industry & distribution - hitachi · 2020. 11. 26. · hitachi review vol. 68, no. 2 152–153...
TRANSCRIPT
-
24.
Industry & Distribution
Hitachi’s Research & Development Group has drawn on the analysis expertise it has accu-mulated over the past 10 years to develop new digital solutions with names such as Black Box Breaker (B3) and Clinical Semantic Linker (CSL). B3 is a deep learning technology with medical explanatory capabilities. CSL is a natural language processing technology spe-cialized for text analysis in the medical fi eld. Th ese technologies have been used in proofs of concept (PoCs) with clients in various business areas, and have demonstrated benefi ts. Th ey form the core of a project begun in 2017 called Hitachi Digital Solutions for Pharma that provides services for improving the effi ciency of client operations. Th ese services include solutions for patient stratifi cation, assisting clinical trial plans, and target searches.
Th e pharmaceutical industry is facing increasingly demanding conditions caused by challenges such as falling drug prices due to rising healthcare costs, and rising research and development (R&D) costs caused by increasingly advanced medical technologies (individ-ualized healthcare and regenerative medicine). Th ese demanding conditions are forcing the industry to carry out more accurate business value assessments. Countries around the world are also introducing healthcare economic assessments to weigh the costs and ben-efi ts of medical technologies during decision-making on healthcare policy in areas such as medical technology insurance reimbursement and price setting. Th e presentation of scientifi cally corroborated evidence is being demanded as a result.
Hitachi Digital Solutions for the Pharmaceutical Industry1
Providing services targeting new management/development strategies and designed to improve the operation efficiency and accuracy of the drug discovery process
straight-through from upstream to downstream
Client
In-house data
• Data cleansing • Security• AI analysis engines
Cloud services
AI
Chemical compound library
Sales data
Open data Real-world data
Electronic patient charts
Medical billing statements for health
insurance claimsOther
Medical papersChemical
compoundsProteinsOther
SalesmarketingSafety
managementPostmar-
ketingsurveillance
Manufac-turing
Clinicaltesting
Platform-providingservices
Management/development strategies
Operation support services
Non-clinicaltesting
Basicresearch
Analyzing data Storing data
Transforming data into value
Data gathering
Business value assessments● Healthcare economic
assessments● Sales forecasts
1 Organization of Hitachi Digital Solutions for Pharma
AI: artifi cial intelligence
-
Hitachi Review Vol. 68, No. 2 152–153
Industry, Distribution, W
aterIndustry &
Distribution
25.
Seeking to help clients grow their businesses and improve patient quality of life (QoL), Hitachi is responding to the demands of the current environment by releasing develop-ment strategy solutions, adding simulation services for business viability assessments and healthcare economic assessments, and providing services designed to improve the opera-tion effi ciency and accuracy of the drug discovery process straight-through from upstream to downstream.
Hitachi will continue to expand the application of this project into the area of medical devices and the entire healthcare industry.
Hitachi AI Technology/Machine Learning Constraint Programming (MLCP) is a plan optimization service that provides a new planning system for deriving more optimum solutions by recreating plans incorporating the implicit knowledge of experts.
Digitalization technology using artifi cial intelligence (AI) enables machine learn-ing-based system growth along with previously impossible conditional mitigation resem-bling the work of experts. Th e design approach used is diff erent from conventional system confi gurations, enabling constraint conditions and plan patterns to be revealed that would previously have been overlooked.
Th e service surpasses the abilities of package products, with the potential to connect diff erent processes, diff erent plans, and people with plans. It is being used in a wide range of areas such as the steel, non-ferrous metal, oil, and food industries. In addition to pro-duction plans, it also supports a wide range of plan types such as vehicle dispatch plans and staff plans.
Using AI to Help Optimize Planning: Hitachi AI Technology/MLCP2
Understanding of business process
and contentA B
Constraint conditions
Expert planner
Utilize the plan(after some adjustment
and confirmation if necessary)
Expert’s planning patterns
Expert planner
Plan (historical)
data
Indicated data
Reproduce expert’s plan with big data analysis technologyand promptly make an optimized plan
Machinelearning
MLCP*
Mathematical programming
MLCP: Machine Learning Constraint Programming* A new constraint programming technology in collaboration with machine learning and mathematical programming techniques
2 MLCP plan optimization service
-
26.
