Predictive Maintenance 

Phy­si­cal main­ten­an­ce can lead to cos­t­ly inter­rup­ti­ons in the manu­fac­tu­ring pro­cess. By means of pre­dic­ti­ve ana­ly­tics, repair and main­ten­an­ce can be sche­du­led based on real-time pre­dic­ted fail­ure pro­ba­bi­li­ties. 

When cus­to­mers ask us for advice on pre­dic­ti­ve main­ten­an­ce, we tell them that Pre­dic­ti­ve Main­ten­an­ce is sim­ply regu­lar main­ten­an­ce stra­tegy that is impro­ved by Data Sci­ence.  

The­re is a strong chan­ce that this is more obvious for you than us, but you can expect that impro­ve­ments to your main­ten­an­ce stra­tegy lead to redu­ced down­ti­mes and more robust pro­duc­tion, lower main­ten­an­ce cos­ts, bet­ter under­stan­ding of relia­bi­li­ty (regar­ding for exam­p­le sup­pli­ers) and mini­mi­zed war­ran­ty cos­ts. 

What is your cur­rent stra­tegy? 
Under­stan­ding the cur­rent stra­tegy helps to iden­ti­fy what is nee­ded for an impro­ve­ment. The dif­fe­rent levels of main­ten­an­ce stra­te­gies can be descri­bed as fol­lows: 

  1. Reac­ti­ve / Heu­ristic → No data, run to fail­ure 
  2. Pre­ven­ti­ve / Plan­ned → Sche­du­led main­ten­an­ce (by time or usa­ge, only) 
  3. Con­di­ti­on-based / Sta­tis­tics-based → Moni­tor con­di­ti­on (sen­sors) 
  4. Proac­ti­ve → Root cau­se ana­ly­sis to mini­mi­ze fail­ures / eli­mi­na­te defects 
  5. Pre­dic­ti­ve → Pre­dict fail­ures befo­re they occur 
  6. Self-Opti­mi­zing 

When you have iden­ti­fied whe­re you are with your stra­tegy, you pro­ba­b­ly also have iden­ti­fied the cor­re­spon­ding data and tools available for it. 

What is the next step? 
You can begin to plan what is nee­ded to move to the next level. This can requi­re new data sources or more data, new tools and the intro­duc­tion of new metho­do­lo­gies.  
That can lead to simp­le methods (from Data Sci­ence) being the most sen­si­ble next step. Simp­le methods have the advan­ta­ge of being easier to use and they are (in gene­ral) more com­mon­ly available. We sug­gest to aim for the low han­ging fruits first. 

What is Data Sci­ence? 
Data Sci­ence is the coll­ec­tion of methods (sta­tis­ti­cal, machi­ne lear­ning, AI) that help impro­ving the main­ten­an­ce stra­tegy. The sel­ec­ted method might stay as simp­le as being “only” impro­ved data inte­gra­ti­on and pre­pa­ra­ti­on, visua­li­sa­ti­on, busi­ness rules, rudi­men­ta­ry sta­tis­tics and/or con­ti­nuous moni­to­ring but can as well include advan­ced tech­ni­ques from advan­ced sta­tis­ti­cal methods, data mining, machi­ne lear­ning and AI, Pre­dic­ti­ve model­ling and model manage­ment if nee­ded. 

What methods fit your stra­tegy? 
In any sce­na­rio you should deploy the tools to inte­gra­te your data, to clean it and to have a pro­per way of visua­li­zing it. 
For the dif­fe­rent levels of stra­tegy some methods fit more nice­ly than others. Refer to this non-exhaus­ti­ve list of popu­lar methods: 

  1. Reac­ti­ve / Heu­ristic 
  • Cost-based ana­ly­sis (e.g., com­pa­ri­son of stra­te­gies) 
  • Struc­tu­red log­ging of events, reac­tions, and reasons 

2. Pre­ven­ti­ve / Plan­ned 

  • Auto­ma­ti­on via busi­ness rules 
  • Eva­lua­ti­on and opti­miza­ti­on of cur­rent plans → Base­line for fur­ther impro­ve­ments 
  • Relia­bi­li­ty / Sur­vi­val Ana­ly­sis, Expec­ted Life­time, Wei­bull, Cox Regres­si­on 

3. Con­di­ti­on-based / Sta­tis­tics-based 

  • Real-time Moni­to­ring, Aler­ting and Noti­fi­ca­ti­ons 
  • SPC (Sta­tis­ti­cal Pro­cess Con­trol)  
  • Advan­ced Visua­liza­ti­on 

4. Proac­ti­ve 

  • Mul­ti­va­ria­te Ana­ly­sis, PCA, Clus­te­ring 
  • Sta­tis­ti­cal methods: Cor­re­la­ti­on, Time Series Ana­ly­sis, Regres­si­on Model­ling, Chan­ge­point Ana­ly­sis, MASS, Root Cau­se Ana­ly­sis 
  • Fea­ture sel­ec­tion, reduc­tion, and engi­nee­ring 
  • Advan­ced Model­ling (DM, ML, AI) to unco­ver hid­den inter­ac­tions 

5. Pre­dic­ti­ve 

  • Pre­dic­ti­ve Model­ling, ML, AI 
  • Model Deploy­ment, Update and Exe­cu­ti­on: on Device (IoT), as Ser­vice (SOAP), Batch-ori­en­ted 
  • Model Manage­ment 

6. Self-Opti­mi­zing 

  • Moni­to­ring of the (pre­dic­ti­ve main­ten­an­ce) sys­tem 
  • Automatic/Continuous model opti­miza­ti­on 

The Pro­ject 
As you can see, pre­dic­ti­ve main­ten­an­ce can be quite easy. It all depends on you and your envi­ron­ment. A simp­le pro­ject to use pre­dic­ti­ve main­ten­an­ce starts with the assess­ment of your cur­rent stra­tegy, iden­ti­fy­ing all its com­pon­ents and then deci­de what the next level of stra­tegy should be and what the gap is that needs to be clo­sed.

Latest News
Your contact

If you have any ques­ti­ons about our pro­ducts or need advice, plea­se do not hesi­ta­te to cont­act us direct­ly.

Tel.: +49 40 22 85 900-0

Sasha Shiran­gi (Head of Sales)