgement and confirmation of characteristic and regulation characteristic.
8. a kind of elastic implementation method of AI boundary operation chip design according to claim 1, which is characterized in that described
Internet of Things perceptional function, algorithm in Decision-making project evaluation step, are marked according to expert database data and sensing function and sensing
Quasi- value is carried out being applicable in algorithm assessment and be determined.
9. a kind of elastic implementation method of AI boundary operation chip design according to claim 1, which is characterized in that described
In the waste verification step greatly of production line scene eight further include: grade separation is reported for repairment with what is confirmed according to expert database data,
Generate the correction strategy for corresponding to and each reporting grade for repairment.
10. a kind of elastic implementation method of -9 any AI boundary operation chip designs, feature exist according to claim 1
In the expert database data include:
Raw material and product is physical and chemically data；
Physical security data table data；
Relevant laws and regulations prescribed requirement, at present the circuit layout critical size limitation of available IP；
The technical characteristic of raw material and product in this procedure for producing；
Order status and default rational inventory scheme and quantity；
PDCA method, man-machine material method is around-France, DEMAC method, management program method, risk assessment method decision principle and program；
Affine transformation evolves, is fuzzy, gene, annealing, heredity, population, ant colony and trigonometric function algorithm application rule；
The information input that line management needs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910763925.4A CN110489861A (en) | 2019-08-19 | 2019-08-19 | A kind of elastic implementation method of AI boundary operation chip design |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910763925.4A CN110489861A (en) | 2019-08-19 | 2019-08-19 | A kind of elastic implementation method of AI boundary operation chip design |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110489861A true CN110489861A (en) | 2019-11-22 |
Family
ID=68552158
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910763925.4A Withdrawn CN110489861A (en) | 2019-08-19 | 2019-08-19 | A kind of elastic implementation method of AI boundary operation chip design |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110489861A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111158967A (en) * | 2019-12-31 | 2020-05-15 | 北京百度网讯科技有限公司 | Artificial intelligence chip testing method, device, equipment and storage medium |
-
2019
- 2019-08-19 CN CN201910763925.4A patent/CN110489861A/en not_active Withdrawn
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111158967A (en) * | 2019-12-31 | 2020-05-15 | 北京百度网讯科技有限公司 | Artificial intelligence chip testing method, device, equipment and storage medium |
| CN111158967B (en) * | 2019-12-31 | 2021-06-08 | 北京百度网讯科技有限公司 | Artificial intelligence chip testing method, device, equipment and storage medium |
| US11714128B2 (en) | 2019-12-31 | 2023-08-01 | Kunlunxin Technology (Beijing) Company Limited | Method and apparatus for testing artificial intelligence chip, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Model construction of planning and scheduling system based on digital twin | |
| Wang et al. | On Petri net implementation of proactive resilient holistic supply chain networks | |
| CN201374004Y (en) | Industrial automation system and industrial controller system | |
| Leng et al. | Cloud-edge orchestration-based bi-level autonomous process control for mass individualization of rapid printed circuit boards prototyping services | |
| CN108320278A (en) | Product defects detect localization method, device, equipment and computer-readable medium | |
| CN111967738B (en) | Risk early warning method, system and medium of power grid information energy fusion system | |
| CN117434912B (en) | Method and system for monitoring production environment of non-woven fabric product | |
| CN116882038A (en) | Electromechanical construction method and system based on BIM technology | |
| CN110489861A (en) | A kind of elastic implementation method of AI boundary operation chip design | |
| CN110378586A (en) | Defect of transformer equipment method for early warning and system based on Dynamic Closed Loop information management | |
| JP2010086219A (en) | Information processing apparatus for maintaining remote machine, program and method | |
| Zhang et al. | Artificial neural network for violation analysis | |
| CN108460294A (en) | A kind of decision proof type block chain algorithm and its physical structure | |
| Sun et al. | Digital twin for energy-efficient integrated process planning and scheduling | |
| Samigulina et al. | Biologically Inspired Unified Artificial Immune System for Industrial Equipment Diagnostic | |
| Tsai et al. | A Fuzzy Nonlinear Programming Approach for Optimizing the Performance of a Four‐Objective Fluctuation Smoothing Rule in a Wafer Fabrication Factory | |
| CN110011971A (en) | A kind of manual configuration method of network security policy | |
| CN109670550A (en) | A kind of distribution terminal maintenance decision method and apparatus | |
| Chang et al. | A fuzzy neural network for the flow time estimation in a semiconductor manufacturing factory | |
| Ma et al. | Application of BP neural networks in garment pattern design system | |
| Chang et al. | Probe of the replenishment strategy and grouping rule in the semiconductor industry | |
| CN112488456B (en) | Digital data modeling method for power equipment | |
| Li et al. | Construction Technology Safety Management under the Background of BIM and Information System Modeling | |
| Zhou et al. | Intelligent decision method for supplier selection | |
| Rong | Analysis of the comprehensive evaluation model of enterprise technological innovation ability based on improved genetic algorithm |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20191122 |
|
| WW01 | Invention patent application withdrawn after publication |