Pastdata sets
Real-timedata sets Quality AI
Energy-saving AI
Precursor AI
Productivity AI
Online(closed network)
Offline data gathering
Data
Expert interviewsAcquired data policies
Continuous dataaccumulation AI capsule sets
Data mart
Data integration
Datacleansing
AI capsulization
Data analysis/AI capsulization service
Data analysis logic development/revision
Analysis result report
feedback
AI-based continuous analysis
Sensors
Heat sources
Work history
Failure history
Sensors
Heat sources
Work history
Failure history
(1)
Dat
a an
alys
is(2
) A
I cap
suliz
atio
n
3 AI servitization
Hitachi pharmaceutical manufacturing execution system (HITPHAMS) is a manufactur-ing management system for pharmaceutical manufacturers that Hitachi provides to clients. It conforms to the Japanese Good Manufacturing Practice (J-GMP)*1 regulations. Version 4.0 is the latest version. It has been refi ned for easier installation and use by users. Th e previous screen design has been replaced with a design that incorporates updated user experience (UX)*2 expertise.
HITPHAMS installation is done by the conference room pilot (CRP)*3 method. Client masters are used while checking actual operation, and the system is installed with minimal customization. Th is method enables smooth and rapid system installation while placing few demands on the client. Th e use of this method has also enabled a design that users can operate intuitively and easily. Workfl ows can be defi ned for business operations, while dis-plays such as menus, notices, and favorites on the home screen enable intuitive checking of
HITPHAMS Version 4.0Refi ned for Easy-to-Use System4
Demand for the use of manufacturing data and AI to ensure quality stability and reduce defects has recently been growing at manufacturing industry production sites. To meet this demand, Hitachi has been providing a data analysis service run by data scientists that is designed to improve product quality.
To enable clients to use the analysis logic created by this data analysis service, Hitachi has started to provide a service that creates AI capsules from this logic. Th ese capsules enable clients to continuously analyze manufacturing data created in real time, enabling faster problem-solving for quality improvement.
In 2018, Hitachi began off ering an AI servitization designed for quality improvement to manufacturing industry clients with process-based production equipment. Hitachi also plans to start off ering the service to clients in the pharmaceutical, chemical, and other industries.
Using Manufacturing Data for Quality Improvement Analysis and AI Servitization3
-
Hitachi Review Vol. 68, No. 2 154–155
Industry, Distribution, W
aterIndustry &
Distribution
27.
tasks to be done and information of interest. Screens displaying work procedures are shown as fl owchart displays, creating a design enabling an intuitive understanding of the tasks to be done.*1 Manufacturing management and quality control regulations for Japan’s pharmaceutical and “quasi
drug” (medicated products) industries.*2 A term used in system design to refer to the user’s emotions and attitudes about using a particular
product, system, or service.*3 Th e work of checking that business operations can be implemented while checking standard functions
on the actual equipment.
Screen design enabling intuitive ease of understanding by user
Flow of system installation by CRP method
1st CRP2nd CRP
3rd CRP
To installation of completed system
Operating environment
Master creation/revision
Operating environment
Master creation/revision
Operating environment
Checking the system using the actual equipment makes it easy for the client to visualize the operation, and enables minimal customization
Master creation/revision
Brush-up
The favorites display shows the user’s screens of
interest
Notification of process to be
worked on
4 HITPHAMS Version 4.0 installation method and screen illustration
Japan’s declining birthrate and aging population are expected to shrink the country’s work-ing-age population to just one-third of its current size by the 2050s, dropping from around 76 million today to just 25 million. Workplaces are responding by making increased use of foreign and inexperienced workers, but quality and work effi ciency are declining. Safety and other issues are also arising as concerns. Examples of workplace problems are listed below.(1) Details of work procedures and lessons from accident case studies are not being shared, resulting in the same accidents recurring.(2) Work procedures and safety precautions are being communicated to foreign workers when needed, but failure to communicate explanations eff ectively leads to mistakes.(3) Th e complexity of managing training meeting attendance and records indicating who has viewed work procedures and case reports is resulting in some workers missing required information, and making it impossible to provide education tailored to the needs of each worker.
