Real-Time-Data Analytics in Raw Materials Handling

144 Real-T ime-Data An alytics in Ra w Materials Handling Christopher Roths chedl, Roland Ritt, Paul O'Leary , Matthew Harker, Micha el Habacher, Michael Brandn er 1 Introduction must abide by the laws of phys- ics must inverse problem mining of sensor data causality 145 2 S y stem Premiss data information knowledge understanding wisdom Fig. 1: The process behi nd the data analy sis system. 146 metadata context significance knowledge human-machine interaction understanding Advanced Symbolic Time Series Analysis (ASTSA) verbs nouns Adverbs adjectives Punctuation Engineering f eedback 3 Data Ingestion 147 Fig. 2: This illustration show s the main processes of da ta ingestion. The top section corresponds to the machine or plant on which data is bein g collected, w hile the bottom part represe nts the data center located at a different location. The data is provid ed in several f ormats afte r it has been in gest ed . 148 Fig. 3: Three single days o f data are assemble d to a contiguous data stream. The il lustrated con- tiguous section cor responds to the time portion a ship loade r needs to load a vessel: this enables ev aluations based on time ranges that are significant to particu lar fields o f inter- est. 149 4 S y stems Currently Being Monitored Fig. 4: Exampl es of four systems that are currently being monitored u sing the described ap- proach: a) ship loade r, b) mobile sizin g rig, c) bucke t -whee l excavator, d) bucket-whe el reclaimer. The senso r channels of these systems are moni tored with a sampling interval of 1s. (Sources: (a ) – http://www.flickriv er.com/photos/impalatermnals_i mages/ 17557941415/, re trieved on 2016-02-08 ; (b), (c), (d) – Courtesy o f Sandvik.) 5 Exemplary Da ta Evaluations Condition Monitoring: Commissioning: 150 Fleet management: Automatic Operations Recognition: Incident Analysis: Logistics Op timisation: 151 5.1 Incident Analysis Fig. 5: This example of incide nt analysi s shows data for a time period of two months, acquired with a sampling ti me of 1s. Each vertical line corresponds to an ev ent; 63 even ts w ere found in total by using Advanced Symbol ic Time Series Analysi s (ASTSA). Every event corresponds to an in appropriate ope ration of the machin e: the data can be zo omed in on automaticall y f o r every single event to perfor m local analysis, i.e., in the se conds and minutes right befo re the occurrence of the event ( see Fig. 6). Fig. 6: Plots of the identi fied e v ents with 1s resolution for three o f t he 63 events reported in Fig. 5. 5.2 Long-Term Logistics Optimisation 152 Fig. 7: Polar histogram o f loading on the slew be aring of a buc k et-w heel reclaimer. The data has been aggregated w ith a sampling time of 1s ov er an observation peri od of one y ear. 6 Conclusions REFERENCES 153