Hitachi’s work safety education cloud service provides conventional headquarters-led safety education while enabling workers to share and accumulate safety skills and exper-tise in a format based on work site participation and communication. Th e service can be expected to provide the benefi ts below for solving the problems above.(1) Videos shot at work sites can be viewed on smartphones and tablets just by uploading them. Th is benefi t makes it easy to share work procedures and accident case studies, and to communicate subtle nuances that are diffi cult to convey verbally.
Work Safety Education Cloud Service5
HITPHAMS: Hitachi pharmaceutical manufacturing execution system CRP: conference room pilot
-
28.
(2) For videos with audio commentaries, multilingual subtitles can be automatically gener-ated and then manually revised. Th is benefi t enables clear communication to foreign work-ers using videos with subtitles translated into natural language they can easily understand.*(3) Each worker’s video viewing history and training meeting attendance can be managed in a unifi ed manner so that their needs can be accommodated when providing information and creating educational plans.
Hitachi will continue to provide assistance for eff ective and effi cient safety education at a variety of workplaces engaging in operations such as production plant work, logistics warehouse work, and local maintenance/inspection work.(Hitachi Industry & Control Solutions, Ltd.)* Th e multilingual subtitle generation function does not guarantee 100% accuracy. It is an auxiliary func-
tion requiring subtitle revision work in each language.
Th e physical security industry in Japan and around the world is experiencing a wave of col-laborative public-private sector eff orts designed to improve security and anti-terrorism measures at locations open to the general public such as event venues, sports stadiums, railway stations, and airports.
Security systems with surveillance cameras have traditionally been used as a way to improve surveillance, but users are now calling for methods to rapidly spot suspicious behavior or fi nd other persons of interest among massive vol-umes of captured video data. But con-ventional technologies such as facial recognition oft en make it diffi cult to
High-speed Person Finder/Tracker Solution6
(The screens shown were created for illustration purposes during development.)
6 Example personal attribute search result display screens
Can be used anywhere, at any time
Automatic video subtitling and subtitle translation Checking viewing/training histories
Easy to shoot, upload, and share
Video manualCase study video
PC/smartphone/tabletViewable anywhere
Videos can be easily shot at work sites using smartphones and tablets, and can be uploaded for sharing
Can be used as digital manual for site
Shooting video
Work manager
Worker
Monitoring
Worker viewing histories can be listed and followed
Voice recognition enables automatic video subtitlingGenerated subtitles can also be translated into multiple languages
English, Chinese, other
Japanese
Automatic subtitle generation
Multilingual subtitle translation
Checking worker video viewing histories
DANGER
Work safety cloud
Work safety cloud
5 Overview of work safety education cloud service
-
Hitachi Review Vol. 68, No. 2 156–157
Industry, Distribution, W
aterIndustry &
Distribution
29.
A recent rise in consumer concern over food safety and security is creating an increasing need for optimum methods of determining room placement when designing food pro-cessing plant layouts. Th ese methods need to reduce life cycle costs (LCC) and incorporate fl ow line planning that maintains productivity and quality. Th ey also need to incorporate the hazard analysis and critical control point (HACCP) approach.
But layout design is done by optimal solution search processes among complex and diverse patterns, and has always required a large number of man-hours and reliance on experienced designers. Hitachi Plant Services has attempted to improve this method by developing an engineering tool that automatically creates food processing plant layouts using AI. An improved genetic algorithm (considered a type of AI) is used as a room com-bination optimization algorithm that standardizes design expertise and forms the tool core.
Th e tool can rapidly optimize conditions such as sanitary zones and fl ow lines of items such as products, workers, and waste. It can create multiple layout plans that have each been quantitatively evaluated. Th e created layouts are nearly equivalent to those of experi-enced human designers, even in terms of key elements such as fl ow line intersections. Th e tool is expected to improve design work effi ciency by about 50% (for Hitachi’s own design work). It has already been used to create about 10 actual layouts. Hitachi is planning to use it in expanding its total engineering work, a business area that provides a comprehensive range of services encompassing everything from upstream consulting to design, construc-tion, and operation/maintenance services.(Hitachi Plant Services Co.,Ltd.)
AI-driven Engineering Tool for Automatic Creation of Food Processing Plant Layouts7
Raw materials
Raw materials
Container washing room
Sorting room Processing room (2)
Container washing room
Processing room (1)
Processing room (4)
Processing room (3)
Sorting roomProcessing room (1)
Processing room (2)
Processing room (3)
Processing room (4)
Products
Products
Coo
king
roo
m
Coo
king
roo
m
Containers
Containers
7 Example comparison of human- and tool-created layouts (pre-made food products)
identify persons of interest in camera footage since the information providing the char-acteristics needed for personal identifi cation is vague, the person is facing away from the camera, or the image is unclear.
To solve these problems, Hitachi has developed a security solution that uses AI driven by multilayer neural network-based machine learning technology (deep learning technol-ogy). Th e technology can identify suspicious behavior in video images using even vague personal or other characteristics, and supports a large number of cameras using high-speed processing. Th is solution will help improve the quality and effi ciency of surveillance work, making the world more safe and secure.(Hitachi Industry & Control Solutions, Ltd.)
-
30.
(1) Creating marking data from CAD data
(4) • Identifying own position using measuring instrument
• Moving, avoiding obstacles and marking as specified by marking data
(3) Measuring/setting site reference core
(5) Marking status monitoring
Example mark
Control terminal
Travel unit
X-Y stage
Laser ranging
3D scanner
Marking unit
Prism target Control unit Wi-Fi*
Wi-Fi Wi-Fi
(2) Sending marking data to robot
CAD
Marking data generation app
Auto-tailing measuring instrument
Automatic marking robot
Control app
Data reading Marking data
8 Automatic marking robot system confi guration
CAD: computer-aided design* See “Trademarks” on page 158.
Hitachi has developed a highly stable optical test environment that enables high-precision optical testing at atmospheric pressure, and is designed for optical inspection applications on board earth observation satellites and astronomical satellites. Th e environment enables major reductions in operation costs and test time relative to conventional vacuum cham-ber-based testing, and can support next-generation large optical systems. It was devel-oped from a joint research project of the Japan Aerospace Exploration Agency (JAXA)
Highly Stable Optical Test Environment Assisting JAXA Space Development9
As part of the eff orts being done to apply information and communication technology (ICT) to construction work, Hitachi has developed an automatic marking robot system that enables a single operator to handle work that previously took multiple skilled workers.
Equipment installation requires an extensive amount of marking work done to provide fl oor indications of the positions at which to install devices such as lines, ducts, and ceil-ing-suspended units. Marking work is done by skilled workers using tools such as mea-suring instruments and inking devices to repeatedly transfer installation positions shown in drawings onto fl oor surfaces. Japan’s improving economy and aging population have recently been exacerbating the country’s skilled worker shortage, creating an urgent need for more effi cient work methods that also require less skill.
Hitachi’s automatic marking robot system is composed of a self-propelled automatic marking robot, a marking data generation app that extracts and creates marking data from computer-aided design (CAD) drawings, an auto-tailing measuring instrument used to recognize the robot’s own position, and a control app. Th e system was used at fi ve air con-ditioning equipment installation sites through the fi rst half of 2018, and demonstrated an error of ±3 mm and marking speed of 3 minutes per point.
Hitachi plans to expand use of the system to sites throughout Japan, while improving its effi ciency by using it for simultaneous marking on multiple related work projects and coordinating it with building information modeling (BIM) applications.(Hitachi Plant Services Co.,Ltd.)
Automatic Marking Robot System for Equipment Installation Work8
-
Hitachi Review Vol. 68, No. 2 158–159
Industry, Distribution, W
aterIndustry &
Distribution
31.
and Hitachi Plant Services. Hitachi is hoping to use it in future large satellite projects and private-sector space indus-try applications.
Th e joint research project with JAXA started in October 2017 with the aim of constructing a highly stable optical test environment. Th e environment was developed by adding humidity and air-fl ow control to a platform of proprietary Hitachi Plant Services technology used for precision temperature control and air conditioning. It has temperature sta-bility of within ±0.01°C and dew point temperature stability of within ±1°C. Hitachi conducted simulated optical inspection measurement using a pro-totype booth created at the Matsudo Works, and demonstrated the equip-ment’s ability to provide measurement accuracy equivalent to a vacuum chamber. Hitachi continued the joint research project in 2018, and is looking into providing support for large test samples.(Hitachi Plant Services Co., Ltd.)
Special air supply port
Special heaterOptical measuring
instrument
Insulating chamber
Intake
Object for measurementVibration removal board
Outside air
Exhaust
Dehumidifier/air conditionerChiller
Airconditioner
9 Environmental test chamber conceptual diagram
Operator fi ne-tuning is a vital requirement for correcting complex sheet steel waviness (shape) when manufacturing sheet steel with cold rolling mills. Operator workload and variations in product quality due to diff erences in skill level are major challenges as a result.
To address these challenges, Hitachi has developed a deep learning-based plant control technology that has been used for shape control on a Sendzimir rolling mill at Shougang Zhixin Qian’an Electromagnetic Material Co., Ltd., a steelworks in China. Data from past operations is used to help the control system automatically learn relationships between equipment operations and work results by itself. Th is approach enables high-precision con-trol, and gives the system an outstanding ability to self-adjust and gain expertise. Th e sys-tem also has a method of reducing abnormal output, enabling learning to continue while preventing the plant from being adversely aff ected by control errors. Th e soft ware has been mounted in a programmable logic controller (PLC) to provide faster control and enable use for applications requiring real-time control.
Hitachi will continue to develop and release systems with higher added value to meet client needs.
Real-time Rolling Mill Shape Control Using Deep Learning10
DL Control Window N.N.INPUT (32-UNIT) N.N.INPUT (38-UNIT)
N.N.-0 OUTPUT
N.N.-2 OUTPUT CLASSIC OUTPUT
32-U
NN.N.
PC for learning and control monitoring PLC Cold rolling mill (Sendzimir rolling mill)
Steel rolling result data
Shape detectorSheet steel
Shape detector
Reel
++
Reel
OperationoutputPlant
protection
Learnedcontrol rules
Deep learningLearning data extraction
Data for learningDocuments
Learning data Control functionsControl cycle (0.5 seconds)
Learning functions
Control rule updating
Controlrules
Trialoperation
Steel rolling result data gathering
Data for learning
Data for control
10 Overview of deep learning-based real-time rolling mill shape control system
PLC: programmable logic controller
-
32.
In December 2017, Hitachi completed an upgrade of the hot rolling control system PLCs of Outokumpu Stainless Oy in Republic of Finland. Th e PLCs were upgraded to the lat-est model (HISEC04/R900), and the system is now operating smoothly. Th e steelmaker’s equipment is a large system composed of one Steckel reversible rolling mill (a special-steel hot rolling mill used mainly for stainless steel), one high-pressure rolling rough mill in the pre-stage, and three quality-adjusting rolling tandem mills in the aft er-stage.
Th e upgrade involved three months of prior parallel operation testing that allowed Hitachi to upgrade 16 PLCs simultaneously with an equipment shutdown period of just nine days. In addition to upgrading the PLCs, Hitachi also provided a remote maintenance service through July 2018 on a trial basis. Th e service offi cially started in October 2018. It enables the system status to be checked from Japan via a remote connection so that optimal assistance can be rapidly provided. It is designed to ensure trouble-free equipment opera-tion and improve work operation quality.
Hitachi will continue to augment the added value of its IT- and operational technology (OT)-based services to help resolve client management issues.
Hot Rolling Control System for Outokumpu Stainless in Finland11
Steel plant InternetHitachi
Support/analysis team Category
Hardware
PLCteam
Support center
Data-gathering terminal
Middleware
Systems/software
PLCs
Real-time operating systems
System management
PIO
Communication protocols
Applications
Networks
HMIdevelopment
AGC
Specialist technology areas
PLC systemAnalysis data
Feedback
Remote maintenance
terminal
11 Remote service illustration
HMI: human-machine interface PIO: process input/output AGC: automatic gauge control
/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages false /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 144 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages false /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects true /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (Japan Color 2001 Coated) /PDFXOutputConditionIdentifier (JC200103) /PDFXOutputCondition () /PDFXRegistryName (http://www.color.org) /PDFXTrapped /False
/CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > > /FormElements true /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles true /MarksOffset 0 /MarksWeight 0.283460 /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /UseName /PageMarksFile /JapaneseWithCircle /PreserveEditing true /UntaggedCMYKHandling /UseDocumentProfile /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ]>> setdistillerparams> setpagedevice